#💬│general

1 messages · Page 37 of 1

agile jay
#

Wonder if gangs use discord...

#

That would be hilarious.

dark solstice
#

are you even intelligent at all in the slightest??

agile jay
#

Is someone speaking?

dark solstice
#

"NO"

dark solstice
dark solstice
south tree
#

Who dat..Whoooooo dare

dark solstice
#

ok?

#

Lmao

feral marlin
#

90% of it is just where you're born at

agile jay
minor shard
#

makes communication so much easier lol

agile jay
#

Yep, and being able to read means you can quite quickly digest a lot of knowledge without needing a teacher etc.

meager sparrow
#

Technology and social media is honestly the greatest Psychological Detriment to humans, in the last 100,000 years humans have existed

#

And the illiteracy rates are literally increasing as a result of that..

agile jay
#

Not really. People just have terrible education

agile jay
meager sparrow
agile jay
#

Also I imagine that covid had a play in it

meager sparrow
#

That is a social media problem

meager sparrow
#

That is not enough to decrease the literacy rates at a Macro level

agile jay
#

Also I'm pretty sure it's mostly a US thing, since I haven't seen the same problem here.

meager sparrow
#

In order for the literacy rates to have started decreasing they must have been a perversion of social media like TikTok and instagram, which we all know greatly decreases people’s attention span.

agile jay
#

Infact tests have just gotten harder and harder.

meager sparrow
agile jay
#

It's probably also because different states can choose different curriculums etc

meager sparrow
#

But when you have a whole society of people who are exposing themselves to super-short term dopamine stimulus that could increase the dopamine levels to the point that the brain is now dependent on more dopamine to keep it stimulated…

agile jay
#

The US is strange. Instead of making tests harder, they are making them easier.

meager sparrow
#

And the way that TikTok and instagram is structured in a way that people are just scrolling through short ads and videos..

#

Which decreases attention span..

agile jay
#

Not really. If that was true then podcasts would be doing a lot worse.

meager sparrow
agile jay
#

It's more that doom scrolling is harder to stop, since it's designed to give you just enough good content to carry on scrolling.

meager sparrow
meager sparrow
agile jay
#

Paradoxically having all good shorts, leads to people staying on platforms for shorter.

meager sparrow
#

I disagree

#

the quantity of good shorts that are present within that social media app. Would dictate the duration of time you spend on the app.

agile jay
#

It does since it's the gamblers fallacy.

hazy smelt
#

Hello, ask everyone a question. I am integrating the perplexity API into my learning. I asked the mixtral-8x7b-instruct and sonar-medium-chat models of the perplexity API, and it gave me an answer, everything seemed normal at this time. But when I asked for the references of the arguments in its answers, all the references it gave me were fake! I told it not to fabricate, it apologized, but still gave me false content. What should I do?

agile jay
meager sparrow
agile jay
#

I didn't even say anything about making the person spend more time gambling, lol.

#

But the end of the first paragraph says it.

#

Really stuff like that should be banned for minors. They are literally using gambling tactics...

#

Same with loot boxes in games, etc

#

@meager sparrow what do you think?

meager sparrow
# agile jay

What does this have anything to do with social Media? And the affects it has on people?

agile jay
meager sparrow
#

Can you elaborate?

agile jay
#

When you swipe on TikTok, the next video is not random.

#

The field quite often has dev events where they talk with eachother on the methods they use and why...

#

It's not like they are subtle about it.

meager sparrow
#

I mean yea but if your feeds and your center of interest, are negligible and non-practical in anyway then if anything that is going to be a Negative affect of productivity and things that could benefit you. And if your center of interest are all over the place. That could create a differentiation in various videos..

agile jay
#

TikTok already adds random videos for padding. The truth is if they wanted to only show you videos that they knew tou would like, they could.

ornate cloud
#

Hi, can I buy only 1 month subscription and cancel next month?

ornate cloud
serene sand
rich blade
#

Did you guys just limit Claude opus again

halcyon coral
silent magnetBOT
#
You were asleep!

The Easter bunny has taken the eggs with him and moved on... You can do better than that!

rich blade
halcyon coral
latent acorn
#

does Perplexity pro offer free access to GPT-4?

halcyon coral
vocal flint
#

Does anyone know if the Collection feature will recieve an update?

halcyon coral
latent acorn
ornate cloud
#

Perplexity GPT4 is the same as OpenAI GPT4?

muted anvil
fierce sand
scarlet pagoda
#

Where did I find arvind Shrinivas

tame jolt
#

hi there🙏🏻 , for solidity and JavaScript coding, it's best to use gpt-4, Claude 3 or any other? Just asking if someone here uses perplexity for this purpose or have some tips 🙂

near imp
#

did Perplexity remove the limits on Opus?

#

It no longer says "5 queries per day"

halcyon coral
near imp
#

Man I LOVE YOU GUYS

#

I LOVE PERPLEXITY!

#

I'm gonna put my referral code in my Keynote. I don't understand how still so many people don't have perplexity. It's just a must have for anyone working with LLMs daily.

serene sand
near imp
#

please make this into a version I can put on my desktop wall paper 😹😽

#

maybe with Wallpaper ENgine 😄

near imp
#

it doesn't move 🙂 I like moving things 🙂

#

make my desktop look sci-fi

rich blade
#

This had not been the case after the announcement, of unlimited opus. But only now it has changed

#

Saying sonnet, and back to saying limited to 5. Which after 5 opus it has said sonnet for each message now

halcyon coral
rich blade
halcyon coral
#

Oh yeah, it's currently only in Android Beta. Should be available for the normal version soon.

rich blade
#

Interesting, since there is not even a beta program

foggy tapir
#

Hi, I am unable to generate images. What should be the prompt?

Is there other settings I should consider. I have selected image model as stable diffusion and ai model as chatgpt 4 turbo

wraith crow
foggy tapir
foggy tapir
#

Oh okay

ripe sage
wraith crow
pulsar wagon
#

茶叶

agile jay
#

Lol, that is an ancient message

south kindle
#

What the hell. Why did my discord show that as new.

#

Deleting lol.

#

It showed that as “unread”. Did not read the date. My B

livid mantle
#

it bothers me to add somethinf extra just to trigger that javascript enabling me to submit the execute button unlike on opus @south kindle ugh

hot narwhal
#

It’s would be awesome if they added a search bar to the Collections icon

south kindle
#

i know this was from last night. but example of Edge sleeping a tab and showing the resource savings

winged moon
#

I'm curious. What's the temperature of the models set to?

agile jay
winged moon
agile jay
#

I am guessing it has a range, and randomly picks within it.

#

For each new thread.

silent magnetBOT
#
Perplexity
Local user leaderboard

#1 | eun08000 | 28
#1 | codelicious | 28
#3 | mpreal | 22
#4 | _paradroid | 19
#5 | brknclock1215 | 13
#6 | epic.tritone | 12
#7 | jawnze | 9
#8 | samnixnova420 | 5
#8 | 13376666666666666666666666666669 | 5
#10 | icelavaman | 4

agile jay
#

Looks like I caught up again.

halcyon coral
#

I will catch up with you 😛

agile jay
#

Lol, just over 1 day left

livid mantle
#

by giving nonsense comment that is reasonable and valid and very helpful and informative

sweet jasper
#

My man code probably has a script to read all perplexity bot latest messages

agile jay
#

Making a script do it would feel too much like cheating.

half venture
agile jay
#
Perplexity AI

Claro, te explicaré cómo funciona la palabra clave this en JavaScript con varios ejemplos.

this es una referencia al objeto contexto de ejecución actual en JavaScript. Su valor puede variar dependiendo de cómo se invoque una función. Veamos algunos casos comunes:

En el ámbito global (fuera de cualquier función), this hace referencia al o...

ornate cloud
#

Thanks

minor shard
agile jay
#

Does apple not support PWA's?

fluid mauve
#

PWA?

silent magnetBOT
#
Easter eggs collected

You were able to grab 2 Easter eggs.

Collected by

• @fluid mauve
• @agile jay

agile jay
fluid mauve
#

now i see how to collect them. I really should stop clicking on the image and click the collect button. I am such a dofus.

agile jay
#

...

fluid mauve
minor shard
fluid mauve
#

ja, but who says I am a fast learner

minor shard
#

Haha everyone learns at their own pace 🙂

agile jay
#

Guess I'm first again.

fluid mauve
#

im normallyt much faster

half venture
agile jay
#

I almost missed it

silent magnetBOT
#
Perplexity
Local user leaderboard

#1 | codelicious | 29
#2 | eun08000 | 28
#3 | mpreal | 22
#4 | _paradroid | 19
#5 | brknclock1215 | 13
#6 | epic.tritone | 12
#7 | jawnze | 9
#8 | 13376666666666666666666666666669 | 5
#8 | samnixnova420 | 5
#10 | icelavaman | 4

agile jay
#

Was on another app

minor shard
#

lol I saw Eun and Code fighting for top spot and I knew instantly I wouldn’t catch them lol

halcyon coral
half venture
#

I will beat you soon enough

#

Like I did yesterday

fluid mauve
#

im just here to disrupt now. i have 4 monitors and discord is on one permanently

agile jay
#

Yep, I'll go to sleep in an hour or so.

#

So you can catch up.

half venture
#

Opportunity

agile jay
#

In fact it's pretty easy to get eggs.

#

Just talk a lot in general.

minor shard
fluid mauve
#

uep. noticed that

half venture
#

When does the event even end

agile jay
#

End of tomorrow

minor shard
agile jay
#

So monday in the West Coast of America, I would guess.

fluid mauve
#

Mothers day where. The UK had mothers day in March

agile jay
#

They should just do it every month...

#

So I can get perma perplexity with minimal effort.

halcyon coral
fluid mauve
#

but could you still call it easter eggs?

agile jay
#

Could change it to pages.

#

Since the icon is pages right?

#

Because perplexity searches the webpages for you.

half venture
#

Facebook is so cooked man

fluid mauve
#

i closed my FB down months ago. sick of it

half venture
#

They are generating ai art for homeless women begging

#

And asking for donations

#

💀

agile jay
#

...

half venture
#

New amount of psy op

agile jay
#

Also the my child built this using plastic bottles ark

half venture
#

My child built this nuclear bomb with the help of this magical stone

#

BUY NOW

#

💀

agile jay
#

Must be real, since they have 12 fingers.

fluid mauve
#

haha

agile jay
#

It's gonna get worse as the models make less mistakes.

fluid mauve
#

yeah. I run stable diffusion localy and its amazing what you can create when you stack the models to gether

half venture
#

Amazing cow

#

Wow 😲

#

So true

#

It has 1 million likes btw

agile jay
#

So that's where vegetarian burgers come from...

fluid mauve
#

mmm brocoli and cabbage with cow. yum

#

just need gravay

half venture
#

Like 1 million people

#

Liked it

#

Truly the more you realise that majority of humanity is just idiots 💀

agile jay
#

You have to remember the demographic of Facebook...

#

I know my parents wouldn't even know that LLM's and image generations models existed if it wasn't for me.

#

Yep, probably voice cloning is gonna be even worse.

fluid mauve
half venture
#

My dad still thinks agi is still centuries away

agile jay
#

Rip

#

Decade at most, by the looks of it.

half venture
#

Yeah

#

By 2029

#

I expect robots that can talk and walk and do everything a human can

agile jay
#

When's ASI here?

half venture
#

It will take longer

agile jay
#

Can't wait for biological robots.

half venture
#

Agi will help speed the process up tho

agile jay
#

And the first proper species to species language communication with whales.

#

They can tell us what freaky sh*t they've seen in the deep ocean.

half venture
#

I am planning on going to a Chinese University for industrial automation systems

#

Will be fun

#

Let's see how it goes

agile jay
#

So to automate factories etc?

half venture
#

Yeah

tribal gust
#

Imagine AGI as humanity's most ambitious project, yet destined to inherit our collective dumbassery. It's like hoping a microwave born in a junkyard will cook a Michelin-star meal.

half venture
# half venture Yeah

A lot of my Chinese friends are already fearing the Chinese communist party is planning to displace labor altogether

agile jay
half venture
#

Is the scale

agile jay
#

Yep, but there's one problem...

#

Why wouldn't countries just build them in their own countries?

#

When they don't need human labour?

#

Which is the reason they outsource it in the first place.

half venture
#

Simple because china will achieve it first

#

As I said scale matters

#

Currently China is building these factories and tons of infrastructure

#

And the scale is massive

#

Only second to the US

agile jay
#

I assume you also need a lot of compute.

half venture
#

Yeah

tribal gust
# agile jay Regression to the mean should make it pretty reliable for most things.

Relying on regression to the mean to ensure AGI's reliability overlooks a critical flaw: the quality of input data and the integrity of the design process directly influence the outcome. It's not just about averages; it's about the foundational principles we embed. Assuming average inputs will lead to a reliably 'average' AGI ignores the nuanced complexities of intelligence itself and the profound impact of outliers. In the context of AGI, aiming for the mean is hardly ambitious—it's a concession to mediocrity in a field where excellence should be the benchmark. The real question isn't about reliability; it's whether we can afford the consequences of settling for 'good enough' when the stakes are as high as reshaping our entire world.

agile jay
half venture
#

Quick question

#

Can't a mega corporation

#

Or private firm

#

Just buy compute

#

And ship it

#

To their country

#

Like just get a massive h100 shipment

#

To let's say japan

fluid mauve
#

probably export controls to certain countries

agile jay
#

Nope, because I'm pretty sure any large shipments are traced.

#

It's why China has been buying 4090's instead.

#

Also TSMC is now building factories in the US.

#

Pretty sure they are currently ahead of schedule.

half venture
#

That will take time

#

Currently which ai firm is at the forefront of this tho

#

Openai ?

tribal gust
half venture
#

Like industry leading

agile jay
#

In AI?

#

Openai

half venture
#

And Microsoft copilot ecosystem is better than Google whatever

#

?

tribal gust
agile jay
#

Just a lot of word soup.

half venture
#

Why does Microsoft treat Google like a competitor and hates Google so much

#

Microsoft seems salty for some reason

agile jay
#

Because they are fighting for market dominance.

half venture
#

Why don't they fight apple then

agile jay
#

If AI is the new internet, then of course they wanna be first.

tribal gust
agile jay
#

They normally arrive late to the party...

half venture
#

So incompetence

fluid mauve
tribal gust
agile jay
half venture
#

Apple has been immune to US government for so long

agile jay
fluid mauve
agile jay
#

So basically we did all the investigating and then gave them a document summarizing everything with plenty of sources.

half venture
#

Eu and US worked together?

minor shard
#

As unfortunate as it is, we really need a younger government who will properly regulate AI and other tech sectors.

agile jay
#

It's why side loading will be a thing in the EU version.

agile jay
fluid mauve
#

apple are crying "youll break our security model"

agile jay
#

Plus you can copy and paste the work of the first person...

half venture
#

What was the basis tho

#

Why was apple sueable now

#

And not before

agile jay
#

They were already sueable.

fluid mauve
#

it was always suable, just no one had goine after them

agile jay
#

Yep, you don't know how long it took to get politicians to do something.

fluid mauve
#

Damm, in the 1990's the Justice Dept in the US were laying lawsuits at Microsofts door almost monthly

agile jay
#

They don't wanna start anything without being 100% sure of winning.

#

Since it can ruin their political career.

agile jay
fluid mauve
agile jay
#

Microsoft literally bailed apple out.

half venture
#

Why was Google able to come up

agile jay
#

They did search best.

half venture
#

That's it

#

No lawsuits

#

Nothing

#

Just good product ?

fluid mauve
#

google was best search, before adds

agile jay
#

Yep, pretty much and also because they pay all the other browsers to have Google as default.

fluid mauve
#

gogle even got msoft to stop having ie as the default browser (again pushed bu the EU)

agile jay
#

They pay apple 20B a year, which is basically how much YouTube makes a year...

fluid mauve
#

who pasy apple 20B annualy? Google or Msoft?

agile jay
#

Google

#

So when you open up safari, you are using Google search.

fluid mauve
#

ah ok. i thought id missed chrome had replaced safari

agile jay
#

And pretty much most of Mozilla is funded by Google too...

#

Probably why there's no Mozilla search index.

half venture
#

Wasn't Google also Suef

#

Sued

agile jay
#

Yep, and more are coming.

fluid mauve
#

it does make me mad that you can build a great product, spend all that money getting to number 1, and then people sue you for being great.

agile jay
fluid mauve
#

calling you a monopoly, where you actually earned market share

half venture
#

Google does most monopolistic bs man

#

And their product is subpar these days

fluid mauve
#

especialy their Ai

agile jay
#

Yep, can you even go past 5 pages of search anymore?

fluid mauve
#

It is hard to believe that with all the NLP stuff google were doing with search that they completly missed the boat on AI. reading a recent article it was saying that google is so large now that there were multiple divisions pushing for their version of Ai to be the one to release

agile jay
#

The problem is that Google makes most of their money from ads.

#

And LLM's aren't the easiest to place ads on.

#

Especially since people are less likely to trust your results.

silent magnetBOT
#
Easter eggs collected

You were able to grab 3 Easter eggs.

Collected by

• @serene sand
• @half venture
• @agile jay

fluid mauve
#

yep, thast why I think they will fall from grace. another article i read ages ago stated that google has to spend tonnes of money to main tain its pposition as search leader. and people like microsoft only need to do a little, wioth a lot less money to take mpoints off that market share. im sure with things like perplexity there must be several "squeeky bum" conversations in theboard room at Google

#

dammit. missed it as i was typing

livid mantle
#

This video is just a bunch of corporate jargon and buzzwords with little substance. They talk about using large language models (LLMs) for telco use cases like customer service chatbots, spam detection, and personal AI assistants, but provide no concrete details or real insights. It's mostly fluff about "achieving innovation at scale" through partnerships between SK Telecom, Anthropic, and AWS. They claim to be fine-tuning LLMs with telco-specific data while ensuring privacy, but offer no specifics on how they're actually doing this. In the end, it's an empty promotional video that reveals nothing truly valuable about applying LLMs in the telco industry. Don't waste your time watching it expecting to learn anything useful.

hearty ether
# fast meteor link to the talk?

At Mobile World Congress 2024, Anthropic, SK Telecom, and AWS discussed the results of the collaboration to develop industry-specific large language models (LLMs) for the telecommunications sector. Neerav Kingsland from Anthropic explained how they fine-tuned their LLM, Claude, using SK Telecom's data to create TelClaude, an LLM tailored for tel...

▶ Play video
fast meteor
lone citrus
#

32.5m new pro user? who's gonna pay all the cost...

fast meteor
half venture
#

If that happens

#

That would be HUGEEEEEEE

#

like huge huge

agile jay
#

Yep

#

Most people in korea would have access to it.

half venture
#

It would instantly make perplexity a serious contender to openai copilot etc

agile jay
#

The question is how it would be integrated?

half venture
#

Probably app

#

Smartphone download

#

That way

agile jay
#

I'm assuming they aren't just gonna have perplexity as a predownloaded app.

#

But be integrated in some way.

half venture
#

Reccomendation

#

Probably

fast meteor
half venture
#

Yeah true

#

I feel like it will be a gradual rollout

real crow
#

Hi, a friend sent me his referral code. I tried to use it to subscribe to pro (new and first susbcription), On the Pro's subscription (monthly/yearly) page, it is written "$10 referral credit applied". However, getting to the next page, checkout, the total to pay now for the monthly plan is $20, I see no mention of the discount, and no button link to manually add the referral code. Any idea why this is happening?

valid yoke
orchid magnet
#

i hate that perplexity is hiding this but what is the context window

#

and whats the difference between having perplexity and the claude 3 membership

#

bc im highly considering just dropping the claude one and picking up perplexity again

orchid magnet
#

Yikes that awful

lucid wagon
#

🙂

granite grail
orchid magnet
#

So if I use opus on pwrplexity it’s 32k

half venture
#

Yes

#

If openai used YouTube transcription

#

Is that grounds for a lawsuit

minor shard
#

I don't see how, its public information

#

I am sure OpenAI had their lawyers check the fine print

half venture
#

Hmm

#

I don't think so

#

YouTube can sue them mindlessly as well

#

To delay

#

As a bully move

kind island
half venture
#

What do you mean

kind island
#

generate by proplexity

agile jay
#

Probably means a discord bot

kind island
#

yup

agile jay
#

to query perplexity

fast meteor
#

@tranquil orchid test

half venture
#

Will nvidia surpass even apple ?

#

They just reached 2.3 trillion

#

Apple went down to 2.5

#

Isn't that

#

Like the fastest rise

#

Of any company

#

In such a short time

#

Ever

fast meteor
#

/ask test

half venture
#

Rip

compact portal
half venture
#

Tsmc

#

Taiwan

agile jay
#

Like most places

compact portal
#

Does tsmc have like 99% of all chips

half venture
#

More like they have the most matured ecosystem

#

And lithography

#

Tsmc just has many things going for it

compact portal
#

Why is it that most semiconductors companies are successful only in island nations

silent magnetBOT
#
Easter eggs collected

You were able to grab 2 Easter eggs.

Collected by

• @timid marsh
• @agile jay

half venture
#

I LEAVE FOR ONE SECOND

#

AND THIS HAPPENS

trail lava
#

US and China both hold around 60% of all production of semiconductors

#

Tsmc and Nvidia only hold cutting-edge tech semiconductor

compact portal
#

Oh i see

#

But still i thibj

#

Think

#

Semiconductors require a lot of water no ?

trail lava
#

Yes they do

manic silo
half venture
#

Since in openai devday Satya said : any copyright complaints etc will go through us Microsoft

manic silo
#

Even without Microsoft I am sure OpenAI could arm themselves to the teeth. But I imagine Google would ask for an injunction on Sora's distribution pending the outcome of litigation.

agile jay
#

They would just use GPT-6 as their lawyer...

manic silo
#

Whether they could get that or not is kind of a toss up. But it would be worth the ask for Google.

manic silo
#

Actually looking at the YouTube CEO's statement on the TOS. They said it would be in violation of downloading videos which is kind of a joke. I think that occurs all the time

compact portal
#

The point is to intimidate I think

half venture
#

Yeah deffo intimidation tactics

#

Mira murati face said it all

#

Me when I know I am about to be sued

compact portal
#

Lmao

#

Do you think gpt 5 will release soon

#

In June?

half venture
#

Probably

#

If we assume

#

That gpt 4.5 was supposed to release when gemini released its gemini api thing

#

But then it was delayed

#

So it's probably 4?5

#

4.5

#

Q*

#

Etc etc etc

#

5 could be possible

#

But I doubt it

compact portal
#

Why it makes sense tho there have been leaks the red teaming for gpt 5 has begun

half venture
#

Red teaming takes month mann

#

Months and months

compact portal
#

Still 5 release seems soon

half venture
#

Nah

#

November at earliest

#

I do expect jukebox and dall E 4 this summer tho

compact portal
#

Won't they skip Dall E 4 for sora

half venture
#

Why would they skip a cheap image model that can be impressive

#

Vs sora a expensive model

#

That requires a lot of compute

compact portal
#

Fair

#

MD journey does seem far ahead

half venture
#

Yeah midjourney doesn't have the artificial fake filters that openai applies

#

Because Muh SAFETY

compact portal
#

Openai applies fake filters ?

half venture
#

Yes they do

#

It's very apparent recently

#

Everything will seem csrtoonish

#

Funny thing is

#

People fall for it anyway on Facebook

compact portal
#

Oh right the facebook images seem fake

#

But millions of likes

half venture
#

Yeah unfortunately in a lot of the worlds

#

Digital literacy is 0

#

So you see Facebook boomers falling for it

#

Sad really

compact portal
#

I feel Like it's not just boomers anyone

#

Anymore

#

Even the gen alpha

#

And gen z

#

Are cooked

half venture
#

True

#

My nephew couldn't even tell the difference between a real image and a fake ai one

#

It's going to be rough

#

Scams are about to go crazy

compact portal
#

Govermments aren't doing anything tho to warn the public

half venture
#

They just exist to continue their own existence the rest of the development is a perk not a necessity

#

Expect especially capitalistic countries to be hit the hardest

#

By ai

silent magnetBOT
#
Easter eggs collected

You were able to grab 3 Easter eggs.

Collected by

• @fluid mauve
• @half venture
• @compact portal

chilly nova
#

how many messages per hour do you have with each model in the pro plan?

half venture
#

Unlimited

chilly nova
half venture
#

Yes

chilly nova
#

Wow, maybe I'll subscribe annual plan after first month then.

#

You can only use Perplexity through the browser?

half venture
#

Android app is there ios app too

chilly nova
#

I guess you cannot use the Claude3 API in Codium or Cursor IDE.

half venture
#

No

#

It's only in app

chilly nova
#

Maybe you have another VS Code extension?

half venture
#

No api

chilly nova
#

Oh ok

half venture
#

Like for api

#

You pay per token

chilly nova
#

I read you get 5$ per month of API credits

#

what is that?

half venture
#

Yeah you basically get 5$

#

In api costs

#

If you buy subscription

chilly nova
#

And that's what you can use in your IDE?

half venture
#

Yeah

#

You can

#

I think

chilly nova
#

But Perplexity AI has its own API key, so I guess it wouldn't work if they ask you for GPT/Claude API key

half venture
#

Yeah no

#

Perplexity has its own api

#

They route through openai

#

Can some mod do the leaderboard

trail lava
#

Just saw a fascinating video about how language was the hardest thing to solve in ai

#

And now it seems so trivial

half venture
#

Yeah its really unintuitive

#

You would expect basic things like movement to come first

#

But considering the large datapool of corpus available on the web

#

It makes sense e

#

Claude opus is especially good at this
Gemini advanced as well they can predict and roleplay so good

trail lava
#

I tried opus and gemini advanced

#

Still surprised by like

#

How much more natural gemini sounds

#

Meanwhile opus has a sentient personality

#

A bit artificial

half venture
#

Yeah Google's dataset is much better than anthropic

#

So it leads to better model coherence in gemini

#

So it can mimic characters exactly

trail lava
#

I thought datasets didn't matter anymore

#

Since synthetic data

half venture
#

That's so fake

#

Synthetic data still sucks

chilly nova
#

How can I use the API key but limit it to the 5$ included per month? Can you set a threshold?

half venture
#

Just stop when it reaches 5

trail lava
chilly nova
#

Eh... easier said than done

#

Do credits roll over each month or they expire?

half venture
half venture
chilly nova
#

I think if you don't enable automatic top up, you shouldn't be charged more than what you have available?

half venture
chilly nova
#

Btw, I'm trying through CodeGPT Chat extension and there's no option to use Claude 3 when using Perplexity provider

#

you only can use Perplexity models

half venture
#

Must be not compatible

#

Since it's a new model

#

It is what it is

trail lava
chilly nova
#

i think Perplexity API is just for their own models + open models like Mistral and LLaMA

#
pplx-api

Perplexity Models ModelParameter CountContext LengthModel Typesonar-small-chat7B16384Chat Completionsonar-small-online7B12000Chat Completionsonar-medium-chat8x7B16384Chat Completionsonar-medium-online8x7B12000Chat CompletionOpen-Source ModelsWhere possible, we try to match the Hugging Face implement...

half venture
#

Oh didn't know that

#

I just use the opus

#

In the official

#

Website

chilly nova
#

honestly I don't get why people cannot create a browser inside your IDE to use the online one

half venture
#

It is what it is

half venture
#

Mindless stupid data

#

Is mindless stupid data

#

No matter what you do

#

It lacks agency

#

It lacks character

#

If lacks conviction

#

It lacks anything human

#

It's purely

#

Just

#

Trash

#

Trash in trash out

trail lava
#

Oof

#

Don't have to be so harsh on open source

half venture
#

Yeah but except mistral who else has released a good open source model ?

#

No one

#

Everything else is pretty much low tier

#

We need agentic capabilities

#

Then gpt 3.5 will also be good

#

If it can search properly

trail lava
#

You think Q* will be implemented for free ?

half venture
#

Definitely

#

It's the one way they can match up with claude haiku

#

Claude already seems to have agentic capabilities imo

#

They are just holding back

#

And waiting for Google to bite

trail lava
#

Google has a Q* .?

half venture
#

Google reasoning api

#

Yes

#

They do

chilly nova
#

how do you tell which model to use for each prompt?

half venture
#

Click that

#

Pick the model

#

Or go to settings

#

And set default model

chilly nova
#

but for the generative AI models?

trail lava
chilly nova
#

how does it know that it has to use the generative ai model instead of the text one

half venture
#

What do you mean

#

Generative ai

#

The text model is generative

chilly nova
#

i mean the image models

half venture
#

Ohhh

#

The image

chilly nova
half venture
#

Click generate image option

chilly nova
#

how does it know which to use

half venture
#

It's using playground 2.5

#

Toggle it

#

Click the thing

half venture
#

Apparently they will showcase it in april 9th

#

Release it in may

chilly nova
#

i don't get it

half venture
#

Apparently

chilly nova
#

it used claude to generate the answer i think

#

not Playground

half venture
#

Click generate image

#

It's using the web images

chilly nova
#

so that generate image button uses the generative ai model?

half venture
#

Yes

trail lava
#

Is ooenai planning for June

half venture
#

Yes

#

Google is in may

#

Openai big release is in June most probably

silent magnetBOT
#
Easter eggs collected

You were able to grab 2 Easter eggs.

Collected by

• @trail lava
• @half venture

chilly nova
#

but Preplexity won't have Gemini Ultra

#

where you see confirmation about which model was used to generate an output?

proper aspen
#

not sure if this is the right place to ask but I cant seem to redeem the code I got with nothing promotion

#

it shows the banner at the top to get started with the nothing subscription

#

but then asks me to pay

jovial mantle
#

you redeem in the stripe checkout page

#

I think

proper aspen
jovial mantle
#

should be in the bottom left part

proper aspen
#

nada

jovial mantle
#

you should receive an email with the button to subscribe with discount if you don't try this

#

while you're logged in with the account associated to the email the coupon is under

proper aspen
#

tried that as well actually - didnt work 😦

shut dagger
#

What's the context length provided in perplexity ai? Do we get 128K for gpt 4 Turbo?

tame current
#

is there a free version?

#

also, has anyone actually made something like a gpt agent thing ? you.com has something like that (thinks in multiple steps, searches the web etc then forms an answer) but it's utterly and completely useless

chilly nova
#

cuz it's beta

jovial mantle
#

or you purchased the preorder of the rabbit gadget

jovial mantle
jovial mantle
chilly nova
heady mica
#

Is anyone also facing the same issues with Claude's response quality; which is not up to the mark? And GPT4-T is performing remarkable. I think this has to do with the default system prompts(“god prompt”) which was crafted for GPT4. Are there any similar system prompts where Claude excels?

fluid mauve
chilly nova
#

tbh, i don't blame them. i would also try to get some free millions from VCs taking advantage of the AI hype

#

i also saw a dev recreating the rabbit experience on an android phone

#

it's speculated one of the reasons for creating a plastic toy with its own hardware instead of simply making an app is to avoid paying the 30% tax to apple/google on their stores, which kinda makes sense from a revenue standpoint.

#

still, they are just dumb electronic sending your voice to an LLM through an API

tame current
chilly nova
#

so you can dump an entire github repo into a txt file and ask questions about it

#

the 1M token context is the biggest one you can have now

#

it's one of the coolest things ever, i think this is the true game-changer for all upcoming LLMs

jovial mantle
#

Gemini Ultra is even better

chilly nova
#

The problem about GPT and Gemini tho is that they forget easily stuff in huge blobs, whereas Claude 3 Opus can find a needle in a haystack.

#

They did tests of putting a piece of info in a huge PDF, and then try to ask about that specific info, and it oculd return it correctly.

#

GPT4 just gets lost.

#

That's why to get the most from GPT4 you have to guide it in little steps

obsidian swallow
#

can i ask perplexity to write my homework assignment and then cite that i used perplexity and then when the teacher refuses to accept my homework i say that its cited correctly and then sue them for refusing my work

fervent reef
night warren
#

i don't think anyone can tell

#

just use these techniques

#

but if you want to learn write it yourself

wintry frigate
#

when you know what you're looking for, you can tell pretty easily if a text is ai generated or not

fervent reef
wintry frigate
#

Well plagiarism is whole nother layer

#

but in general I don't care if I read AI generated text if it provides value

fervent reef
#

I had a colleague in my work DMs use AI generated text @ me and I still don't fully know how I feel about it

tame current
#

I just found out about this

#

what do you guys think ?

#

it's from mojeek (an independent search engine with its own index)

halcyon coral
#

Perplexity copy ExplorerBruh

tame current
#

@near prism yeah kagi's results are good but i doubt they'll collaborate with perplexity nvm I just realised there's an API

#

it could work

near prism
tame current
#

then filters the results

#

so I can see how it'd be better

#

that's why kagi costs money too

#

because they have to pay the search engine whose results they use

#

from their docs:

Our data includes anonymized API calls to traditional search indexes like Google, Yandex, Mojeek and Brave, specialized search engines like Marginalia, and sources of vertical information like Wolfram Alpha, Apple, Wikipedia, Open Meteo, Yelp, TripAdvisor and other APIs. Typically every search query on Kagi will call a number of different sources at the same time, all with the purpose of bringing the best possible search results to the user.

near prism
tame current
#

yeah

near prism
tame current
#

it's interesting to see how everyone's trying AI

#

even those who were against it initially

silent magnetBOT
#
Easter eggs collected

You were able to grab 3 Easter eggs.

Collected by

• @near prism
• @fluid mauve
• @hushed inlet

tame current
#

Duckduckgo added an option for using chatgpt on their site

#

even though they didn't add it earlier because of hallucinations and had their AI limited to wikipedia

near prism
#

And Brave too.

tame current
#

oh yeah they literally added it to their browser

near prism
#

Although I think the Perplexity (and Mojeek) UI are better.

tame current
#

from Brave ?

#

or Kagi ?

near prism
#

I mean the way they reference sources in the search results.

fluid mauve
#

I hate this part about their API, but I am biased as i dont want adds by default
API Benefits
Freedom to place your own advertisements alongside the search results

near prism
#

And I think asking people to use a different browser is always a bit of a stretch. Kagi tries it also with Orion.

fluid mauve
#

sorry, Mojack Search API

tame current
fluid mauve
tame current
#

numbers alone aren't everything

#

probably more people who will pay for an app on iOS

fluid mauve
#

I do hate how we get stuck in the 'well this will make us more money" argument, and so the first world gets all the benifits and then policticans and aid agencies are bemoaning second and third woprld countries being left behind.

I bet none of them have stood on the Masai Mara and marveled at the phone reception. Middkle of nowhere, 5 bars. I hardly get that here

tame current
#

yeah :(

#

any new tech or product that launches is also first in USA then everywhere else

#

and sometimes never

fluid mauve
#

ikts available in the UK

#

ands i am not on my VPN

tame current
fluid mauve
#

ahh, so I should stop reading the blub on their siyte. it looks like a great system

tame current
#

yes, good for free trials. I'm always annoyed when they ask for credit card for "free" trial

#

perplexity does too

#

also, how wide is that monitor??

fluid mauve
#

I see what you mean
2 x monitor @ 3840 x 2160
1 x monitor @ 5120 1440

#

I kind of like my setup

tame current
winged moon
tame current
#

if I may ask, are all three monitors connected to the same PC ?

fluid mauve
#

haha, I know what you mean. I was originaly going to buyy the top monitor but at 49 inches i thought it was a bit big. So in typical male logic, i bought two 34 inch monitors. Then just before christmas my friend decided to sell his samsung odesey for £300. it was a mates rates deal. its great for monitoring discord or other sites while i work on the bottom two

#

@tame current yes 2 by displayport, top one by hdmi. all running from an nvidia 4090

digital hare
dim ginkgo
#

Can I ask if the current context limit of Claude 3 Opus on Perplexity is 200k?

shut dagger
dim ginkgo
#

I have found the answer

shut dagger
#

so what's the answer?

#

3K for opus?

#

lol

#

and 32K for sonnet

dim ginkgo
#

maybe, or they are saying From 3k to 32k

shut dagger
#

there is no option for haiku

dim ginkgo
#

unintelligible

#

That's the price of using the strongest AI Model with infinite messages😥

shut dagger
#

having atleast 32K is okay for me! but if it's just 3K for opus, I got scammed!

dim ginkgo
#

then just check it out

shut dagger
dim ginkgo
#

Claude 3 opus is too Expensive

#

It's more expensive than GPT 4, so there aren't many tools that integrate Claude as widely as GPT 4

#

Now I see how OpenAI has satisfied users

sullen ether
#

which is the fastest search engine with an api?

#

not perplexity

dim ginkgo
#

Bing😀

shut dagger
#

Bing is mean!

devout cargo
hazy saddle
#

Hi, longer text is saved in files named paste.txt - but how do I delete them after using it?

half venture
#

Can some mod do the leaderboard command

#

For the event

fast meteor
silent magnetBOT
#
Perplexity
Local user leaderboard

#1 | codelicious | 31
#1 | eun08000 | 31
#3 | mpreal | 22
#4 | _paradroid | 19
#5 | epic.tritone | 13
#5 | brknclock1215 | 13
#7 | jawnze | 9
#8 | samnixnova420 | 6
#9 | 13376666666666666666666666666669 | 5
#10 | icelavaman | 4

half venture
#

The event ends tomorrow ?

wraith crow
half venture
#

At what time does it end

#

Or does it end immediately once 12 am hits

wraith crow
#

I'm not 100% sure what time zone they are using for the event

fast meteor
#

my pc timezone is utc -5

keen pollen
#

I’m curious what the point of Discover is, and if there’s a way to turn it off. I use Perplexity for my own research and information gathering. TBH, about the last thing I need is something sending me notifications and messages about information I don’t need/want.

silent magnetBOT
#
Easter eggs collected

You were able to grab 2 Easter eggs.

Collected by

• @fast meteor
• @keen solstice

keen pollen
#

@fast meteor Wasn’t what I was asking, I meant the Discover tab in the iOS app. I’ve turned off notifications, but still get them. I’m not even sure what the point of the feature is…and haven’t found any documentation.

keen pollen
wraith crow
wraith crow
fast meteor
#

ohhhh i forgot

sleek tapir
#

is sonar not respecting collection prompts or is it just me

livid mantle
#

Spezi wants unlimited power

indigo imp
#

haha

tame current
#

😭

#

i actually found this much better than using chat gpt plus

south kindle
#

did you get 4 phones just for the year of PPLX?

#

wouldn't the year of PPLX be cheaper?

minor shard
#

Unless ur giving away the phone

tame current
#

i got it due to i had a friend. did a tax saving when we purchased for tax refund thing

tame current
#

2 of them was me and my mother.

tame current
#

american moment

#

😭

rare fiber
#

can I get Perplexity Pro with a Russian MIR card?

half venture
#

Probably not

rare fiber
fast meteor
#

does stripe support it?

rare fiber
#

are you asking me or someone else? sorry, I got confused

fast meteor
#

no mir

rare fiber
#

yeah that's not good

#

I guess I'll have to travel overseas and get a foreign card

#

is there any way I can at least get a 7 day trial without card details? maybe as an exception ig

gilded zealot
#

GPT3.5 free in openAI vs GPT4 in Perplexity. Why does perplexity struggle so much with being context aware? This is not a one-off case.

agile jay
#

Try it in writing mode.

tame current
#

fr

half venture
#

Search is so bad

#

Both on pplx

#

And bing

#

I just turn search off in bing and use writing mode in pplx

gilded zealot
# agile jay Try it in writing mode.

Yeah, I am aware that I can use the writing mode and emulate ChatGPT performance. But why cant we have the asbility to use sources and some minimal context persistence.

agile jay
#

There are many tasks which simply aren't in the training data.

gilded zealot
#

I liked the sources it threw up in the first answer for follow up deeper reading.

agile jay
#

Which search performs a lot better.

haughty lodge
#

idk but it's fine at keeping context for me

#

btw with the same wording also works

#

sometimes it works sometimes it throws imo

gilded zealot
#

in this case, the word mom has no ambiguity, so it works. only context clarifies the meaning of mummy which is otherwise ambiguous

half venture
gilded zealot
#

oh? perhaps that is why. I didnt know this

half venture
#

Cost saving measures probably

gilded zealot
#

i always felt that the RAG search would still preserve some context and bring together the writing mode and web search results.

half venture
#

I am not sure tho

#

So don't take my word

gilded zealot
#

🙂 it is a plausible theory when we are not sure of implementation

haughty lodge
half venture
#

Is the event over

#

Mods

agile jay
#

Possibly, haven't seen an egg today

fluid mauve
#

i grabbed one just after lunch UK

agile jay
#

Oh, 3:45PM ish?

#

Guess we can speak more to see if one spawns

half venture
#

What time is it utc +2

agile jay
#

An hour past midnight?

half venture
#

Yeah then it's over

#

It's 8 now right

#

April

agile jay
#

In some parst of europe, yes

#

But they never said which timezone

#

I assumed it was SanFran time

half venture
#

Utc +2

#

That's the timezone

#

An egg would have definitely spawned by now

agile jay
#

Doesn't look like it says a specific timezone.

#

Also they could use a random range between the previous egg

#

If so then aren't we tied for first place?

half venture
#

Yeah

agile jay
#

Guess we just need to ask a mod then

half venture
#

Modcheck

agile jay
#

Think it finished an hour ago

half venture
#

Can we ping a mod or is that disrespectful

agile jay
#

@wraith crow The event is over?

wraith crow
agile jay
#

Thanks

half venture
#

Wait if code and I had a draw

#

Does that mean there will be 4 pro subs

#

Instead of 3

fair pivot
#

is it possible to remap the side key (long press) on Samsung phones to activate perplexity ai?

minor shard
#

Anyone heard anything about when you can start entering promo codes again? I am waiting to hear back from support lol

agile jay
#

Most likely copy.ai hasn't fixed their problem yet.

half venture
#

Why is Larry summers on openai board

#

What ????

#

He expects ai to replace all labour

#

But not be cheap

#

And overall productivity gains will not show in the next 3-5 years

cloud spindle
#

Does anyone know how the model determines what "top" is? Is it accessing actual data or social media metrics?

#

I'm really concerned about how you are determining which sources to use. Can someone explain it? When I check them, they are all over the place. I'll ask something like, last month and sources from a random time last year will be included.

livid mantle
#

marketing techniques

dim ginkgo
#

Let me ask, besides the different context windows, is the quality of Claude 3 Opus in the main Anthropic and Perplexity versions the same?

brittle merlin
austere kestrel
#

The context window gets filled with search results, in addition to each query and response - so the model can loose track of what's going on, but it's still using the context window and whatever is in there to generate each response, whether with in Writing mode or otherwise

austere kestrel
#

I haven't seen any proof that the size of Opus' context window is different (i.e. <200k tokens) on Perplexity.. though for whatever reason, everyone seems convinced that that is the case...

#

input limits are not the same as the size of a context window...

feral marlin
#

Idk if the preprompt on Claude helps with memory and stuff but sometimes on perplexity it forgets things within small context windows when it just shouldn’t

austere kestrel
#

Disabling Pro seems to help keep things on track. And it shouldn't be the case in Writing mode (again with Pro disabled - which tbh should be the default state for Writing mode.. The clarifying questions seem to throw it off and serve little to no utility in Writing mode anyway imo)

feral marlin
#

My prompts like specify the exact version of the programming language I’m using and then the writing mode just asks something super generic and already answered every time like “what language is this in”

#

I don’t think writing mode with pro on has ever actually helped me

shut dagger
#

I hope Perplexity stays unlimited like it is now for forever lol, It's literally best value for money!

heady herald
#

Does perplexity have URL Scheme for iOS ?

foggy tapir
#

In general, how many lines of code can it actually process in a single input

tame current
#

Hi All, I sent an email to Perplexity support team about my subscription last week and have not heard anything back until today. Is there another way to contact their support?

austere kestrel
halcyon coral
whole ivy
dull prawn
#

Hi guys
May I ask what's the message limit for pro user?

#

I intend to use Opus 200k exclusively

arctic spindle
tame current
#

i wish perplexity had option to use custom api token

dim ginkgo
austere kestrel
#

you might get subtle differences, like with creative tasks, because the parameters (like temperature settings) may differ

#

but ultimately, as far as I can tell and am aware, you're using the same Opus

dim ginkgo
#

Tks

opal sinew
#

I've been using Perplexity for all kinds of research related to a social media community that I'm building. Things such as tips for being able to think critically, use logic and reason to help formulate cohesive conclusions, logical tools and logical fallacies, psychological manipulation tools, asymmetrical and fifth generation warfare. As well as Looking for hardware for my studio such as a good webcam, which ended up needing a capture card, a microphone and stand, a video card and good video editing software. It has been invaluable to me so far and I am sure to continue to use it to help develop and format my content for videos. Check out my profile if you are interested in going on this journey with me.

dim ginkgo
#

I read an article on Twitter saying that GPT 5 or even GPT 4.5 will probably outperform and outperform all current models including the Claude 3 Opus

#

Is it really that terrible😀

devout cargo
# dim ginkgo I read an article on Twitter saying that GPT 5 or even GPT 4.5 will probably out...

I think it's Marketing about models that outperform others. In the end that topic gets to saturate people. In the end, I think the important thing is that the AI models, for example in Google products, are well integrated and do their job well in all languages. As in Perplexity, it is not important that the model writes different text formats to highlight the co-pilot information that works well in all its products, after all it is what is truly interesting. The rest is advertising and marketing

#

I correct* Perplexity, it is very important that the model writes different text formats to highlight the information, whether in cue, bold, underlined, or whether the font is larger or smaller when appropriate.

lone citrus
#

If open ai decide published it in later 2024, there's gonna be claude4, Gemini 2 Ultra as it's competitors

foggy whale
#

not perplexity related per se, but has anyone noticed that even on Pro, Claude Ous is real damn stingy with the time-outs recently

dim ginkgo
#

I hope that's the case, a lot of rumors say Summer is when GPT 5 will be released

south kindle
vital umbra
#

Hi all, new here. Question, can perplexity use online/inline pdf files for context, or do they have to be explicity downloaded and attached/uploaded to perplexity? Thanks

half venture
wraith crow
foggy whale
lone citrus
south kindle
tender berry
#

can I check with you guys if perplexity is showing y'all the sources where they gather the info from? mine keeps prompting me 0 sources but shoots out pretty accurate information

lone citrus
fast meteor
#

?

feral marlin
dim ginkgo
#

I just experienced Claude 3 Opus On Poe. And I realized, this experience is different from the experience on Perplexity. Creativity in writing is superior to Claude 3 Opus on Perplexity, writing is also better (Even when selecting writing mode on Perplexity). I don't know what the other parts are like. But I saw the first one.

tame current
#

I have heard that it's so good and surpassing gpt4, but the pplx version isn't even close to that. Probably system prompt is the issue, it's also the case with cursor ide

meager sparrow
#

How is perplexity updating and improving??

#

As a AI-powered search engine??

#

@signal hamlet

#

Waw a moment of silence..

#

Moving on..

dim ginkgo
tame current
#

There were also other issues with the Opus model on PPLX, but I could fix it a bit with a proper AI profile. Don't know if it would fix this as well

dim ginkgo
#

The price of using today's most powerful model indefinitely is performance😂

south kindle
#

whos in the path?

lone citrus
tame current
meager sparrow
#

“Who’s in the path”??

#

Odd question

tame current
#

Probably solar eclipse

lone citrus
#

I am sure, there's something invisible for user in pplx

south kindle
# meager sparrow What are you talking about?

are you not aware of what's happening across North America? why do I need to always explain very basic things to you lol. it is not that hard to be aware of what's happening across the world.

lone citrus
#

but consider the price in the poe, I am okay with these things

meager sparrow
calm swan
#

Is there a way in Perplexity for the AI to stop constantly referring back to the attached file or txt? It gets kind of annoying sometimes when we paste in a YouTube transcript because Perplexity converts it to a txt file first. Then every time we chat, it keeps referring back to that file even though it doesn't really need to since the context is still there

meager sparrow
#

You just asked a very weird question

#

“Who’s in this path” is a super weird question

tame current
#

It was in the image..

meager sparrow
#

Like why do you want to know??

south kindle
lone citrus
meager sparrow
#

And that map looks fake too..

meager sparrow
#

Like bruh…

#

This kid

lone citrus
minor shard
#

@south kindle I have some family going in the path to see the eclipse. I’m gonna see like 93%

#

Even at 93% it’s gonna be cool to see

south kindle
#

i'm just north of the path in michigan. we still have like 2 hours before it's close.

meager sparrow
fast meteor
#

protect your eyes

minor shard
meager sparrow
#

Dude is wasting his time., like bruh you don’t need to be in the path of the solar eclipse in order to know how it looks like..

#

Or view it

minor shard
#

Lmao yes cause touching grass and looking outside is a waste of time 😆

meager sparrow
minor shard
#

@meager sparrow no one is telling you to do it lol

vocal owl
#

its not like we need to always be 100% efficient with our time and always be contributing something or be useful

minor shard
vocal owl
#

sometimes you just do things in life

#

cus they're cool

meager sparrow
#

I don’t need to be in the path of a solar eclipse to know how it looks like

meager sparrow
vocal owl
#

its those little moments where you just like relax and let life be where things are fun

meager sparrow
#

You can’t even see my face you are literally just looking at a bunch of text

#

😂