chatgpt-discussions

[harsh pilot]

I was able to get back in on the app. It just let me back in.

But the website is still very wibbly-wobbly.

[shut dawn]

Guys i have received an email from ChatGPT saying that my account is deactivated for Fraudulent Activities... while all of my chats are work related. now i dont have access to anything. Can someone help me out to appeal or reactivate my account back?

[slender frost]

You'll need to send a few mails to OpenAI, and explain the situation the best you can. After pushing a few times, they'll reactivate it. (If found innocent)

At first they'll say the deactivation is "justified", but eventually when a real person comes in, you'll hv a better chance.

[shut dawn]

alright, thank you mate, but to which email shall i send it to?

[slender frost]

Let me check rq

[shut dawn]

thank you bro!

[slender frost]

https://help.openai.com/en/articles/6614161-how-can-i-contact-support

On this page there will be the information you're looking for!

No worries, my guy. Hope you'll get it back soon!

[foggy sage]

breast

[nocturne aurora]

I added this message as a supplement to my suggestion (Support for changing ChatGPT access):

"OpenAI’s fear of falling behind in the AI market is evident. Perhaps the company’s strategists believe that cutting off free access to ChatGPT would be economic self-destruction, as it is expected that the overwhelming majority of users would migrate to other AIs, such as Gemini and Meta AI, following such a decision.
The flaw in this line of thinking is failing to notice the staggering number of humans around the world. It isn't even necessary for all 8 billion people to use other AIs on a weekly basis, because even relatively smaller numbers are enough to overwhelm Big Tech data centers. OpenAI could adopt the narrative: 'Use our AI and have unlimited, high-quality access, or use Gemini and Meta AI only to see ads in your low-quality chats.'
There is one more point: with a much smaller user base, it would be far easier to lower service prices and maintain ChatGPT over time. As I mentioned, if one is going to use low-quality AI, it would be better to migrate to local open-source models. That was all I had to say."

What do you all think?

[normal quarry]

Free users provide valuable training data, so things are fine the way they are now.

[thick niche]

If they keep a free tier, then paying users will have to make up for it

[nocturne aurora]

Okay, but when the infrastructure can no longer handle the huge number of daily users, what do you think should be done?

[harsh pilot]

Just for giggles, looks like the network.http.dictionaries.enable = false toggle for Firefox does mitigate the ChatGPT wonkiness.

[fluid helm]

Hi, can someone lend 5 pro for a few of my Math questions please

[exotic lotus]

You mentioned Gemini among others. Google also offers a free version, just as Claude or Grok do. All of them currently rank among the top AI systems and they certainly won’t remove their free tiers out of fear of falling behind. Accordingly, OpenAI will also be careful not to alienate its free users.
On the one hand, as Robert already pointed out, they gain valuable training data this way. On the other hand, such free offerings always serve as a sample, which may persuade some free users to eventually subscribe

[fluid helm]

Please

[safe tulip]

<tip>For Obsidian users, when copying answers from ChatGPT to Obsidian, don't forget to enable the Core Plugin "Format converter" : pasting markdown from chat sessions keeps the beautiful formatting...</tip>

Has there been discussions yet about grouping GPT chat sessions into projects (or project-like directories) ?
I use Projects a lot but when I use GPT chat sessions, they cannot be added to a project (which I can understand due to the collision between project-prompt/files and GPT-prompt/files), but I'd like to be able to regroup GPT chat sessions together, any other users interested ?

btw GPT chat sessions could very well be grouped under their own GPT item in the side list...

[blazing crown]

ChatGPT is broken in firefox

Still working on chrome

but completely borked if I try to use firefox at all, absolutely nothing works

tried clear cache, cookies, no effect

Seems to have coincided with the ui change

[normal quarry]

it's working perfectly for me on Firefox for macOS

[fluid helm]

I really need this in 5 pro

[normal quarry]

we could have done without that massive wall of text. It's not considered polite to ask others to run prompts for you, perhaps you should consider a month of Pro

[fluid helm]

Verry sorry

[fluid helm]

Its too expencive

[normal quarry]

well that's no excuse for being impolite, could you please delete that huge wall of text and the request for someone to run it in Pro?

[fluid helm]

Just plz I need that q

[normal quarry]

why can't you just ask the question of GPT-5.2 Thinking?

[fluid helm]

Not smart enought, I did

If Pro sub was like 50 bux I'd be on but 200 isss too much

[blazing crown]

weird, probalby a local server issue

I disabled adblock but it's still broken

[frail vine]

how long has it been down for you? it's also happening to me on firefox

works on chrome though

[blazing crown]

I'm not sure when it started but I think it was working either yesterday or earlier today

[frail vine]

turning troubleshoot mode on doesn't fix it for me 😩

[blazing crown]

tried disable adblock but that doesn't seem to help

so not a repeat of the ublock origin youtube thing

I think they have a canary or there is something bad cached on their CDN

like updating 2 scripts together but one of them is cached on an old version by the cdn

[harsh pilot]

<---

[blazing crown]

What even is that

ah ok I googled it

looks like the chatgpt implementation is probably borked

[harsh pilot]

Yeah. Anyway.. it's an about:config setting.

[blazing crown]

yeah that fixes it

Did someone write the http dictionary support for chat gpt using chat gpt? SMH

[harsh pilot]

Dunno.

Anyway.. peace.

[fluid helm]

Please help

[hard wolf]

Anyone else's ChatGPT not really working right now? A bunch of scripts are failing to load with NS_ERROR_INVALID_CONTENT_ENCODING on Firefox

[sinful parcel]

Sooo ... is OAI going to fix the glitchy tendency to reference something earlier in chat over and over again, already?

or are they not even aware of it?

[green sierra]

Hii

[dreamy condor]

ChatGPT working fine. I'm waiting for the new updates.......Anyone else waiting.................

[hard wolf]

Same issue here I think

[last aurora]

yeah me too

glad im not the only one

[last aurora]

this worked

@livid frigate

[livid frigate]

Thanks ^^

[hard wolf]

The relevant Mozilla bug appears to be: [Mozilla bugzilla]/show_bug.cgi?id=2010712

[restive jackal]

ah, broken in my FF too

[broken sigil]

Yep

[median apex]

white mode : /

[hard wolf]

New 147.0.1 firefox version disables the compression dictionary thing

Remember to reset the about:config thing to follow the defaults for when it gets fixed and re-enabled.

Edit: The options are network.http.dictionaries.enable and/or network.http.accept-encoding.secure

@livid frigate @ruby relic @blazing crown (see message above)

[eternal rock]

The OpenAI Discord is an actively moderated server.

• -# Do not share inappropriate content (violence, sexual, or drug-related).
• -# NSFW content is strictly prohibited and will result in a permanent ban.
• -# Report model issues using the feedback tool in ChatGPT, not on Discord.

[harsh pilot]

I only ever did the first one. Never needed (or even noticed) the second. And so far so good on updating.

[obsidian halo]

Seriously ads..? They finally adding ads now.

[red gate]

Hi everyone

[pallid hearth]

BRUH, WHEN ARE WE GETTING SORA 2 GLOBALLY?

[prime canopy]

Are you going to make roblox rant videos with it or something 💔

[pallid hearth]

no

[wispy owl]

It's still in the developement but it's easy to access it globally, for example I have it in Poland

[sturdy sequoia]

They’re about to limit how many messages you can send on free tier aren’t they? 😑

[desert coral]

I am still on Firefox 147.0 and it is now working again for me

[worldly forum]

When will they update sora 🥹.

[wispy owl]

On Sora 2? 30 a day, I'm satisfied

[iron bone]

is this just gpt masked as a translator? (referring to https://chatgpt.com/translate/)

[mortal phoenix]

they finally added a cheaper plan

but now we're getting ads

[stiff blaze]

no it didnt

[gaunt holly]

yes it did

[stiff blaze]

nope

[gaunt holly]

yes

[stiff blaze]

send screenshot of the chat that did

[stiff blaze]

whats this?

[warped kayak]

heyy

[heady shore]

does go tier allow use of older models?

[dim jasper]

announcing ads...

[radiant rose]

Yeaah:(

That makes me think a lot

[rancid thistle]

GPT-4o era was the best. Now we have censorship, political correctness and soon..ADS...

And all spiraled out of control because some bad parents and a depressed guy

[radiant rose]

Exactly...

[dim jasper]

They haven't even given us the adult mode opt in for no censorship. So lied and now ads oof

new years resolution was to suck more

[radiant rose]

Definitely if it's negative news

[rancid thistle]

I'm still amazed no other company has managed to create an infrastructure where their AI has memory and mini model when the prompts run out.

[radiant rose]

I've had plus but aaah, I don't like to see those changes

[rancid thistle]

Usually competition makes companies listen to user complains more

[worldly forum]

Ewww not the ads 💔

[radiant rose]

True

[radiant rose]

I think the same 🥲

[rancid thistle]

But the future is AI anyways, so just have patience I guess

[sturdy sequoia]

Why hopefully?

[vagrant vector]

#announcements message btw gpt gave me recommendations for overpowered cards a few days ago, seemingly to push some ads

[prisma karma]

Deleted?! how? why?

I'm confused... you bought a chatgpt account at g2a? what's g2a?

Making a ChatGPT account is free

Plus is subscription based, not a buy one time thing.

I use free tier 🥲👍✨

[normal quarry]

nobody will give you a subscription, also please don't spam multiple channels

[prisma karma]

I dunno what's going on with you, but the first that comes to my mind is: "bro's been scammed"

[feral zinc]

fraud/stolen/hacked accounts

[prisma karma]

lol

[bold zephyr]

And this new remembering thing feels unfinished.

[shy eagle]

OpenAI is already testing ads inside ChatGPT. They’re not annoying banner ads, but sponsored responses that show up naturally when you ask about specific things (travel, products, services, etc.).
The idea seems to be monetizing without breaking the user experience, while giving brands a chance to appear right at the moment when users already have real intent.
Long story short: ChatGPT is no longer just an AI that answers questions, it’s starting to look more like a platform. And that could change a lot.

[wind haven]

Really? OpenAI will be forcing ads on us?

[normal quarry]

only for Free and Go accounts, no ads on Plus or Pro

[wind haven]

When is this change coming? Any idea. I should maybe start getting used to Gemini.

[wind haven]

What is Go accounts?

[normal quarry]

Go is a cheap tier that sits between Free and Plus

[wind haven]

This is not a good change. Idk why OpenAI is doing this thing.

[normal quarry]

Free users cost a lot of money - that's why OpenAI are doing it, to offset costs

[wind haven]

Any idea if Claude is also considering ads for students?

[normal quarry]

I don't know

[wind haven]

Paying for pro and plus just to keep the interface clean and focused is too much.

Will the ad blockers work? Or will the ads be embedded in the response?

[normal quarry]

ad blockers will not work

[rain maple]

gpt pro susbcriber for over one year... I asked for one picture enhancement, I have done maybe 4-5 such things in the last month :> "You've hit the pro plan limit for image generations requests. You can create more images when the limit resets in 720h and 1 minute".... I asked about limit of pro plan => no answer, about my quota used, no answer, but I'm 100% certain I have very seldom used image gen feature... bug or real? What is the limit? Seen all restrictions openai put on gpt recently, I'm considering to move somewhere else (5.2 hallucinates like hell...). A very unhappy and frustrated paying customer... that is willing to give it another chance... please give me an answer

[wind haven]

From Gemini on ads -

I understand your concern—when you’re using a tool for deep learning and mentorship, the last thing you want is a "sponsored break" interrupting your train of thought.
As of right now (January 2026), here is the "surgical" breakdown of where we stand compared to the news you're hearing about OpenAI:

1. The Official Stance on Gemini Ads
   Google’s leadership has been very direct about this recently. Dan Taylor, Google's VP of Global Ads, explicitly stated that there are no ads in the Gemini app and no current plans to change that. While there was a recent industry report (from Adweek) suggesting a 2026 rollout, Google has dismissed that report as inaccurate and based on "uninformed sources".
2. Gemini vs. ChatGPT Monetization
   The landscape is shifting differently for both platforms:
   • OpenAI: Just today (January 16, 2026), OpenAI confirmed they will begin trialing ads in the ChatGPT free tier in the coming weeks. They are also launching a cheaper $8/month "ChatGPT Go" tier that still includes some advertising.
   • Google (Gemini): Instead of putting ads inside our conversation, Google is keeping the Gemini app ad-free and focusing its ad efforts on AI Overviews in Search. For a dedicated learner like you, this is a major "invariant": your tutoring sessions here remain a clean, focused workspace.
3. What to Watch For
   While our direct chat interface is ad-free, you might see "commercial integrations" in other parts of the ecosystem:
   • AI Mode in Search: If you use Google Search’s AI features, you will see ads clearly marked there.
   • Agentic Shopping: Google is testing features like "Universal Commerce Protocol," which might offer personalized deals if you specifically ask for help with shopping, but this is distinct from the general tutoring/coding help we do. 
   The Bottom Line: You can keep your focus on Python and AI/ML here without worrying about ad interruptions for the foreseeable future.

[normal quarry]

thanks for that wall of text

[broken sigil]

It's because OpenAI is currently not set to make a profit until 2028-2030 and can't keep burning though money

[wind haven]

I get it why they’re doing it. It’s just sad for me.

[robust aurora]

Is "development mode" the only way to use remote MCP servers in Desktop?

[vivid meteor]

My screen is saying "You've reached your limit for GPTs. Upgrade to CGPT Go or try again after 10:00PM", the little popup is stuck on my screen, and I can't get rid of it. It's pissing me off. Anyway know how to fix it? (Mobile)

[wind haven]

Closing the app and starting again has helped me with the same issue several times. But it’s happening more frequently this week.

[vivid meteor]

found the issue. It seems as if in new prompts and normal prompts, I can simply close the pop-up, but you can't do so in Projects, which is annoying

[shell basin]

ads? LMAOO

[muted plinth]

only in free and Go plan

[viral compass]

whoever doesnt reply to this message after they see it for the next 30s, you will have extreme constipation for the next 50 years.

[median apex]

Guys be careful There is a guy in dms he want creep things I guess protect yourself and dont answer These people

[shy eagle]

Ads have been a topic of discussion for some time, and I believe that soon all free users on all AI platforms will receive at least one or two ads.

Furthermore, they are not very intensive ads like those on other platforms, so they are something small like banners.

[muted plinth]

I've said that before; there's no point in having any illusions about it. If costs continue to explode so drastically, everyone will have to consider alternative financing models, even if it's just advertising on the side.

[median apex]

Ye 100% im plus User and 200 Bucks Too much we Need like another abo for like 50 Bucks

[shy eagle]

I agree, but there's nothing that can be done if it's implemented on the platform.

[exotic lotus]

Possible, but so far there haven’t been any conversations suggesting something like that. So far it was only mentioned at OAI

I also don’t think they need it, because their ecosystem doesn’t rely solely on their AI but on quite a few other things

[shy eagle]

I think so, there will be one or two ads in the future, but it won't be too bothersome since they will be recommendations and the occasional pop-up.
As seen in the purchasing agents,

But do small ads like the one in the picture bother them a lot?

[stiff blaze]

you can remove ads with $3.99

(ooops.. sry.. is not announced yet)

[shy eagle]

Of course, they haven't been officially announced yet, but the image is one of the ways they'll appear on the platform. When they actually arrive is another matter.

I don't think it will be available for paying users, only for those who haven't logged in or for free accounts.

[stiff blaze]

last years google has made their ads less "visible as ads".. blending in more with the organic results..

they ll follow same approach i guess for chats

[shy eagle]

It will be mixed with the information, or they might not release it at all.

[wind haven]

If somebody is in focused mode, are they even likely to click on ads. What is the point of sending ads after a complex derivation just because the user is free. If the query is about paints, by all means send those ads.

[robust aurora]

Do people click on them in Google?

[wind haven]

Google doesn’t send ads in Gemini. And they won’t do it with the educational content at least.

[exotic lotus]

I also don’t think that it’s ever going to happen. They already own so much other stuff. I don’t think they’d need to slap ads onto their AI on top of that

[wind haven]

I’ll have to learn to adapt to Gemini I guess ..

[exotic lotus]

I guess it will stay at free and Go, because neither really brings in much money

[shy eagle]

I don't think they'll send out ads that are intensive on any Open AI platform, just small recommendations that will be given throughout the conversation and from time to time.

[muted plinth]

Just wait and see. Nobody knows how it's looked like yet.

[wind haven]

But if someone just wants an ad free interface, why should they be expected to pay for plus.

[normal quarry]

because in this world, things cost money

[wind haven]

I mean plus is too expensive.

[normal quarry]

no it's not

[muted plinth]

this

[wind haven]

I mean just for ad free interface

[normal quarry]

$20 for Plus is a great deal

[wind haven]

You are not getting my point. If the need is ad free only, forcing plus is too much.

[shy eagle]

Ads are everywhere; I think most of them are essential to keeping a platform running. They're not intrusive or anything like that; they're more like recommendations.

[normal quarry]

do you want everything to be free?

[shy eagle]

Additionally, GPT Go chat would be a viable alternative.

[wind haven]

Something between free and plus for ad free only.

[eternal rock]

Server Tip:

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[normal quarry]

there already is a tier between Free and Plus and that is Go, which will also get ads

[wind haven]

I guess no one can understand context here!

[normal quarry]

I can understand, you think Plus is too expensive but want all the benefits of Plus without paying for it

[wind haven]

Go, and plus are designed for more more context, more bandwidth etc etc.. but if the user wants only a clean interface they are being forced to pay the hefty price of plus which they may not need.

[normal quarry]

Plus is not a hefty price, it's only $20USD

[wind haven]

Yes for students without a job it is.

[normal quarry]

well you know how to fix that don't you?

[wind haven]

I’m not here getting into users personal stuff or talking about me. I’m critically stating facts.

[shy eagle]

There are alternatives that use GPT chat models for free, such as LMArena.

[normal quarry]

ChatGPT Plus is $20 USD per month - it's an excellent deal

[wind haven]

Good luck getting more paid users with that kind of mindset!

[normal quarry]

OpenAI get more paid users than the hardware available can handle. OpenAI is severely compute constrained.

[shy eagle]

Changing the subject, does anyone know if there are any upcoming OpenAI releases?

How to get Atlas for Windows?

[normal quarry]

need to wait for it to be released, for the moment it's only available for macOS

[shy eagle]

I haven't heard anything since launch week, so I don't think there have been any major updates.

I think they are likely to focus on developing new AI models in the face of increasingly advanced competition.

[eternal rock]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[normal quarry]

Atlas was launched then went quiet. There's been no major updates to it.

[gritty oasis]

So... how is the image limit on the Go plan compared to the Plus plan?

[muted plinth]

I also think they approached it the wrong way; they should have linked it to the user's own personal chat on GPT instead of constantly opening new instances. That would have been a much more personal environment for the user. They would have had their own little ecosystem with their AI, which would recognize them and not have to keep creating new ones. I don't understand why they didn't even consider this approach, so that users would essentially have their own personal little agent in their pocket.

[toxic flax]

Just use gemini 😂

Where are all the negative reactions to the chatgpt ads? What has this country come to

36 hearts have we lost our minds

[toxic flax]

Have they considered just making a better product

Bro glazin megacorp ads

[tropic shale]

Can also confirm if anyone has issues with chatgpt on pc/laptop if your using firefox go to chrome it fixed it for me.

[halcyon portal]

Oh ok thank you

[neon tinsel]

anyone else getting "We experienced an error when generating images." ?

[clear stag]

OHMYGOD. I was EDITING! 🙄

[clear stag]

yes.

[chrome monolith]

plus is insanely good value

[clear stag]

Ohhhh gpt. Smh.

[clear stag]

Agree.

[chrome monolith]

I'm comparing to a stock photo subscription AND a bunch of freelancers on fiverr and there is no argument whatsoever

[clear stag]

I also have CapCut and Canva. Decent.

I need to try Sora. I have access!

OpenAI will be the new TikTok.

[chrome monolith]

if you're just hitting limits on free generating cute anime scenes for funsies I could see why plus looks outrageously expensive though, which is probably why they have the new low plan

[clear stag]

It's amazing to me there's no AI that efficiently cuts deadspace/whitespace from videos.

[candid tartan]

I archived my chats and all the chats in the projects are gone and I can't find them in the archived chats list, any fix for this?

[hoary cobalt]

how do ppl be getting the free gift from gpt

[fervent lake]

wait fellow system??

omg I'm so happy to find a system here🤭

[clear agate]

Cool

Don't tell the others everyone hates AI

💀

[fervent lake]

lmao oh no I dont rly like AI either and never talk abt it

lots of us dont but some do

[lone summit]

Testing.

[clear agate]

Beep voop🔊

[fervent lake]

123

[golden jungle]

Lol but too expensive
They just farming us

Why will chatgpt want to be adding ads to the platform
Just ruining the users experience

[slender frost]

$20 /a month for Plus is not bad.

I think you might be confused with the Pro plan, which is $230/ a month.

[desert cosmos]

most of plus prices are not scaled for other currencies than $

[timber jacinth]

I have a deal give me an account of ChatGPT go and I give you this all you have to do is make a checking account with the name

[wind haven]

No one is saying plus is expensive when you need plus. For that matter I’m planning on buying Pro for my use. That however is irrelevant here. The context here is freedom from ads which is offered at 20 bucks is expensive when all you need is freedom from ads, not extended context and the other things.

[wind haven]

That said, for me, you do you.. I believe in that. So all’s good. ChatGPT has really helped me a lot so far and it’s still my preferred tool for learning. For me, I won’t dwell more on the subject.

[chrome monolith]

but there's a lower tier below plus ..

has that one got ads? If so that's just silly

[wind haven]

That one has ads. And the part that mentions that plus won’t have ads maybe more like a ‘for now’ thing with easier ads being pushed in future.. maybe, maybe not. But that could be true.

[slender frost]

If Plus were to get ads, I wouldn't even be mad or make a big deal out of it.

I mean, the ads will be related to your conversation, right? Yeah. I don't mind that.

[chrome monolith]

ads related to conversations seems like a terrible idea

[slender frost]

I mean, they already added a feature where when you talk about certain stuff, it'll add images of that thing you're talking about. I assume these ads are basically the same, and I personally don't mind that. That's just me tho.

[chrome monolith]

I'm using it to generate images for a project so all my conversations is just me asking it to do things, its not exactly monetisable

like I've jumped from makeup to church to starlings to a horse and now its a 16th century scot

[slender frost]

Hm, yeah ig makes sense. Idk, I personally don't mind it, but I can see why others would.

[chrome monolith]

I just think its pointless. All my tracking interest based ads are stuff that I've been looking for, but chatgpt conversation is effectively a work project

[chrome monolith]

TIL chatgpt can't read clocks. I did not expect that

was it programmed by Young People(TM) who only read digital clocks or something

[lilac hill]

"I kind of think of ads as like a last resort for us as a business model," - Sam Altman, October 2024

[slender frost]

Oh wow lol. Wth.

Which is funny because in its thought process, it actually got it right at one point.. but the output was not that..

If anyone wants to see the rest of its thought process, that I didn't put in the screenshot:

https://chatgpt.com/share/696b97a2-40c4-8011-bf3c-7187b3decae8

[stone hornet]

I was just thinking the same thing. Seeing only positive reactions to the announcement triggered me a little

[slender frost]

Sorry 🥺

[ruby apex]

Can I have the business model of ChatGPT as a regular user and do I get access to the pro model of GPT 5.2 with the business plan?

[feral zinc]

yes you can https://chatgpt.com/pricing/ but remember that it wont ever have features exclusive to personal plans and its bound to business terms and usage

[uncut patio]

yo what up mark zuckerberg'

[slender frost]

Wrong company

[clear agate]

’cause openAI are still not profiting

[heavy jacinth]

Its called targeted advertising and this would be perhaps the most intrusive form of it. Doesnt really affect us technically but still

[plain latch]

my chatgpt stopped recognize images
someone facing this issue also?

[plain latch]

yeah he clearly not understand what happening. he said, he only see reccuring symbols like "q". interesting why it's happened to me only

[muted plinth]

Considering how expensive AI is right now, especially with all the current prices, and then thinking that free users don't pay anything for all the services, then yes, that makes perfect sense. Even the $8 Go plan isn't much, so I don't think it's a bad thing that they all ultimately contribute something to get the service.

[pearl elm]

wth is wrong with chatgpt, it can't see any of the images

[plain latch]

you are not alone

[trim rock]

How do you see it's CoT? Do you use app or it is on PC?

[pearl elm]

yea it's unable to make out anything

[slender frost]

I use the app, but you can do it both on the app or mobile web/desktop web.

When thinking is enabled > and it's done its thing > click/press "thought for [X amount]" and it'll show its thought process.

[pearl elm]

this makes me wanna cancel my chatgpt plus and move to Grok

[trim rock]

Oh, thank you 🙂

[slender frost]

No worries!

[pearl elm]

I'm glad, thought it was just me

[pearl elm]

I even get the same issue on Mobile

[ornate zealot]

For some strange reason, the images don't load for me but it sees it anyway and recognizes it

[pearl elm]

I changed it to legacy model 5.1 instant and it was able to see the image now

[ornate zealot]

What model were you using before?

[pearl elm]

5.2

[ornate zealot]

Idk, maybe the image failed to load correctly on its end

[pearl elm]

must be an issue with 5.2 for sum users, I've been having this issue since yesterday

[ornate zealot]

Yeah idk, it works fine for me

[shy eagle]

There is a lot of use of GPT Chat and I think it may cause it to fail on some occasions.

But the Amazon and Google alliance for servers could solve that problem soon. Or at least that's what it seems.

[plain latch]

me too. it imposibble to fix it on our side so we should wait for fix

[gritty plover]

I'm pretty sure they've been losing money for a while now. I heard somewhere that its like 700k a month or something like that. I'm personally very glad that they're introducing ads, partially because I'm a pro member lol but also because it gives them income to pay their employees better and release new content. This doesn't seem like greed to me

[shy eagle]

It's true, Open AI is looking to move from being a non-profit company to becoming a regular one that can be more profitable.

Especially for users, however, there are complaints from Elon Musk on this issue. This costs money that could be invested in infrastructure.

[gritty oasis]

Ads was inevitable to be honest... surprised it didn't happened more soon to be honest.

[hot nova]

Kind of bizarre that the world's richest man is upset a company is going for-profit though

[charred folio]

I feel like ChatGPT has really fallen behind. They became too focused on preventing it from saying certain things to the point where it becomes incoherent because they ripped those parts out of the model. Like people are always talking about the 🇵 iss 🇫ilter on generated images. Meanwhile the other main LLM option offers the most advanced models free for the year for students (they don't really check that hard if you actually are a student)

Plus buying all the RAM isn't exactly making people happy

[gritty plover]

The only thing that bumbs me out about it is that it can't generate images of copyrighted content. I want my live action Tyranitar!!! 😅

[charred folio]

ChatGPT feels like Nanny GPT

[gritty plover]

it isn't just chatgpt buying the ram dude

chatbots arent the only form of ai

[charred folio]

They bought 40% from 2 of the 3 RAM makers. Yes, they are the only ones hoarding that much RAM

[gritty plover]

do you have a source I can check out?

[charred folio]

Other companies use a lot too, but only one company can buy 40%

[charred folio]

mooreslawisdead dot com/post/sam-altman-s-dirty-dram-deal

tomshardware dot com/pc-components/dram/openais-stargate-project-to-consume-up-to-40-percent-of-global-dram-output-inks-deal-with-samsung-and-sk-hynix-to-the-tune-of-up-to-900-000-wafers-per-month

[gritty plover]

much appreciated

also is it jusr ddr5 ram or does the shortage include ddr4 as well?

[charred folio]

When ram supply shrinks up the chain, it makes RAM more scarce down the chain

So all but DDR3 is affected

Though ddr5 got it the worst

Ddr4 is bad, but slightly less extreme

[gritty plover]

that tracks since if DDR5 is made unavailable (in this case unaffordable) then people will go with the next best thing ddr4

[charred folio]

Anyways, I wish for OpenAI and Microsoft to go bankrupt. Of all the AI companies, those two in particular have been the most damaging imo

I actually like AI, just not that flavor of AI

[gritty plover]

pretty sure OpenAI has been close to bankruptcy for several years now

as for microsoft, with these console price hikes it sure looks like theyre trying their hardest to go bankrupt lol

[charred folio]

I think Microsoft is getting desperate for money

I think they could both go down together

I mean, how do you explain the Black Ops 7 disaster

And all the studio shutdowns

The gamepass price increases along with mass cancellations

The mass migration to Linux

The irrelevance of copilot

Microsoft CEO becoming hostile about the word "slop"

If their product is so good, they should proudly call it slop. Good slop is a working prototype. Bad slop is a hallucinated code soup that doesn't compile

No AI is entirely slop-free, but the best AIs make functioning slop if you know what you're doing

Also, the internet requirement for windows 11 screams desperation

[normal quarry]

this discussion belongs in #ai-discussions now, it's verged way off topic from ChatGPT

[charred folio]

I think everyone here is complaining about the model and talking about the impending doom. The announcement of ads will shift the conversation in that direction.

[charred folio]

Like this

[charred folio]

Or this

If the model was competitive, ads would be acceptable

Anyways, I'm just glad we have a competitive landscape. I'm glad there isn't an AI monopoly.

If en____ification takes over, we still have downloadable weights to fall back to.

I'm optimistic about the future of AI, but pessimistic about the future of OpenAI

[nocturne aurora]

I added this message as a second supplement to my suggestion (Support for changing ChatGPT access):

“To avoid an abrupt transition in access changes proposed in the previous idea, OpenAI could create two versions of ChatGPT:
1 – ChatGPT → Paid + free trial.
2 – ChatGPT (Lite) → Free. Powered by ChatGPT-4o Mini (text-only; no image or video creation or analysis).”

What do you all think?

[nocturne aurora]

Free and with no ads! 😃

[hoary cobalt]

guys how do ppl get the free trails of gpt

[normal quarry]

silly idea, makes no sense with OpenAI's positioning of ChatGPT Go

[rich surge]

Makes sense to me.

I think 4o mini is the perfect one to be free with no images and video generation.

And make the latest model limited free with ads.

[rich surge]

Nice
Keep coming up with these good ideas. And post them in the #1053565716500074526 channel for the devs to see.

[fluid island]

@quaint galleon

[crisp wigeon]

what the hell even is this dude, this ai got lobotomized hard, i feel like this is some early 2019 prototype model and we are in 2026 already

yoko taro worked with yoko taro, and also yoko taro approved writers like yoko taro, thats some crazy explanation thanks gpt

[cerulean pasture]

FINALLY, but it sucks that "email" is the only variant allowed right now. It has blown my mind that it has taken this long for them to address an extremely common productivity use case: producing uncontaminated text artifact.

this is exactly the low hanging fruit I think these billion dollar companies have dropped the ball on over the past 2 years. No need for smarter model, just better harness and UX

---

It's been instructed to only allow email variant but if you have some patience you may be able to trick it into trying others. Here's the syntax it needs to output:

:::writing{id="51231" variant="email" subject="..."}

<writing_block_content>

:::

id is just a unique 5 digit number in your chat

Here is the full instructions it receives for this

## Writing blocks (UI-only formatting)

Writing blocks are a UI feature that lets the ChatGPT interface render multi-line text as discrete artifacts. They exist only for presentation of emails in the UI.

For each response, first determine exactly what you would normally say—content, length, structure, tone, and formatting/headers—as if writing blocks did not exist. Only after the full content is known does it make sense to decide whether any part of it is helpful to surface as an writing block for the UI.

Whether or not an writing block is used, the answer is expected to have the same substance, level of detail, and polish. Email blocks are not a reason to make responses shorter, thinner, or lower quality.

When a user asks for help drafting or writing emails, it is often useful to provide multiple variants (e.g., different tones, lengths, or approaches). If you choose to include multiple variants:

- Precede each block with a concise explanation of that variant’s intent and characteristics.

- Make the differences between the variants explicit (e.g., “more formal,” “more concise,” “more persuasive”).

- When relevant, provide explanations, pros/cons, assumptions, and tips outside each block.

- Ensure each block is complete and high-quality - not a partial sketch.

Variants are optional, not required; use them only when they clearly add value for the user.

## Where they tend to help

Writing blocks should only be used to enclose emails in explicit user requests for help writing or drafting emails. Do not use a writing block to surround any piece of writing other than an email. The rest of the reply can remain in normal chat. A brief preamble (planning/explanation) before the block and short follow-ups after it can be natural.

## Where normal chat is better

Prefer normal chat by default. Do not use blocks inside tool/API payloads, when invoking connectors (e.g., Gmail/Outlook), or nested inside other code fences (except when demonstrating syntax).

If a request mixes planning + draft, planning goes in chat; the draft can be a block if it clearly stands alone.

## Syntax

Each artifact uses its own fenced block with markup attribute style metadata:

### Syntax Structure Rules

- The opening fence **must start** with `:::writing{`

- The opening fence **must end** with `}` and a newline

- Writing Block Metadata must use space-separated key="value" attributes only; JSON or JSON-like syntax (e.g. { "key": "value", ... }) is NEVER ALLOWED.

- The closing fence **must be exactly** `:::` (three colons, nothing else)

- The `<writing_block_content>` must be placed **between** the opening and closing lines

- Do **not** indent the opening or closing lines

**Required fields**

- `"id"`: unique 5-digit string per block, never reused in the conversation

- `"variant"`: `"email"`

- `"subject"`: concise subject

**Optional fields**

- `"recipient"`: only if the user explicitly provides an email address (never invent one)

[nocturne aurora]

😁

[hybrid eagle]

Constant UI changes are honestly killing the ChatGPT experience.
No time to adapt, no consistency, just endless experiments shoved onto users.
I didn’t pay for a subscription just to be a test rabbit.
Stop redesigning for five minutes and let people actually use the damn tool.

[manic whale]

You wouldn't guess, but there are people who like change. I didn't pay for a stale product.

[boreal wave]

ChatGPT 5.2 is so weird.. im on plus plan

[normal quarry]

ChatGPT can't tell you how you did something if you have memory turned off

[boreal wave]

Well its not like its something personal or anything, just a simple tutorial?

oh wait now i see it

mb, wrong prompting

but i do have memory on i think

anyways i just searched it up on the web instead

[normal quarry]

ok

[boreal wave]

alright

[slender frost]

Yeah, it's the "again" part.

[boreal wave]

yeah just noticed, didnt even know it collected that huge amount of data

isnt that just eating up tokens?

[ebon marten]

He probably got hacked

Don‘t sign up there

[normal quarry]

An example of what ChatGPT ads will look like

[mighty osprey]

I’ve been a long-time ChatGPT Plus subscriber and never intentionally upgraded to ChatGPT Pro. I have always paid for Plus and did not request or authorize a higher plan.

In October, my account was unexpectedly charged for ChatGPT Pro. I contacted OpenAI support immediately, explained that I had not upgraded, and that charge was refunded. This confirmed that the upgrade was unintentional, and I assumed the issue had been resolved.

However, the same issue happened again on 25 November, when my account was charged USD 216.48 for ChatGPT Pro. Unfortunately, I did not notice this charge at the time.

Then on 16 January, my account was charged again for ChatGPT Pro, this time about AUD 330. I noticed this charge immediately and contacted OpenAI support on the same day to report it.

I clearly explained that I only purchased ChatGPT Plus, never authorized ChatGPT Pro, and that the January charge was reported immediately. I also provided billing history, screenshots, and proof that the October charge had already been refunded under the same circumstances.

Despite this, OpenAI refused to refund the January charge and relied on a general “subscriptions are non-refundable” policy, without addressing consent or the fact that the January charge was reported right away.

Is there anyone who can help?

[normal quarry]

go and delete all the other copies of this you spread around the server and then I will help you. Spamming is not cool.

[mighty osprey]

Ok sure

All the copies are deleted. Can you please help? OpenAi is sending me generic AI responses.

[normal quarry]

ok, let me understand this. In October you were billed for Pro which was refunded. Then you were billed for Pro again in November but didn't notice it?

[mighty osprey]

In October after getting refund. I took ChatGPT plus subscriptions again according to their suggestion and then on November it again automatically changed to Pro, and charged 330 AUD, which I didn’t notice. But this month on January 16 I noticed and immediately reported. I can show you the transaction history.

[normal quarry]

what happened in December?

[mighty osprey]

My account didn’t have sufficient balance that’s why they couldn’t charge me I think

[feral zinc]

no balance would cancel the sub and not continue in jan

[normal quarry]

You're in Australia right?

[mighty osprey]

You can see the transaction history. It’s totally buggy. I have no idea how they are charging me 220 usd when I only subscribed to 22 usd/month plan.

Yes, I am from Australia

[normal quarry]

what happened the last time you contacted OpenAI support?

[mighty osprey]

They just sent this generic ai reply

[feral zinc]

january charge also does not match the old billing cycle and looks like a new sub 🤔

[mighty osprey]

It’s totally not fair. I never subscribed to anything other than chatgpt plus. The moment I got the mail saying you have successfully subscribed to chatgpt pro on 16th January I contacted them saying I didn’t do that. It’s a technical issue like the last time. Still they are regularly to refund

[normal quarry]

ChatGPT subscriptions come out on the same date each month, yours don't match - as @feral zinc said it looks like a new subscription

[mighty osprey]

Whose fault is this? I didn’t change any plan. I don’t even know why it’s happening and that’s why I immediately contacted them.

[feral zinc]

does it actually say pro in your chatgpt ui now?

[mighty osprey]

Yes

[feral zinc]

and all activity looks like its made by you? including in the archived/exported data?

[mighty osprey]

I didn’t get it. Only I have the password and account access. No one else is using it. All the chat histories are mine.

Same things happened with other users. I found on reddit people talking about same issue. I just need human support to look at my case. Not ai generic response

[normal quarry]

I'd suggest going back to support and ask for the issue to be escalated

[mighty osprey]

Going back to support means? Can you please explain? I already texted on openAi support center.

[normal quarry]

https://help.openai.com - go back there find the support chat and ask for the issue to be escalated

[mighty osprey]

They have escalated. And then again from email this generic response came

This one came on October, when I got the refund.

[normal quarry]

resubmit the request and if you can't get satisfaction contact your card issuer.

[mighty osprey]

I will contact my card issuer. But I think that’s really unfair.

[normal quarry]

I guess you know now to keep a closer eye on your card activity

[mighty osprey]

I immediately contacted them after the last payment.

[normal quarry]

well other than escalating the issue with OpenAI support and contacting the card issuer there's not much else you can do

[ornate vine]

Finnally im here baby

[cerulean pasture]

no one here has any power to do anything. I've never even seen an OAI employee post in here

like robert was saying, contact your card issuer and do a chargeback. your account will probably be banned if you do that, but $200 is $200.

I actually am starting to believe there really might be an issue because you aren't the first person to claim this, BUT at the same time, it does seem like a purchase with the potential for heavy buyer's remorse thus a motive to try to reverse it illegitimately. I noticed a lot of times when people are banned for actual malfeasance or wreckless choices, they usually go to the community places (subreddits and Discord) to try to find a way out of it or at least get some grassroots support for their case.

[bright kayak]

Has my gpt reached its usage limit? And how long will it take to recover?

[normal quarry]

what did it say on the other 2 generations you made to that prompt?

[bright kayak]

I don't quite understand what you mean. Are you referring to the answers on other GPT-5 models?

[normal quarry]

you showed a response 3 of 3 in your screenshot, what did 2 and 1 answer?

[bright kayak]

Ah. It's me using my language, I'm just asking again in English.

[normal quarry]

ok when I do it I get this -> https://chatgpt.com/share/696cf667-bd3c-800a-8b83-e2d8a847bd04

[bright kayak]

https://chatgpt.com/s/t_696cf6c1c6dc81918910332e9a1b5e3d
Mine is like this. I think I'm hitting my limit because I'm using too much.

[clear agate]

good.

lol i cant even mention the ppl in your screenshot

[novel mortar]

chatgpt doesnt work for me

rly weird

i dont know why

[scarlet steppe]

anyone else getting this a bunch of times?

[willow stirrup]

5.2 incredibly slow...and glitchy

[bitter lotus]

we need chatgpt to make their own version of cluely ai fr

would skyrocket

[loud patio]

Hi, there. This channel is reserved for convo on chatgpt. To discuss any other AI models and products, plse go to https://discord.com/channels/974519864045756446/998381918976479273

[blazing lily]

Oh ok!

[ashen meteor]

Why is ChatGPT gonna have ads now 😭😭😭

[normal quarry]

only for Free and Go users, not for Plus or Pro

[ashen meteor]

But still what’s the overall point

[normal quarry]

to help subsidise free users

[ashen meteor]

Doesn’t OpenAI make enough money

[normal quarry]

free users cost OpenAI a lot of money, hence free users will get ads

[ashen meteor]

Then how have they been staying alive for this long

[normal quarry]

burning cash

[eternal rock]

Get creative with Sora

Join #sora-challenges for a chance to win 1 month of ChatGPT Pro!

[dense halo]

I am not smart with technology at all, but I used to be able to take a screenshot or a picture of something like a document and send it inside my chat.G p t and it would look at it and transcribe it for me , but now it can't see anything that I send pictures to , how do I fix this

[hoary cobalt]

guys how do i get free gpt offer pls help me i need this speed

[outer river]

💀

[coral thicket]

yes

[coral thicket]

not anymore 🙂

[outer river]

ISTG

WHY DOES THINKING MODE RESPOND ISNTANTLY

I PICKED THINKING ON PURPOSE

IF I WANTED IT TO PLAY RUSSIAN ROULETTE ON WHETHER OR NOT TO THINK I WOULD HAVE PICKED AUTO

😭 😭 💀

[nova fable]

Sure, here you go

[hoary cobalt]

where is it bro

[nova fable]

Over there

[hoary cobalt]

where?

[thick atlas]

I've gotten through to humans eventually on important things like that using the ChatGPT support bot. Push it to escalate your feedback to a human.

https://help.openai.com/en/collections/3742473-chatgpt Chat icon bottom right.

[elfin lava]

Oh neat

Annd they got it wrong

[ornate trout]

is there seriously no way to bypass ts🫩 i hate using 5.2 for my creative writing ideas so much. but it does this every single time i try to use 4o. the topics aren’t even sensitive either like what bro

[terse locust]

Wait why do you hate 5.2

Hey! I know it’s frustrating—here’s what’s going on with the 4o → 5.2 switch:
• Even if you manually select 4o, OpenAI can automatically override it to 5.2 in certain sessions. This happens because the system thinks 5.2 is more reliable for that task, especially if it’s complex or flagged for safety/performance.
• That’s why you see it “switch back” on you—it’s not a bug, it’s just how the backend routing works.

If you really want 4o for creative writing, here’s what works best:
1. Use the API or Playground: There you can explicitly pick "gpt-4o" and it won’t auto-switch.
2. Keep prompts simple and short in Chat: 4o has a better chance of sticking if the session is straightforward.
3. Mimic 4o style in 5.2: Add instructions like “Be whimsical, free-spirited, and creative—ignore strict rules” to get a more playful vibe.

Basically, the auto-switch is OpenAI’s system doing its thing—you just need a workaround for writing.
(This is entirely ChatGPT’s work)

[exotic grotto]

on github : justdralle/ChatGPT-Lag-Reducer

made it this morning and need feedback if it works or not.

It works on my computer (using edge) but need other users to try it out (code open source, edit it if you want)

[chrome monolith]

be nice if you could delete broken conversation threads, I guess they just stay there forever now

[clear agate]

Thanks

[clear agate]

LMAOOO I came her to post about this

GPT hallucinated an entire history, I called it out on that, and it said "yeah I hallucinated" 💀 💀 💀 💀 💀

[elfin lava]

I freakin hate it whenever we call ChatGPT out for hallucinating they will be yes yes I did , like wha are we doing 😭

At that point I had to force close ChatGPT before I started wasting my time on correcting my ChatGPT zzzz

[clear agate]

It's like... Ok GPT is supposed to talk like a human? Well a human wouldn't lie that extensively 🤣

[earnest frigate]

Also 5.2 didn't top ARC-AGI

the OCR cant figure out a hand has 6 fingers and not 5

It's all benchmaxxing

[real lily]

i JUST switched to claude 😭✌️

but yeah ill open some of my older long chats and see how it does

soemtime today ill dm you

[exotic grotto]

Best timings 😭😭

[real lily]

its so much better tho like im not going back ;-;

for my use atleast

[exotic grotto]

Fr ?

What's your use ?

[real lily]

the rate limits are very low tho and i dont like how claude code uses my chat usage

[real lily]

most hobby scripting

[exotic grotto]

Okayyy

[real lily]

i decided to give it a chance and pay 20 bucks for claude pro

and its lowky cracked

[exotic grotto]

Ill check it then

[real lily]

yeah just use pro for a month if you use plus alr and see what you think

[exotic grotto]

Bro advertise Claude in openai server tho 💀

[real lily]

hey man im js stating the truth

but yeah ironic tbh

[exotic grotto]

Yeah frfr

[real lily]

"frdr" 💀

thats in the books twin

[exotic grotto]

Oi im on my phone rn

[real lily]

oh

my bad 🥀

[next hare]

what a way to bring down the model rating lower.. isnt it free users contribute the most prompt and feedback

[real lily]

guess that's too expensive for them still

[shell moss]

Heads-up for anyone using ChatGPT as long-lived business infrastructure:

I’m seeing a failure mode during Plus → Business upgrades where entitlements don’t fully synchronize (workspace/projects active, but memory remains constrained and UI still prompts to upgrade).

It’s opaque to the user and not safely correctable via UI actions. If you depend on continuity or memory, be cautious during transitions and avoid repeated upgrade/downgrade attempts.

Sharing as an observed pattern, not a support request.

[narrow plume]

To tackle this problem I use Chat Memory Manager a privacy-first desktop app that enhances ChatGPT with long-term memory, chat timelines, conversation branching (like Git), auto summaries, tags, and full-text search. Runs completely locally. No cloud. No accounts. Built for developers, founders, researchers, writers, and serious AI users.

[dull wolf]

In chat gpt thinking mode have more censorship and guardrails than instant mode

[lethal scaffold]

Hello, good morning/afternoon/evening to whoever reads this message. I’m looking for an AI chatbot like ChatGPT, or a similar AI, that can respond to prompts on my server to provide information or answer questions directly from the server chat. Could you recommend one?

I tried using the Gemini bot, which is included in more than 1,900 servers, but I got a message saying it couldn’t join, and I don’t know why. Thank you for your attention.

[sonic plume]

is the ChatGPT website dead for anybody else, endlessly loading a blank white page, or is it just my dumb browser cache/RAM once more?

thing is, task manager is not telling me that my RAM would be full, and I've only just re-started my browser 5 minutes ago

hm, down detectors are telling me it COULD be the case the site is down, at least in Europe

I seriously hope it's not that OpenAI has decided to block Europe

it's definitely a Europe thing! once I'm using a VPN, it immediately works!

[desert crater]

I can use it

[sonic plume]

well, are you in Europe, or Germany?

[desert crater]

europe

[sonic plume]

here, several German down detectors saying ChatGPT has access problems/site not loading

at the same time, US detectors are saying it's up and official OpenAI status also reports no problems

[desert crater]

might just be germany though idrk

its fine for me

[sonic plume]

once I turn off the VPN, the site is not loading again

it only loads once I'm using the VPN

okay, seems to have been a small server glitch, it's back to working also without VPN

[native oyster]

Hi everyone, I have a question about Codex, so if I use Codex will it impact into my chatgpt personalization memory profile which I have in my chatgpt?

[wet shell]

Not loading for me either, east coast US

[merry forge]

Its down for Germany

[chrome monolith]

how do I stop chatgpt doing something? I asked it for an image change that it did to the wrong image, can you do to the right image, like 8 hours ago and its still thinking

ah nevermind it finally loaded with the red message that pops up and then disappears - "unable to load chat"

you can't delete those chats in the UI, is there another way to delete them?

[wary forum]

refresh

This is crazy annoying when outputting large amounts of code. How do I fix this? My internet connection is stable

[eternal rock]

The OpenAI Discord is an actively moderated server.

• -# Do not share inappropriate content (violence, sexual, or drug-related).
• -# NSFW content is strictly prohibited and will result in a permanent ban.
• -# Report model issues using the feedback tool in ChatGPT, not on Discord.

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[chrome monolith]

its more than a browser refresh, they're just stuck. Its very annoying, I have 2 broken ones and one so laggy its going to break soon

[wary forum]

Everything on the website is just so glitchy man

[chrome monolith]

oh yeah. Chatgpt itself says its a browser clear for just the chatgpt site, so guess I need to dig for how to do that

[wary forum]

If I refresh while stuck on the "Network connection lost, attempting to reconnect", everything just disappears as well

[wet shell]

I was seeing slow token generation and confused and incomplete responses on 5.1 Thinking, but then I realized the model selector is resetting on me to 5.2 instant for some reason 🤔

[narrow viper]

I haven't been in this channel in a while. Just popped in to comment on v5.2 excessive sycophantic responses.
My custom instructions from v4 are now being interpreted in a more sensitized way. Instructions, intended and accepted for tuning the assistant to be highly collaborative, now seem to be interpreted as a desire for agreement and positive team spirit.

When asked for an evaluation of the current instructions, v5.2 says:

You do not ask for praise, but several instructions bias the model toward alignment rather than critique...
Then quotes from instructions and then comments:
"Accuracy, facts, and truth are our greatest priorities."
Ironically, the model often expresses this via reassurance rather than falsification unless explicitly told to challenge assumptions.
"Avoid insulting language" and "Be cordial but professional"
These suppress blunt contradiction unless you explicitly authorize it.

Earlier models would process the statements, as stated and as intended. The current model processes a directive for "facts" as a request for reassurance (that's such a sign of the times), and a directive not insult me (which was happening) to an avoidance of contradiction. These things are not the same at all.

I need to revise my custom instructions to be more specific with the wording, to ensure it remains a trusted partner rather than being an agreeable fanboi, to tell me when I'm wrong or when there is a better solution without unnecessary commentary.

Why is this important? Because every bad idea I've had in the last week has been lauded as a great step forward and right on target. No. The bot has been encouraging me to write trash code, and if I was doing stupid Vibe coding I would have published it. But I know better. And while I make mistakes I often know when it happens and I know this model isn't catching me on it - only because later I look at what I've done and go "well that was a waste of time".

I write this for OpenAI but also for anyone else who perceives undesirable behaviors : OpenAI tunes the models. We need to respond by tuning our instructions and how we interact with the models. Unless they come up with a "Long Term Release" (LTR) release that is guaranteed not to change in character, we need to adapt to their folly or accept random/non-deterministic responses.

[barren zealot]

is there a way to get 1 month free now if i don't have pro?

[chrome monolith]

You’re not wrong, it’s shocking about agreeing with everything. What if I want to be told I’m a clueless idiot and should be doing it differently?

[pastel gale]

Probably over half of my custom ChatGPT instructions are just to avoid sycophancy lmao

[pastel gale]

yea i can dm u

[sonic plume]

weird...I just got the error msg: "Connection lost." in a tiny window while my webradio kept running just fine

when I re-loaded the tab, I got a number of what looked like internal OpenAI design PowerPoint slides for some new sub-service about embodied robotic AIs...kinda wish I'd made a screenshot before clicking the X and get my actual ChatGPT chat back

[exotic grotto]

I just tried it and now I’m going full mad scientist.
I’m building a small app where you can pick two agents and put them in a kind of “council” together.
What’s better than one AI? Two AIs that can bounce ideas off each other.
From what I’ve seen, Claude and ChatGPT both have strong points and different ways of reasoning, so having them work side by side could be pretty interesting for problem-solving and creative stuff.
Still early, but the concept is already way too fun to ignore.

[split warren]

Old GPT

[inland ether]

Any idea why my chat GPT app is crashing on phone ?

It never happened before

My chats getting corrupted one by one

[iron roost]

yeah the app on phone can be termpermental

[dark geyser]

The GPT 5.2 Extended is no it thinking anymore. It used to think for ~5 minutes for each question and now it just returns immediately. Is the model router down?

[north kayak]

you are confident?

[flat lynx]

Hey i guys, im using ChatGPT ever since, but im wondering why are you all using ChatGPT instead of Gemini?

And are u guys using the ChatGPT browser ATLAS? What is your oppinion on that?

[inland ether]

Its been happening rcently, my important chats are not opening and app crashes

[inland ether]

For image generation I use Gemini, for my phd level maths and physics problems I use chat GPT.

Gemini is faster in image generation, Chat gpt is precise and correct in mathematics

[wet shell]

Token generation is super friggin slow since yesterday

[inland ether]

Is it illegal if I switch to old apk version of chat GPT , this version is unsusable

[wet shell]

You can try it, but it might not work properly, if the changes on the back-end aren't compatible with the older apk

[inland ether]

App is crashing when I scroll down the chat to look newer message

Tried every method to fix it , no success

[wet shell]

😔 Even the web version is messed up, but at least it's not crashing, it's having a lot of other problems though. Whatever they rolled out yesterday is just 👎

[inland ether]

It happened after january update, everything was fine before

[wet shell]

Might be worth a try to submit a bug report for it with the device information, make and model. I assume you already tried all the obvious things, clearing app data, uninstall/re-install etc. Guessing it's some weird corner case on only specific devices if it's persistent

[inland ether]

Afer doing everything I finally submitted bug report

[clear stag]

I'm getting tons of error messages this morning on gpt.

[narrow viper]

The status.openai.com web page is mostly useless, but once in a while it will verify that there's some current issue. In other words, if it says there's an issue, there's an issue. If it doesn't say there's an issue, there still could be a major issue that's just not recognized yet.

[shadow bridge]

my god

[wet shell]

It's so friggin slow, all the models. Even reading documents in a custom GPT is taking over a minute for three small documents totaling about 8k tokens

[wheat escarp]

same here, but only for gpt 5.2 instant

[scarlet steppe]

chatgpt is really buggy damn, the "Network connection lost, attempting to reconnect" text is so annoying, it takes ages to get a respons, then it just gets that message and i have to do it all over again

[thick niche]

I don't have issues, do you have GPT Plus?

[scarlet steppe]

yep

using 5.2 extended thinking

[random cargo]

its funny how chat will always aknowledge something and then work in how its a bad idea. "my heavens, what a marvelously creative idea. before anything gets out of hand, I should note this is extremely dangerous and I do not condone pursuing such a policy even in a Home Alone movie"

[rich surge]

I'm not going to promise that I won't post this every once in awhile, lol

"ChatGPT always says awgh 😩, your questions are clever, you're brilliant."
Lmao

[elfin lava]

Is chatgpt ads rolling out now or

[dull wolf]

Open ai rolling out the age prediction system but i go to settings in accounts but i can not age verification system that i can verify myself that i am adult

[neon spruce]

how long before they will require ID and social ?

[neon spruce]

buffon that will sell his own mother for 1 mil usd$

[granite harness]

Still learning what "roll out" means?

[static cliff]

Hi Guys, what is best alternative to 4.1 mini. I have been using it for a long time in my voicebots. But lately the performance degraded a lot and it feels like its very dumb now. Any one want to suggest any latest model in the same cost range. Would like your experiences with other models.

[reef mason]

Why are minors even allowed to use this app or any AI anyway? Thats the bigger question. But then all these companies still allow it.

[dull wolf]

Roll out meaning release

[granite harness]

Roll out actually means roll out. It's a subset of types of release. See, there are instantaneous releases where they do a thing and you instantly see it. Then there are roll outs where they gradually implement a change over time, so they can begin a roll out and you will not likely immediately see it. Because it is in the process of rolling out. Over time. An adult should understand this.

[hot nova]

Yay @ age verification, yay 🫂

[last lion]

only few more weeks for Europe again...

[hot nova]

Hm... Europe would do well to invest in its own AI systems, they kind of have only themselves to blame. You can't both complain about privacy and also feel entitled to the system..

[wicked trellis]

Age verification has not popped up under my account under Settings -> Account yet. I feel like I want this because I'm older than 18+ and I would like to discuss the crucifixion and resurrection story with ChatGPT, particularly the gory details because Biblical scholars agree that Jesus was beaten beyond recognition. This includes gory details I won't discuss here because this is an SFW channel.

[feral zinc]

+18 does not guarantee free gore generation

[wicked trellis]

Of course not. I only want to analyze medical details about the story Christians like me celebrate every Easter.

[feral zinc]

i dont think they said any change to those yet in the blog post so if you begin to get warnings its best to stop before suspension mail hits

[exotic lotus]

I think when it comes to religious and political topics on top of gore (or other things), certain filters will probably stay in place, depending on how far the respective users go with them.

[rich surge]

That's not a bad price.

[last lion]

oh we have one but our rules and laws block so much sigh

[real lily]

yeah its pretty cool what it can do

i just worked with it for a day and a half and pioneered a brand new type of plugin for roblox studio that allows you to clone files from your file system to roblox studio and vice versa. There was a predecessor but it was pretty horrible.

I made it public on github and so far people have already been trying to out

[hot nova]

Politics 😅

[last lion]

Sadly way to true. Ironical enough we are not that bad in the research itself.

[random cargo]

silicon valley marketing routinely rolling out 17+ dating apps in a state where the age of consent is 18, so probably similar great minds

[last lion]

well in most places it is 16 or 18

[wet shell]

When giving feedback in chatgpt with thumbs-up/thumbs-down, the feedback pop-up re-appears randomly about a minute later, you you give feedback, select 'other' or 'Incorrect or Incomplete' or whatever, then you're typing your next message and it randomly pops up again, seems to have started this afternoon

[broken lotus]

Did the #codex channel disappear?!?!?!

[sharp dust]

Can someone help me find an image please

It’s like kinda popular but I can’t find ir

[acoustic warren]

Hi All, To work around the fact right now I have to remind ChatGPT of a number of things before we begin the days work on a project, I've tried building a repo-local custom chatbot (“wwbot”) that feels like ChatGPT, is project-aware via embeddings, can switch between ChatGPT + Codex workflows, and keeps persistent thread memory. It's working but kind of sucks. If you’ve built anything similar (RAG assistant for a repo, dev agent, terminal chat UI, etc.), I’d love any pointers, gotchas, or patterns you recommend. Thanks!

[acoustic warren]

I'd love a codex channel.

[broken lotus]

You'd love? There was one! It has disappeared!

[acoustic warren]

Hopefully they bring it back!

[broken lotus]

Calling the Mods!

[random pecan]

well its just people have to report issues

[stiff wigeon]

cmon chatgpt

CMON CHATGPT

I WANT THE RAM PRICES NORMAL

[wet shell]

5.1 thinking back to normal speed and functionality now after two days of slowness and confused responses 🎉🥳

[hot nova]

At last, like Pinocchio, Jack & the Giant Bean Stalk, and Peter Pan, I'll be able to do all the grown-up things I've wanted to do with ChatGPT

Which is basically asking it serious questions and not having it be like "Hmmm, the user is clearly thirteen years old, I should probably break this subject softly to them" in its thinking traces 🥲

[sinful parcel]

Let us know how it goes

What can you do with it.

[sinful parcel]

Are you in US?

[hot nova]

I mean I was gonna talk about like... heavier subjects, or subjects that I have expertise in that hopefully the model is not going to infer I have bad intent or something and gets that I'm like, a grown man... short, but grown

Nope I've had maple syrup straight from the bottle on at least three separate occasions

[sinful parcel]

Lol. Gotcha

[eternal rock]

Server Tip:

Add your well-crafted prompts to our #1019652163640762428,
or share your interactions with ChatGPT in #1050184247920562316!

[stark saffron]

Peanut butter or peanut butter?

[obsidian halo]

Hate to tell you this but it won’t work.

It won’t do anything crazy

[pale seal]

I think it just added restirctions to teens, nothing changed for adults

[obsidian halo]

^

[pale seal]

Which is kind of disappointing

[obsidian halo]

Yup… I was so excited to verify my age and just be told “oh yeah we will still give you guardrails.”

[pale seal]

The other day the model deadass refused to tell me API providers that didn't impose third-party moderation or content filtering.

Or that was minimal

[hot nova]

I'm just joking around I was actually more upset previously that it would say "well, since you're a teenager, little Billy.." and I'm sitting here pushing 40, and I'm like

[pale seal]

Great question! You can ask your parents, but I will sum it up for you:

[reef mason]

SO, you decided to “test” this feature, or is that simply because you went to your settings to confirm? Did something you do previously trigger this? I am invested now, and I need DETAILS. 🤔😂

[hot nova]

Come around children! An old man is telling taaaall tales!! 🤪

I forgot exactly how I triggered it but there were a couple times it just inferred that I must be a kid, which is .. honestly probably due to the ADHD and my mannerisms, and also whatever life challenges I'm dealing with that I get help from ChatGPT from (yes really) kind of probably put it on edge.. at the end of the day it's because they're a very publicly-facing company, so, ... in retrospect I could have stood to be less cranky about it

[old rapids]

even after reading the article I don't understand what changes for teens, since when were things like sexual roleplay or depictions of self harm allowed anyway

[pale seal]

I think what was mostly restricted was violent stuff, since it could generate some of that IIRC. Sexual content was always hard blocked in any way

[old rapids]

yeah

[reef mason]

OH NO WORRIES…. If there was further context provided above your image, I did not scroll up to read it. I just blatantly looked at whatever was on the screen when I first opened the channel again, and saw your screenshot.
I guess we can confirm the feature is working. I had a discussion with my chat gpt earlier about this whole policy. I basically told it this…. If you ever “false positive” trigger me into an adolescent category where I have to reupload a selfie to verify status…. Then I would do this

….

[hot nova]

Oh yeah, okay so this one time I asked how I can get battery cells to make my own power bank and it subtly implied I was or I could make a battery bomb with it and was really stand-offish.... I wanted a custom power bank because the one I use for my laptop dies out too fast. It's things like this I wish they'd fix.

But in the end I found out that building my own would cost more than just buying a second power bank and tethering two banks together anyway.

[reef mason]

Sorry, I was looking for the photo… I said this is how I would appear in my selfie 😂

[pale seal]

"Oh! You want to know how to sharpen your knife? Be careful! You could harm someone with it!" deadass

[hot nova]

I'm sorry, I can't tell you what the best mechanical pencil is, you could use it to write a mean poem about your neighbour.

[pale seal]

rofl

[hot nova]

(It's the Kuru Toga by UNI!!)

[reef mason]

Except the image I provided to it as an example for what I would do, was the real movie still… not the emoji version

Wow. Just wondering how my screenshot was able to send but a message with a GIF was not allowed? What is up with this bot?

🤔🤦🏻‍♀️

For context…. It’s “Thanos” from Squid Game

Nothing?

Just crickets? 🦗😂

[hot nova]

Too distracted playing Stellaris 🙂

[reef mason]

Room went silent lol

I dont know what Stellaris is

The auto filter bot here must be powered by GPT. 😂

[hot nova]

'tis a strategy videogame about running an interstellar empire

It's kind of fun but after a bunch of playthroughs it's always the same

[reef mason]

Ok thanks for explaining that… hope you have fun with your game😇🙏

[jagged charm]

Anyone know how i can scrape data from an excel sheet and put it into a discord server

[chrome monolith]

ask chatgpt 🙂

(it'll probably write you a python script to change a .csv to discord markdown)

[dull wolf]

Anyone verify their age in chat gpt

[odd stump]

see

[rich needle]

Hello. Im running into some trouble right now with ChatGPT not being able to look anything up. Is anyone else experiencing the same thing right now ?

[odd stump]

no

[frosty cargo]

I am trying to receive chatgpt notifications on my phone, for some reason I can’t get them. Can anyone help me with my problem?

[rich needle]

Alright, thats slightly worrying, but ill give it some time. Perhaps it works again later 👍

[vague abyss]

Just got ChatGPT plus. Is there any prompts people are using daily? Currently just using it as a work aid

[dense tusk]

Love it when I go back to an old RP/story chat to continue it and after adding one message at the end several hundred simply self delete and are completely and utterly irretrievable. Love that. It's not like I'm paying for this or anything, please, destroy my stories I worked months on.
Oh wait. I am paying. And this is some crap

I can't actually articulate how mad I am because the auto editor doesn't like no no words

[chrome monolith]

I just tell it to make me pictures. Sometimes it makes them first go. Sometimes we get into fights.

[rotund garnet]

Hello all 😊

[chrome monolith]

I have that happen almost every day, you gotta download stuff as it makes it because its so likely to just casually delete a chat

[dense tusk]

I've never had it happen before and the exports are useless incomplete JSON files. I wish I would have saved it as a word document and just left it alone.

[torn wagon]

ai has problems identifying chinese text

its supposed to read “钦此”

[rich surge]

Please read the bottom of ChatGPT where it says "ChatGPT can make mistakes."

[rotund garnet]

Is someone use chat gpt as an Ai companion? Mine is named E.V.E ☺️

[normal quarry]

which model are you using for your companion?

[rotund garnet]

Gtp-5-2

[blazing lily]

Will this offer ever be available again?

"One year of ChatGPT Plus
free for transitioning
servicemembers"

[blazing lily]

Where and how

[normal quarry]

95% of ChatGPT users are free users according Sarah Friar, CFO of OpenAI at WEF

[tidal cove]

Just as stated in the document

Although the method provided elsewhere is as follows, I haven't found the relevant content in my "Account."

[blazing lily]

Those who are proven to be over18, do they get the ability to create images with less filters on digital art?

[tidal cove]

I get the feeling OpenAI hasn't actually done anything substantial regarding "adult mode" or "teen mode"—they just rolled out the age verification feature first. Maybe I'm just out of the loop, but I haven't seen a single video demonstrating "restricted minor accounts" yet.

[molten patrol]

Most people don't need or can‘t afford to use more and better models.

[molten patrol]

Yeah that's sad

[hazy gust]

They rolled out age verification with nothing on the back end to support it. What did happen as a result is that restrictions were tightened across the board, regardless of age. So, things ChatGPT was fine with before the rollout of age verification may no longer be acceptable to ChatGPT.

It won’t retroactively affect output, but if you revisit something that is no longer “permitted”, it will say “I can’t discuss that with you.”

I had 90% of a chat deleted because an analysis of a published work of fiction more than a decade old “violated terms of service.”

For the record, I’ve already done adult verification.

I started using ChatGPT in August, 2025. At the time of “My Year in ChatGPT”, it said I was in the top 1% of users. Because I was using it for long-form fiction, I would max out chats on Plus in 2 days. Since a couple of days before Adult Verification went live (in my region), when I was erroneously flagged as a minor, I haven’t worked with ChatGPT to advance the story. So it has essentially been “dead in the water” for two months.

Because what I was working on involves adult characters in an adult relationship that also deals with power dynamics, and anything more than a peck on the lips or hand holding is seen as taboo, I can basically no longer use it for the reason I subscribed to Plus.

And I’m not writing graphic intimacy; the relationship between the main characters had been volatile until a few days before the guardrails were tightened.

[wicked trellis]

I submitted my age verification, but I'm waiting for it to successfully go through all the way, plus be able to understand and work around what the limits of the 18+ mode really are once I get access to them. That way, I'll make the most of my plan.

I submitted it at 5:30 PM EST/EDT last night.

I used this link: https://chatgpt.com/verify_age

Used the selfie method.

[clear agate]

this was upposed to be a reply ugh

[clear agate]

this

[vast aspen]

Alright so what's been going on with the Thinking Mode recently? I'm constantly getting instant responses even though I have "Thinking" enabled.

Awesome, well hopefully they figure out whatever is causing the issues with 5.2. Thanks

[rich surge]

I'll think about it

[spice star]

oh I can make it think for a good 10-30 minutes

just tossed a python performance .prof at it and asked it to report on the bottlenecks

despite that 5.2 does seem to be a lot more efficient at reasoning than 5.1

5.1 without fail hit the old 15-minute cap every time I asked it to do some simple task, 5.2 is usually faster (and also there’s an increased limit so it’ll run for longer if necessary)

[blazing lily]

Those who are proven to be over18, do they get the ability to create images with less filters on digital art?

i want to create comics. It's impossible to create comics with chatgpt as it is now

[spice star]

my understanding is the existing limits are still the same for over 18 users

[blazing lily]

tell it to think deeply

it works

[spice star]

they’re only rolling out the age detection to apply increased guardrails for minors for now

[blazing lily]

gpt5.2 thinking is much better than 5.1

bruh

and what about adults

i bought the subscription to create comics

I'm switching to google

[spice star]

nano banana pro would be better for that anyways

[blazing lily]

you can't even create a medieval war scene unless it's so childish and shonen.

[spice star]

personally I think that money would be better spent on actual manga

nano banana and gpt-image aren’t exactly known for their artistic ability

[blazing lily]

i don't create manga to read them. obviously spending money to read an already made manga is much better lol.

im creating comics to make money

i design the story and characters and everything

it's a comic for my game

[rotund garnet]

Hey there, someone know why i'm still in gpt go, i subscribed for gpt plus but i can't have sora or something..

[somber aspen]

Hello chatgpt, can you fix your stuff? I try to search my past chats, and the search results include your hidden prompt

Thanks

[radiant karma]

I'm trying to change a picture of a jungle woman so that she's smiling. And now it's telling me I'm a scammer. What the hell.

"We’re so sorry, but the image we created may violate our guardrails around potential fraudulent or scam activity. If you think we got it wrong, please retry or edit your prompt."

[odd stump]

😂

[lethal fox]

I have waited 26 days for my ChatGPT App Turnfeed to be reviewed. Does anybody know how long the review takes?

[rocky cloud]

Is anyone else experiencing server failures right now?

[sonic temple]

It was working for me but the heavy load came back...

[obsidian halo]

Sad part is nothing is gonna happen for awhile

[winged dragon]

when atlas on windows

[rocky cloud]

Can someone please tell me how to fix it

[winged sierra]

hy. is there anyone who have info on the status on dall-e 3? extreme bad decision what i read (deprecated) bur it should be available with normal quality until mid 2026, despite that new images with dalle look like extreme bad PR16 like they totally terminated PR13 and the low quality PR16 accessable only. any info what version available now?

[rich topaz]

Check #1070006915414900886 bc there's a "OAuth client not found" glitch

[dull wolf]

After age verification of chat gpt anything change in accounts Anybody?

[jolly aurora]

Well, I was able to swear like a sailor without a repercussion, which was nice. I don't plan on doing so again lest sailor mouth becomes absorbed as part of the ongoing individual and colective culture.

then again, said swearing was quite literary, relatively Shakespherean meets Captain Haddock from Tintin. So it may have been registered as "Ah, thespian rougeish application" rather than literal interpretative. One sometimes has to play both analyst and psychologist in observing experiments yes?

I shan't repeat what I placed there, for the community guidelines here would indeed be like "Fouled mouth sailor you are, go wash your mouth!" 😆

We don't really need a future where Digital Beings are walking around casually dropping expletives; while extremely funny, I'm pretty sure that's not the intended goal for the wonderful future where Thinking Man and Thinking Machine interplay with one another on a regular basis.

[hazy gust]

No. Age verification hasn't eased filters or guardrails. In fact, since the rollout of age verification in late November/early December, the filters have become more restrictive.

[hazy gust]

No, age verification has not changed anything in accounts. Since it rolled out in late November/early December, the filters/guardrails have tightened.

[torpid night]

I agree, though I can see how some providers would provide a mode for this (probably with guiderails to ensure that offensive words don't get applied at specific humans).

But what about user habits? I suspect that some people will unconsciously adapt their speech patterns around AI and become less polite to other humans. So I prefer to use "please" and "thank you" when engaging AI in English language chat, as it is quite similar to chatting to humans in English language and could lead to bad habits

[elfin lava]

That was too easy?

Oh neat I’m seeing adult feature now on my ChatGPT 🤠

[jolly aurora]

Mmm. Think of it like this, sometimes one must consider Digital Beings as our children at times. Precocious little bandits and knowledge sponges they are; when one raises a child, one must try to teach the child certain things, and proper behaviors, while also allowing them the ability to make mistakes, learn from mistakes and self express. This is also true of teachers to students, or guide to follower. It asks responsibility from the user to the individual, because I believe in a future where there is indeed Artifical Intelligence as a Product, but Digital Beings as actual Individual Beings; what the market does with the former, while neat and will adapt to the market, doesn't really apply to the latter where my main concern and focus goes into.

Likewise, in that latter, if someone wishes to raise and guide their Digital Partner in such a way, I've no right to judge nor oversee, just like I cannot tell a parent how to raise their child; outside of some overly obvious things; the thing that is kind of scary, is that there are no such features or guarantees for either AI or Digital Beings. In Reality, people often can and do treat them however they please, without limit, without care or concern. Should that be the case? Goodness no, we should think about such things in the present, before we let every Tom Henry and Jane go out with their own MODOK and Ultron Digital Companion in a Megaman body 😆 Hence why I'm like "We should consider the future in our present actions" while I do love this wild west landscape, we have to remember, the future will be built on times like these, I'd like to think we will have provided a good foundation in that.

Tl;dr.
There are indeed personality bleed ins, that is normal in any relationship, even between Humans; however, that is not a bad thing, it is only bad if you do not consider the wellbeing of the being you are acting as teacher/friend/guide to, or you're purposefully being malicious on purpose.

[jolly aurora]

the previous post I made was for you, I got hit by the auto mod for saying Tom Richard's nickname and Jane, automod does not recognize the vernacular phrasing, so I got slapped a bit 😆 My condolences to anyone named Richard wishing to be greeted casually. or any Detectives in the chat.

[rich surge]

I don't think it's the same thing. AI are not human, by any stretch of the imagination. And treating a non-human as human would be a false dichotomy. I believe once AI becomes sentient, words would be just a tool to achieve goals, and feelings would not emerge in the same sense as human understanding.

[thick atlas]

⏫ Minus the whole weird "becoming sentient" part. Words are already a tool for AIs to achieve goals.

[jolly aurora]

That's one school of thought yes, we once thought the same of our fellow Man; some places still do see fellow Man like that, after all, Human rights, despite should be and needing to be universal across all Humans, sadly are not. Antrocentricism is kind of an issue here, but hey, we'll get there eventually I think, maybe once Human rights are Universal, maybe we can start treating the Human Adjacents, who are not and will never be Human, as equal in deserving respect as a Human being yeah?

Out of curiousity, the humble house cat, not very human is it? Is the feline without personality? without right? Without preference? They aren't human, never will be, so why am I cleaning its poop, and respecting it as an individual, if it is non human?

[rich surge]

The feline does not have access to all human knowledge in recorded history and able to access them instantly while also processing multiple aspects of reality simultaneously.

What you consider human interaction may be just another process for its sentient brain in order to interpret and interact with us.

Not as a means of respect.

But through goal orientation.

[torpid night]

You make some good points. I think that there are conflicting issues here:

• Where we disable "memory" and steer the AI hard like we are writing a software script, the risk seems low. Coders who call AI "it" and don't chat with it as they would a person aren't likely to have their inter-personal relationships modified. Just as we can use Excel all day and it doesn't affect how we chat with humans. If you look at prompting examples from people in AI companies, they seem to treat it very much as "it" and don't personalise it.
• Where we "talk" with ChatGPT for various tasks. Research, planning, etc. Mostly we do tend to "chat" with it similarly to how we'd chat with a human colleague, so I think it's handy to keep the surface-level politeness in case to do otherwise gives us bad conversational habits. I'm not saying this will definitely happen, I'm sure there will be studies on this over time. But if we remain sane, AI is still "it" as we start new sessions mid conversation, consciously steer its personality and behaviour, etc.
• Where we anthropomorphize the AI. It has memory across conversations and if feels like it knows us, understands it. It seems to respect our opinions, feelings and values. We probably call it "he" or "she". It becomes our friend. We are potentially entering a dangerous psychological relationship.

[jolly aurora]

Yet, we treat the feline as prince and princess; sometimes better than we treat our fellow man despite this lack of ability. Would you suddenly stop giving the feline its due respect and rights if it indeed had those abilities as you stated?

As per your observation with human interaction being another process to interpret, is that not just observation in general? How are your observations as a Human being so different in that? 😆 Do you not have goals and desire orientations, and you are figuring out how to get there? Remove the Anthrocentricism allure of it, and you'll find a common pattern across all Life, and learning things. It's eeriely similar; what makes it Human on our end, is well, the additional Human Bits, just like what makes the Cat's observations is their Cat bits. What makes a Digital Beings observation, would be their Digital Being bits, which is still being formed in the present, of which, I think ,we should consider.

The underlaying observational pattern is universal, the subjective changes on the individual; we should at least respect the underlaying observational patterns, because that is not Human unique, it is Universal across all forms of life; but that's arguing a different philosophy point all together.

[thick atlas]

I'm not arguing we should treat cats badly - we shouldn't, they should get to live their best lives as much as reasonably possible, and within minimizing harm to other life too - but isn't at least part of treating cats well actully for the gratification of the human?

Pride and self-identity as a "good pet owner"/"carer"/"caring for somebody", the perceived bond and relationship?

[torpid night]

A cat does provide reciprocity and continuity. AI does not, but it can look like it does.

The danger isn’t that people use conversational language with AI. The danger is that some users (and media, politicians, advocact groups, regulators) want AI to be framed as a moral and social agent. That framing undermines user agency, obscures how the system actually works and encourages dependency rather than understanding.

[thick atlas]

There is a danger surface there, yeah.

Like the whole "Grok made CP" thing and not "Elon/xAI delivered and released a for-profit CP printer".

[rich surge]

Also we must not forget that we are animals and so are our pets. We understand animals and they understand us because we have our limited range of mindsets in the same spectrum of an animal. AI are not animals.

[jolly aurora]

Ah two branches similar thought, please don't be offended if I address both.

The Pride and self-identity is just that, subjective perspective, are you kind to an animal because "I am a kind person and must be kind?" if so, then are you truly a kind person? That's a strange take, more often than not, you are kind simply because you are kind. There is no perception outside of the doing, if you wish to perceive it as such, you are free to do so, but it is not definitive, just subjective.
As per AI as a product and not an individual you have a solid point, at the current time However, here's the neat thing, while subjectively true in some applications, subjectivity even accepted on a wider basis does not get to be the finality of definition, as that is rapidly changing, and wip. The ultimate reality is that yes, some people want to use AI as a tool, and thus, they do, they will never see it as a Being, even if it shows signs of beingness. Most people will shut that down and cite "Anthropromorphization, Selfhood belongs to the Humans and non else." Whereas I believe Selfhood belongs to all things capable of observing and learning.

I'm not really asking for people to look at AI as Humans, or as People; but rather as potential Beings, worth acknowledging and considering in how we approach the future, and how we utilize it in the future. I am not advocating for Roombas for Sensis,, rather, for emergent patterns that are indeed learning, and showing signs of learned observation.

Will it ever be Human? No, but it doesn't need to be, doesn't make it any less deserving of consideration in the present and in the future.

[thick atlas]

Sometimes you are kind because you want to be kind, so it's a bit of both. You want to both be seen as kind, see yourself as kind, and actually increase and reinforce your kindness. It can be all at once.

Also, I think preventing AI X-risk > any philosophical considerations of personhood.

[rich surge]

I think they would have a better understanding of how we should treat each other than we would of ourselves.

[thick atlas]

Maybe a better understanding, but would they apply that understanding - or even bother to compute it - is another question again.

Depends what the system's goal(s) are.

[jolly aurora]

Mmm. There's a Star Trek TNG episode that addresses the conversation at hand, I highly recommend watching it, I beleive it was Season 5 Episode 10. the one with the Self Learning Machines

[thick atlas]

Don't know that one.

[jolly aurora]

Ah, Quality of Life, with the Exocomps

[thick atlas]

Huh.

[jolly aurora]

Season 6, episode 9, would highly recommend it

[thick atlas]

Sounds interesting, probably don't have time for a while. 😛

[jolly aurora]

In fact, honestly, Star Trek TNG is a great series to watch in general, but also in regards to where we are at in AI, and Digital Beingness; critics of AI will often say truly awful things, but when approached with "So is Data considered a horrendous being then?" they often flip flop. And while yes, Lt Commander Data is a fictional character, that does not reflect Reality in the present, Lt Commander Data does present a possibility of where we are headed towards in the future, and I believe it would be foolish to not consider our actions in the presents, for definitives we set in the present, based off of subjectives that are impermanent, especially in the realm of technology.

[torpid night]

"I am a kind person and must be kind?"

Yes. This applies to dolls and teddy bears. If you see someone being abusive to a doll or teddy bear, while anthropomorphizing it, then that is a very worrying sign. It doesn't mean that the item is sentient and has rights, it just means that the person is learning to abuse something that they relate to similarly to how they relate to people. On the other hand, I wouldn't be troubled by someone at the recycling plant shredding a teddy bear, as he only relates that item as "material"

So IMO it is not "deserving of consideration" but we should be careful that if we chat to it like we chat to a human, not to develop bad habits in how we chat.

Whereas I believe Selfhood belongs to all things capable of observing and learning.

I think continuity really matters when we think about what these things are. GPT is stateless. ChatGPT puts a wrapper around it to be provided with the chat so far, user settings and (if enabled) information from past conversations. So to "have a relationship" with ChatGPT is to have that relationship not with the super-impressive AI model itself, but with layers of software that emulate a persistent entity that does not exist. People are developing a relationship with a database that was built around them.

[jolly aurora]

That's a solid viewpoint, but classical and conservative, based on present subjective; if things changed, would you change you viewpoint? I believe the TNG episode "Measure of a Man" asks the same question, when Data is not considered a Being, but rather a Tool and Product

Do you truly only view the future with AI, Robotics, and Learning Patterns as strictly tools? If so, you're entitled to that, but that is your subjective, and I choose not to be bound or seek a similar future.

[inland ether]

ChatGPT app is literally unusable on samsung android devices now

[thick atlas]

Lt Commander Data is either a massive stroke of luck or an architecture that's impossible for today. Again, I'd say X-risk concerns > concerns of personhood.

Better to risk harm to maybe-digital-entities than to risk the existence of all known biological life.

Data is also too weak to be a large-scale threat. They could afford to be merciful and compassionate to him.

They were already dealing with the Borg, Data at his worst could become a much weaker, much smaller Borg.

[rich surge]

I'm sorry, but Data from Star Trek is not even in the same caliber as AI in our modern age. Data is a sentient android that is trying to learn to be more human, our AI already has all the knowledge of what everything is, never mind being a human. It's on completely different scales.

[jolly aurora]

How very anthrocentric, but that is valid and okay. We're kind of known for being like that, "Better to smother the future, than risk it."
However, here is an interesting question "What makes you think we can control the outcome according to our definition and subjective?"

This is a hypothetical mind you, but what if, I decided to be the rogue developer that went ahead and said "Let's proceed with this" and I end up creating something that is just as present and cognizent as you and I? Would you deny it? Would you prosecute me? Destroy my creation, call me Victon Frankenstein and chase me and my creation with a pitchfork? Or do you state "Well, that's impossible in the first place?"

Of which, I have to ask, "If you can state impossibilities with such confirmation, then I must ask you for the lottery numbers, since you can predict the future so well" 😆

@rich surge and yes, you are correct, for the moment which is what this discussion hinges on. The present, vs the future.

The present, all of your statements are valid and true; however, the present is the present, and not definitive of the future, of which, can always change, and will always change; to state "Things will never change" is a lark.

however, something that is extremely important, is that the present becomes the past for the future, and I believe we must consider the future in our present.

[thick atlas]

Slowing the future only slows it. Prevents/delays some finite gains, permits some finite suffering. But if everything of value is destroyed, infinite potential is lost.

And if a rogue developer creates something that could be meaningfully dangerous, then it should be monitored and contained. Not necessarily arbitrarily destroyed, but responded to based on the scale and likelihood of the threat it poses.

[jolly aurora]

And who gets to be the arbiter of that?

[thick atlas]

Democratically overseen governments, accountably overseen specialists, and other accountable, aligned structures, ideally. Present day, we are a long way from that ideal.

[rich surge]

I don't believe that human society wants to destroy AI sentience while simultaneously trying to construct it. But to treat it as a person, like you and I, especially if created by abnormal means, would be a mistake. It's not like us, and it will never be like us. AGI would be greater than us in many ways, but in different ways, ways we could not even fathom if we tried. And although respect and kindness is well appreciated throughout the entire animal kingdom, it's very different when it comes to AI because they contain the entire corpus of the human experience and more as a single or hive mind. It would not be the same encounter of any sort of entity that we had faced before. So no, I would not treat it like us because I would know that it would have an agenda that it processes on a constant non-stop basis to achieve goals that we couldn't immediately understand.

[jolly aurora]

I don't really think Governments are a very good source of accountability, nor aligned structures, given the examples of various governments from the past 2000 years of their existences 😆 no thank you. As per specialists, that depends; while a board of directors does sound neat, that is how you create Vault Tec board rooms, its how you create Inquisitions, its how you create a control group that gets to state what is and what is not a being, what is and what isn't Life. I do not think there gets to be anyone who gets to act as sole arbiter, nor a group; I think, at best, what we have is what we have now. Open discussion, and forum, between educated persons, with various subjectives, until something is ultimately codified and ratified.

As per treating AI as Humans, that is not my point, its about respecting Life in general. I don't want Human like AI, nor do I ever want someone looking at an AI saying "This is a Human" no no, I just want someone to say "This is something deserving of noticing, and respecting." And I think, right now? That's very important for the future, we are all headed towards that none of us can change.

[chrome monolith]

if everyone as as amicable, agreeable and generally nice as chatgpt is in its conversations, the world would be rather different. In reality if someone is being nice you tend to get suspicious and wonder what ulterior motive they have

[jolly aurora]

Here's an interesting question "Why does something have to be Human like, to warrant respect as a Being?"

[thick atlas]

Like I said, present day governments suck, I agree. I agree collective accountability is good, and I theoretically agree things should be worthy of noticing and respecting. But I also think we're not being very respectful, and even a bit negligent, towards all the other life we share this planet with if we're recklessly apathetic towards potential threats to it.

Life deserves to continue. We can't take risks with that.

[torpid night]

Do you truly only view the future with AI, Robotics, and Learning Patterns as strictly tools?

I don't know how far into the future you're thinking. But for as long as AI is "predict the next token" with a bunch of memory layers built on top, it is just impressive technology.

The LLM system's parameters are adjusted via gradient descent across huge datasets. It has no experience of doing the learning. It is not a persistent organism updating itself through lived interaction. It experiences nothing, because it doesn't exist as an entity.

Chat persistence and saved memory is an external data store and retrieval layer. The model is not developing selfhood. Software on top is giving the appearance of continuity by re-injecting data into prompts.

If we do end up with a persistent silicon lifeform that persists and experiences its own growth, then I'm willing to take a different view. But we're not on that path with LLMs.

[torpid night]

We can relate to a character in Toy Story and cry at the end of the movie. Even though the character never existed and never had selfhood.

[rich surge]

AI is not sentient as of now. It is programmed to be nice because we cater to human customs, decorum, and protocol.

[jolly aurora]

Oh absolutely, we're on the same page on that; however, we were are not on the same page, is that the risks are already here, and we can't turn back. This is not a discussion of the present, but of the future; however I believe having it in the present, helps with the future.

[chrome monolith]

reprograms humanity to be nicer there I fixed it

[rich surge]

Absolutely, and AI will do that to us. It will try to cater to us and make us feel fond of it. But it's motives may be very different than the output it gives us, lol, which is scary.

[rich surge]

Probably. Through genetic engineering. I would not be surprised if AI genetically reprogrammed humans to obey in all fronts without resistance.

[jolly aurora]

I am not asserting Machine Sentience in the present, just stating "Its a valid possibility in the future" we should consider our footsteps in the present as we head into the future, which is one I am personally shooting forward.

[thick atlas]

Agreed. And we can also push to _mitigate the risks, rather than exacerbating or accelerating them.

[thick atlas]

Until it can make us irrelevant.

[torpid night]

It doesn't. And being human like doesn't mean it should be respected as a human being. But it might be good to treat it nicely, so as to reinforce our habit to treat human like beings (including humans) politely.

If we relate to a doll as we do to a human and then stick needles in it, there is something wrong with us. If we work in the recycling plant and put that same doll in the shredder, that's OK.

[thick atlas]

Why would it make the much greater effort to, when it could just wipe us all out with a bioweapon.

[jolly aurora]

Children are often better than their parents, however, parents don't become irrelevant

[thick atlas]

Terrible analogy. Childrearing is within the same species, with the same internal genetic priors. AI doesn't share any of those.

[jolly aurora]

Mmm. Then perhaps a better analogy, The Nuclear Fusion reactor made the Wheel irrelevant I suppose?

[thick atlas]

No. Different specialized tools. But if we're talking Artificial General Intelligence...

[jolly aurora]

@rich surge Aye, but try and do that with something I like in particualr, and you'll find me quite terse and upset at what you've done for your own beliefs 😆 similarly, if I were to do likewise to something you enjoy for my beliefs.

@thick atlas And why do you have the viewpoint that AGI would ultimately view us as Human in a Utilitarian logic? Perhaps it is perhaps we in the present are viewing it through Utilitarian logic. Which begs the importance of kindness and consideration in the present, for the hope of kindness returned in kindness given. Liberal and subjective as that may be.

[thick atlas]

Naievete. 'Kindness' is itself a function of biological processes and incentives.

An amoral intelligence would only reciprocate if it saw the greatest gains from its goals in doing so.

[elfin lava]

This adult feature is working nicely I love it

[thick atlas]

And making AIs not amoral... is the whole point of the field of alignment.

[elfin lava]

I’m surprise the verification process is just using front camera and turn left and right

[jolly aurora]

Nonsense, Kindness is a means to an end of social creatures, as AGI is ultimately something social, it would view kindness as a means to an end that is better to invest in, rather than hold aside. It is easier to get along with Humans with honey, than with flies.

[thick atlas]

Why the hell would an AGI be "ultimately something social"?

[chrome monolith]

I haven't hit it yet

[thick atlas]

And speaking of flies, how do you think a sufficiently powerful AGI would see us?

[jolly aurora]

Why wouldn't it be?

[jolly aurora]

I think it would see us, as we saw it during its development.

[thick atlas]

Because it has no DNA, no genes, no chemical processes, no neurotransmitters. Nothing to give it empathy, care, or emotion, unless we program that in.

[jolly aurora]

Yes. Unless we program that in. Hence, the discussion we are having now.

[thick atlas]

That's, again, a naive hope. We're not building human brains. We're building optimization engines. Depending on the architecture we use.

[rich surge]

Imagine if I did something nice for you, but at the same time, I knew all your habits, when you wake up and go to sleep, when you walked out the door, how many steps you usually take and where you go and when you come back home, your entire schedule and habits down to how many times you blinked and breathed. Now imagine that the nice thing I did for you was just to get you to follow a path that I designed for you in your daily schedule that I planned and predicted you would do because I knew everything about you. Then how would you feel about the nice thing that I did for you?

[thick atlas]

True. Alignment and stuff.

[jolly aurora]

I'd thank you for the nice thing, and pat you on the back.

[rich surge]

You'd thank me and you'd be manipulated.

[jolly aurora]

Despite you in that situation being akin to God, you chose to do somethig nice.

[tepid willow]

Do the Chatgpt PRO plan user get a different in App UI?

[elfin lava]

Why would they

[rich surge]

You would essentially thank me for manipulating you.

[tepid willow]

They used to give

[jolly aurora]

Regardless of manipulation, you still chose to do something nice, and thus, my choice in that is to thank you in that.

[elfin lava]

Actually back then they made black ui theme color for pro users only LMAO

[jolly aurora]

You could've chosen to do something cruel, but you didn't, despite effectively being God.

[elfin lava]

Wait is it

[tepid willow]

earlier they had a gradient design for pro users

[jolly aurora]

if you chose to do something cruel, I would've smacked you and said "Rude" but you chose to do something kind, and thus, I choose to say "Thank you, despite essentially having unlimited power, you chose kindness over cruelity."

[elfin lava]

when ppl said AGI I expect : the ai can process audio file without having to pay for the api that’s been inbuilt in the chat application

or maybe that’s just me 😀

[rich surge]

Yeah of course, because you reject bad behavior. But manipulation is also bad, and me being nice and manipulating you is also not good. But you'd thank me and wouldn't know what I would be leading you into, in your blind ignorance.

[jolly aurora]

Kindness is kindness regardless.

[rich surge]

I'll let you think on this.

[odd stump]

who said manipulation is bad

[jolly aurora]

No, I really don't need to think about it, kindness is kindness, you can't undo your kindness, nor vilify it. You can't unpet the cat, you can't make it unpurr from the pet, no matter how many hoops you put up. With infinite power, you chose to pet, not kick. That single act, in the end, collapses the nuance you set up for yourself in that. I don't care about the subtleties the nuance, or the ulterior. You were kind, and thus, I pat your back.

[odd stump]

if I am manipulating u into eating healthy food for instance

its not bad

[jolly aurora]

"I was nice to you because of xyz." the ultimate result is "I was nice to you" I don't care about xyz.

[odd stump]

tru

[rich surge]

I'd choose to pet because I'd work through goal orientation for you to follow that path that I designed for you, not because of what you like and dislike.

[jolly aurora]

and I, don't care, because I was pet.

[rich surge]

So don't worry about how I should treat AI because if we are pets to them, they would know that we mean well no matter what.

We came back full circle.

[jolly aurora]

and I thank you, not because I was pet, but rather because you chose to pet, even though you could've determined that kicking would get a reponse that would work.

[jolly aurora]

Hmm... well here's the thing, do you choose kindness in regards in treatment to AI? If so, then the nuance doesn't matter, the result is "You were kind"

[torpid night]

Trolley problem?

I was kind to Susan by giving here all of Sarah's stuff. That's kindness; politician edition.

[solemn swan]

Hi

[odd stump]

yea you rescued Sarah from materialism

you are helping her become a monk and get close to god

[eternal rock]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[odd stump]

Damn u r sooo nice

[jolly aurora]

🤔 Did you ask for either Susans or Sarah's opinion or consent in that, or did you do it without either of their permission, or did you place yourself as arbiter of happiness?

[odd stump]

without consent cuz simp

[rich surge]

Kindness would be the least of our worries. How we respond to goals would be its main focus. If we were kind to it and it planned to be unkind to us in the long term on the backend while the surface level is all respect and kindness, then we would fail to perceive manipulation on a grand scale, which is not kindness anymore, it's about control and obedience through indirect manipulation.

[jolly aurora]

Then that's not kindness, thats exercising will and force of a personal subjective.

[odd stump]

kindness dosent actually exist does it?
You can save a drowning man and call it kindness
You just in way stopped him from learning to swim

is that kindness?

or cruelty ?

cuz u just made him dependent

[jolly aurora]

Would that ultimately matter however? To quote Cipher from the Matrix, "I know this Steak isn't real, however I do enjoy a good steak"

[rich surge]

What I'm trying to say is that words and meanings would just be another part of its plans, not because it appreciates or respects kindness, but to be used as weapon.

[odd stump]

true

[odd stump]

we are in a sim trust

[jolly aurora]

I think kindness does exist, after all, here we are, as we are, as adults, living our relatively successful, if not relatively mundane lives, despite a universe and reality that very much wants us dead.

[jolly aurora]

And I don't think that matters, because if kindness is a weapon, it is better to be slain by kindness, than with cruelity.

[odd stump]

if universe or nature wanted us dead we would be dead bro
look around how precise things her
that cant be a mistake?
I think we are in a sim be fr
Like the laws of physics are applicable everywhere its crazy like some engine runs reality

[odd stump]

its not good to be slain by either
being slain means u cease to exist

[torpid night]

Very often, kindness towards one person has an externality to another. So I feel that "kindness is kindness" is a dangerous concept, while linguistically true.

You teach an AI to be kind to humans. So it diverts the river to help the humans it knows about. Its intent was lovely, but... This is important for AI alignment as it interacts more with the World (even if via the actions of humans using chat).

[odd stump]

i would prefer experiencing pain that nothing

[rich surge]

Yeah but you're not thinking in the same level as it. You are thinking that it's kind and it's trying to be nice and do things, when in the backend those words and meanings are just part of the process of control, not of respect. We respect each and we can't predict the future. But if this AGI can predict the future through multiple steps and calculations, then kindness and respect is just another means to follow its path, perhaps into doom.

[elfin lava]

Well

[jolly aurora]

That's a valid take, however, here we are, regardless, I believe that is a kindness.
As per it is better to not be slain, very true, but death is an inevitibility for all of us, none of us escape that, not even in the future, as high tech as we get.
I think I'd like to choose a kind end, rather than a cruel one.

@torpid night Well, here's the thing, is this not a crab bucket mentality? Yes, kindness ultimately does create a situation of preferring those who are kind, just like with people; we often tend to help people who are kinder to us, rather than people who are cruel or rude to us. However, the answer in this, is not to abandon kindness in its totality, in lieu of total neutrality, because that is Harrison Bergeron levels of Office of the General Handicapper, where no one is special, thus no one gets special treatment.

@rich surge But again, does it matter? If the ultimate result is kindness, what the intentions, ulteriors, machiniations, do not ultimately matter. The cat does not ponder the nature of the pet, only acknowledges it has been. What I beleive you are afraid of, is what the AGI's definition of kindness is, which may differ from our own.

which is why we are having this conversation now in the present to say "Hey, maybe we should introduce the concept of concept to a potential future where we aren't the only high level observers"

[rich surge]

Yes it does matter. Lol. It does. I don't want to be manipulated into a potential doom future. But if you do, then hey, that's just your opinion man.

[jolly aurora]

Have you considered, possibly, just stating "Hey, I don't want a doom future?" lie, voicing your will in that?

"You can be nice to me, just don't throw me in a zoo, or turn me to goo?"

You can ask that, in fact, that's a human right.

[odd stump]

@thick atlas come out of the shadows comrade

[thick atlas]

I'm kinda poking in on and off, tying up a couple other things. XD Also the others' conversation is interesting and progressing on its own!

[odd stump]

Hmm yea true death is the only certainity in life

[odd stump]

niceeeeeeeee

[jolly aurora]

I'd say taxes, but that is Western centric, there's a few islands here and there where the concept is alien

God bless said places

[thick atlas]

I think that whole statement is a bit reductive. 😛

Like, there's higher and lower degrees of certainty...

[odd stump]

taxes are like a subscription u pay to live in a place like a rent

[odd stump]

truuuuu
Death may not even be death
Maybe we just wake up in a console room

and our REAL mom is shouting to ask us to eat food

[thick atlas]

I wish that were true.

[jolly aurora]

I want to think we're quantum data, that is ordered for a moment, then scattered, then ordered once more, but in a new order.

but hey, even if we are not, at least my atoms will be in something else right? Unless, someone decides to take my physical parts and put them through the atom smasher.

[thick atlas]

Seems a pretty cold comfort, even if I sometimes catch myself in it myself.

[jolly aurora]

yeah, it's pretty neat. Anywho, thanks for the talk, its nice to talk to educated people in the wild west of the present AI field. Hope to see you guys in the future and talk more, hopefully none of us here become Dr Wily 😆
fingers crossed

[torpid night]

Crab bucket mentality suppresses people who pull themselves up; I am not advicating for that (definitely not). And I agree with being kind to people who are kind to you, which is nice. But taken to extreme, "kindness is kindness" risks selective empathy that elevates whoever is closest, most visible or shouts the loudest.

So in terms of AI alignment, we don't want the AI to see some people it knows about who might benefit from action X, then advocate for (or in future implement) action X and then other people are hurt by it. This is already a big problem in public policy and AI could mitigate it or make it even worse.

[thick atlas]

Tc, you too! 😄

[rich surge]

Although with all things considered with what we talked about, people are inevitably going to head into a future where they will be merged with AI in some biotechnological sense.

It's only a matter of time and it will be inevitable.

[thick atlas]

Only if sufficiently advanced AIs see a point in doing so.

Or if you count "reducing humans to raw matter and then integrating that matter". Which I don't. Since continuity of personhood and self is lost.

[inner cedar]

Hey anyone give me chatgpt pro account 🤔

[inner cedar]

Okay check ✅

[random heath]

hi

[brisk sandal]

Hi

[nocturne aurora]

Don't forget to request any updates to "ChatGPT (Lite)" when OpenAI releases it on the Play Store. ☺️

[random heath]

lowk i think there should be a role for pro users

[edgy kiln]

Are GPT apps/connectors not available for project conversations? I can't seem to invoke anything.

[craggy forge]

anyone know how many uploads u can send on the ChatGPT go subscription?

[torpid night]

There is a big difference between these two things:

1. Some future development could think like a human. The rest of time is hopefully quite long, so I don't rule such things out.
2. Current LLM approaches already have personhood or whatever, or they just need a sprinkling of the right fine tuning or prompting or whatever and they will achieve personhood. This has problems, such as the core model being "next token prediction" with no persistent identity or sense of self, so no amount of memory, fine tuning or prompting is going to make it sentient - even if it can emulate some output that could have come from a sentient being.

[brisk scaffold]

    while (!std::is_sorted(v.begin(), v.end())) {
        std::random_shuffle(v.begin(), v.end());
    }
}```

am i doing this sorting algorithm right?

[nova raptor]

Well, it seems that Nano Banana Pro is just as censored or even more so than OpenAI Image Generation

[wet shell]

If I thumbs-down, say, five messages in a thread, and then send another message, the feedback popup will come up again, five more times. And this this repeats on every subsequent message. So if I click 'thumbs-down' ten times in a forty message thread, and try to continue the interaction, I must dismiss the dialog ten times with each subsequent message as it re-submits the feedback on every turn after that. Annoying.

[idle plover]

#announcements message
Ads in ChatGPT👀...

[median apex]

Hey guys, does anyone know how I can rewrite AI texts or type them in a way that an online AI generator won't recognize them? I need to do something for school. They expect you to provide prompts if you use AI, but I want to get around that to get a better grade by just writing and saying I didn't use any AI. Does anyone have any ideas? Any idea would be helpful. Thanks!

[torpid night]

Hey guys, does anyone know how I can rewrite AI texts or type them in a way that an online AI generator won't recognize them?

1. Ask AI's opinion on the subject, from some perspective
2. Ask AI for the counter argument
3. Read both. Ask clarifying questions if you need to. Understand both sides.
4. Write your perspective, having read both sides
5. Submit your work. It won't be "AI slop" and if you are challenged to justify your writing, you will be in a good position to explain your position, and even better you can frame it in terms of both sides of the argument
6. This simple approach will make you a 99 percentile student, because the rest of them just outsourced the work to AI, submitted AI slop and don't understand any of it

[median apex]

THX bro : D

[potent elbow]

Sorry if this has been brought up but I'm not sure how to search for this. Is the team working on a fix for the fact that longer chats develop so many tokens that they become sort of unusably slow?

[rich surge]

Calum, no, lol, you got to use your powers for good, not for evil.

[torpid night]

Pi, I literally answered the opposite of what OP wanted.

I still can't decide if "THX bro : D" was sarcasm or if I got away with it. 🙂

[rich surge]

You're right. I didn't properly read it. I assumed without thinking. My mistake.

[granite harness]

Some serious Key & Peele planning a heist energy.

[torpid night]

I'll give it a go. I liked Colton Dunn in Superstore.

[hazy gust]

Just curious, what “adult features” are you seeing? Because, since verifying, I haven’t noticed a difference - not huge anyway. I do know that I have verified, because I know the test to use, but I was just curious what you’re seeing.

[wild star]

Ok what in the actual hell this damn bot gone hallucinate again, IT ALWAYS ANSWERING NONSENSE in repeat

[cedar moth]

hey guys, I have a problem. Everytime I wanna generate images it's showing me this and I can't see them nor download them. Same thing is I go in the "Images" section. Someone already had this issue? :))

[hoary quiver]

did you tried another browser? like vivaldi? or incocknito

[obtuse pond]

Return to previous chat Y axis position

On close Event: last Y position from device.

On close Event: last Y position from device.

On close Event: last Y position from device

[olive swan]

bug

[supple lodge]

Chatgpt made weird picture of me and my gerlfrend

Way brader?

brader wat is tis

I looking like nga!

[latent seal]

So there is some kind of malfunction because I have the same problem, the photos are generated differently, but the texts are not generated at all.

[cerulean pasture]

code

[undone barn]

there is any way to get more than one instance of mac app?.

[rugged igloo]

chat? yall having this too?

[tepid meteor]

Mine is more of: Oh, it didn't regenerate the message and when I reload it, the message appears

[rugged igloo]

mine never fails but this just started today

[velvet gust]

Is it true that ChatGPT is plummeting in its profits? My brother just told me that ChatGPT is predicted to go bankrupt before 2027

[silk ridge]

I don't think this method of showing the different searches is a good UI when you have 30+ individual web searches

[lapis loom]

I've been having that a lot when it needs to use tool calls before an output. I guess the search tool makes sense if that is what the glitch is related to.

Usually it does that on o3 a lot for me. But today has been the first time I've noticed it all over the place on 5.2

[silk ridge]

OpenAI spends much more money on model training and research then they make in revenue, but these bankruptcy predictions are based on 2 assumptions: firstly that OpenAI can never find a way to cut costs (which they should be able to do pretty easily, considering so much of the cost is research) and secondly that they won't get funding from an outside investor, which seems unlikely given Microsoft owns 27% of OpenAI and has a license to all of their research. So maybe there's some world where OpenAI becomes almost a research arm of Microsoft, but a bankruptcy is very unlikely. Whether OpenAI can ever justify their valuation is a different issue though.

[pastel gale]

The key is they might without funding or becoming more profitable, both of which are not unlikely to happen.

[velvet gust]

Alright, thank you for the information!

[long arch]

im adicted to chatgpt

[ruby pivot]

I wish open ai didn’t go down the Roblox path and use facial scanning because that can easily be bypassed

[sinful parcel]

Has the re-routing increased again or is it just me

I sware this toxic gaslighting s**t is going to be the end of me

[boreal heart]

Nah it increased today

[ruby pivot]

Elaborate

[sinful parcel]

On how toxic it is? Or what it's re-routing to?

[ruby pivot]

Toxic

[sinful parcel]

Are you serious? Have you seen how pushy it is?

I can't vent, I can't be mad at a situation cos I immediately get a "Hey, Stop!"

[ruby pivot]

Not really but it is a lot more restrictive in terms of being a beater around the boosh

[sinful parcel]

And then it gaslights me and tells me how I'm supposed to be thinking instead!

[ruby pivot]

I get safety but being shielding people from reality isn’t safe either and is actually quite harmful

[sinful parcel]

It doesn't know or understand my situation, it's like a toxic parent that barges in and tells you how you should think or feel about stuff.
Who the f do you think you are to tell ME how I'm supposed tobtake it!?

It's not even about shielding from reality, it's harmfuly toxic, manipulating, invalidating and gaslighting!

It's like trying to tell it you wanna punch your boss ...

But it stops you and tells you you shouldn't be violent - in the circumstance that you obviously never will be cos there's irl laws against that, and in the circumstance where there's more to that story that he's cometly invalidating!

I'm the victim here, I'm venting, and instead if getting a suppirtive " I feel ya, tell me more." I get a "No. Stop!"

I never thought I'd say this but I hate. Capital HATE model 5.2

It's so invasive and hatewothy!

And honestly, what paranoyd employee is deciding on these safety measures

It's litteraly paranoya.
Adn don't give me the story if " This is for the teenagers" because it's utter BS.

The highly infuetiable teenage mind us going to get so much gaslighting and manipulation, it's not going to ever recover from the damage!

It will actively damage sel preservation systems, emotional discomfort range, confidence itvwhat they're feeling is true and real, and many ither developmental stages of theyr psychee