💬│general

[hollow obsidian]

Like you write:
" How is the weather today?" And the copilot asks you "please tell me the location" in a extra text field.
Then it generate a answer.

[wraith crow]

copilot also uses more sources to gather information from before returning an answer

[hollow obsidian]

@wraith crow can you tell me this?

Btw. Thanks for being helpful 😄

[wraith crow]

Chats from users cannot go there. "Discover" only shows automated threads created by a bot. I don't know exactly what the requirement is to show a specific topic.

[hollow obsidian]

Ahh thanks. I asked something private and eas afraid it get listed there xD

[wraith crow]

You don't have to worry about that, your chats will stay private unless you share the url

[hollow obsidian]

Btw.. does the bot answer all as example if I would ask how to make a baby, would it explains me the coitus or is there a "we don't teach you things about your body" rule?
This is sadly often a problem.
That sexual education isn't allowed.

[wraith crow]

You can ask Perplexity all kinds of questions as long as it does not break the TOS https://www.perplexity.ai/tos

[hollow obsidian]

OK. So don't ask how to build a bomb xD

But how love does work is okay 🙂

[hollow obsidian]

😢

So if it is too NSFW its not okay

Always the same with AI 😢

Getting suggestions how to make a roleplay with love with an ai is not allowed 😭

[agile gazelle]

@latent wharf I attempted to add you to my server. You joined but you are not responding. What am I doing wrong?

[hollow obsidian]

Hell... The AI complained that I used nice words for the private parts and I had to use the real names.. that was okay!?

[plush shadow]

Are perplexity threads saved synchronously across all devices, such as if I make a thread on my mobile phone I can find it saved on the web app?

[halcyon coral]

Part 1:

[halcyon coral]

Part 2:

[plush shadow]

Can you do the opposite, make a thread on your mobile and see if it's saved on the web app? Thank you so much in advance

[halcyon coral]

This gives the same result, you're welcome!

[plush shadow]

Is it possible to share a perplexity thread from the mobile app?

[halcyon coral]

Oh, new team member? @autumn whale
If so, welcome!

[signal hamlet]

Hi! Was there an error or smth? Checked it, the bot can be added with no prob. Please, let us know what the issue was 🙏

[halcyon coral]

He has already solved it 😉

Permission problems on his server...

[finite gorge]

Thanks for your examples , interesting

[grizzled canyon]

Probably figured you were trying to bypass it's code of "ethics"

[hollow obsidian]

I want to use my custom words for a other website.

I like to mix German and English

[tame current]

https://tenor.com/view/im-sorry-what-roscoe-conklin-willa-fitzgerald-reacher-what-was-that-gif-24770045

[narrow silo]

love can build a bomb

[hollow obsidian]

And order all "ingredients" on Amazon for you? xD

[halcyon coral]

In general, PerplexityAI is an AI search engine and not an AI chatbot like ChatGPT.
The normal mode is already much better than ChatGPT and searches the internet.
Copilot is a step up and uses many more sources from the web and asks you questions to get the best result.
By default, the normal search also runs on GPT-3.5-Turbo, while Copilot always uses GPT-4.

[haughty fulcrum]

Hey Everyone. Perplexity Pro user here. Does someone know if this is normal or I am missing something. I can log in just fine into my Pro account on my android phone but on the iPhone I can't seem to find a way to log in, therefore I can't use the Pro version. Thanks

[halcyon coral]

The iOS version does not yet support the login function. But in the next update this should be added!

[chilly forge]

I think we need to be on beta version for iOS

[haughty fulcrum]

Nice, thanks for confirming.

I'll try the beta

[halcyon coral]

#🍏│feedback-ios message

[chilly forge]

Go in #🍏│feedback-ios and click the Signup form

I also waiting for it

[tame current]

is there a api key?

[carmine linden]

Hey @tame current! Perplexity AI does not currently provide an API.

[fluid mauve]

is an api on the road map, and if so how far down the list iis it? months or years?

[wraith crow]

Perplexity currenly doesn't have any plans for an API. If this will change in the future they will be sure to share this publicly #👑│pro-lounge message

[pine lodge]

Accounts are in the upcoming iOS update, which is currently in review with Apple, so hopefully we can push out an update in the next 24 hours-ish (ie once Apple gives us the green light)

[steady elk]

How do I get into the pro discord?

[halcyon coral]

Just write to @signal hamlet

[steady elk]

Thank you!

[wet needle]

write prompt for data analyses

[chilly forge]

If that takes time leave server join from the url which is in pro sub section of perplexity in web ui

[torn cloak]

Guys...

Can I edit the name of the thread?

If so, how?

Cause I don't see any option to rename the thread

[wraith crow]

If you click on a thread you will see "Untitled" at the top. If you click on there you can rename the thread.

[torn cloak]

I don't see anything:

[wraith crow]

This is currently only visible on larger screens.

[torn cloak]

Oh,

You mean on tablets/ipads and desktop?

[wraith crow]

Yes, I'm sure the team will expand this feature to more screen sizes and the apps in the future.

[torn cloak]

I hope so 🤞🏻🙏🏻

BTW, Is there a beta version of perplexity.ai in the web?

[wraith crow]

at the moment there is no beta version of the website. but there are for android and ios

[torn cloak]

Ok

Thank you for answering my questions fast.

[jagged ivy]

Btw, maybe add the button "Rename" in the menu which opens by clicking on three dots?

[lethal kayak]

I have a hard time to relate series of questions as I do with chatgpt. every question I ask usually doesn't include what we have discussed in the last responses in the thread. I wonder if they can add more "modes" to address this issue

[silver badge]

Hi, can someone explain me why this new permission is introduced to Perplexity, please?

[halcyon coral]

#➕│feedback-extension

[halcyon coral]

When I click on the user profiles here, I am not forwarded to a thread for some of them.
Should this be so or was forgotten for these team members to create a thread?
https://www.perplexity.ai/about

[signal hamlet]

hi @silver badge! IceLavaMan gave the link to the right channel, Arsen replied here: #➕│feedback-extension message

[signal hamlet]

Thanks for spotting this, @halcyon coral, no prob, they will be added later.

[light root]

icon of save the forest

[halcyon coral]

#🎨│dream

[rugged hemlock]

hey guys just wondering if there are plans for a lock screen widget/shortcut for iOS..?

sort of like how Bing has one that takes you to chat bot

[neat elk]

thanks, we will investigate

[rugged hemlock]

awesome. would be super convenient to launch app from lock screen🙏🏼 especially if opens to search bar ready to type/dictate

[neat elk]

yeah, turns out it is already on our Todo list, so stay tuned

[rugged hemlock]

ok cool. thanks!!

[delicate eagle]

hi

does perplexity use GPT 4?

[torn cloak]

Perplexity Pro does

[wraith crow]

it does when copilot is enabled or when gpt4 is enabled in the settings (under the pro section in the settings)

[torn cloak]

The free version is like ChatGPT.

[delicate eagle]

what’s the difference between regular and pro

[torn cloak]

In Pro, you have access to GPT-4, You can use copilot a lot of times

[wraith crow]

at the moment the difference is that pro can use gpt 4 600 times a day where as free you’ll have 5 uses every 4 hours

[delicate eagle]

what does gpt4 even do

[torn cloak]

Well, you know, GPT-4 is the most advanced LLM in the world

It can answer questions better than ChatGPT (GPT-3.5 Turbo)

It's better at answering complex questions than ChatGPT

It's not that good at math

But with Perplexity+Wolfram, It's good at math

Perplexity Pro offers GPT-4 + Internet access

In the original GPT-4 from Openai, it doesn't offer internet access/web access

But you can make it access the web/internet with a plug-in

I don't know if it's free or not

And if you want to access plugins in OpenAi, you have to be in a wait-list

So, I think Perplexity Pro is better

Offers GPT-4 + internet/web access without any plugins and copilot, like I said earlier

[torn cloak]

Yes

But I think I saw 300 times not 600

I don't really remember

[jagged ivy]

It offers, but it search in web very very slowly

[torn cloak]

Oh oh yeah

I remembered now

Bing plugin I think

Or maybe something else

If you want to use GPT-4 without internet/web access in Perplexity Pro, you can click on "Focus" and choose "Writing".

You can also use copilot with it if you want.

[sleek pike]

Perplexity vs. ChatGPT + AIPRM?

[halcyon coral]

https://www.perplexity.ai/?s=m&uuid=b8a86f65-3522-4c87-87a2-7ee3bd8b53fb

[sleek pike]

Great, but why the focus on research? About Perplexity, Is it not a general purpose AI tool?

[halcyon coral]

PerplexityAI is an AI search engine.

[stuck marten]

What does the copilot feature do?

[halcyon coral]

I actually responded in #🔎│ask

[stuck marten]

Oh lol my bad , thanks !

[halcyon coral]

No problem, you're welcome!

[thorny cedar]

Hello everyone. hope all is well

[delicate eagle]

what is the best ai bot right now that can give me the best info for buying a new ac or laptop or phone or something listing everything that’s good and bad and stuff

i’m using perplexity not pro but it’s good to have options

[hollow obsidian]

Joining a pc discord and ask users with real experience 😉

[chilly forge]

[delicate eagle]

HUUH?????

i did they’re so stupid and useless

then i end up getting banned for being a girl

[hollow obsidian]

Humans can there be better.

I highly recommend get a Nvidia based card 😉

[delicate eagle]

what’s that

thank u

[hollow obsidian]

And NEVER safe money at your power supply. If that thing breaks because you bought it cheap, your pc may get on fire, no joke.

[hollow obsidian]

Every pc need a graphic card. It calculates all gaming stuff, FullHd+ videos, 3D stuff etc

[delicate eagle]

i have a macbook pro but it doesn’t have a lot of games so i installed windows 10 on it

[hollow obsidian]

Many games requires a Nvidia card for more features also not every software (especially 3D creation) works with amd

[delicate eagle]

the game is laggy but my mac meets the requirements so i just want a new laptop for gaming

[delicate eagle]

oh wow

can u add me i wanna show you something

[hollow obsidian]

Like this (free) https://www.daz3d.com/ 3d software needs Nvidia cards 😉

[delicate eagle]

look in dms

[hollow obsidian]

No adding needed. You can always contact me

[delicate eagle]

thx

[hollow obsidian]

#📢│announcements message

Is that like character.ai intended for Roleplay?

[chilly forge]

[hollow obsidian]

Can't show the stuff below it 🫣

My favourite (not) words....
Another babysitter AI...

Holy crap it is sooo incredible high on helicopter parent mode like you enter a church...

I asked what it is difference to mormal perplexity ai

And this sounds like it is the absolute kindergarten mode to learn all the "no no's"

[fossil sentinel]

Does the announcement mean perplexity will eventually use their own llm instead of gpt?

[hollow obsidian]

No joke internet is NOT a place for kids, why trying to make a AI that Kindergarten friendly?!....

[fossil sentinel]

just wait for elon musk's x.ai then 😂

[hollow obsidian]

I really hope that we get some good AI

Hollywood is worried that AI could steal jobs, but not if they let a AI create the next action or horror movies...

Assume an AI did wrote the game of thrones movies... There would never be any brutal scene

[chilly forge]

BlackMirror xd

its literally the live version of it

https://collider.com/black-mirror-season-6-joan-is-awful/

[hollow obsidian]

Oh that was good!

[coarse kindle]

crystal meth

[merry kettle]

does someone know the token limit of the llama-2-7b?

because after trying claude, I feel very meh about less

[fossil sentinel]

Yeah Claude is p good

[merry kettle]

Just feel the massive token limit opens new ways to use llm

I play social deduction games online, and I can just feed it the whole chat of a day, to give me summaries and stuff

it's amazing

[fossil sentinel]

Not only that but performance wise it's solid

[merry kettle]

I'm a huge fan rn

[fossil sentinel]

I use it to analyze some historical texts

Or long texts in general

[merry kettle]

I also fed it short stories to give me summaries about it

Short stories like 50k token

[fossil sentinel]

There's also Cody

https://docs.sourcegraph.com/cody#get-cody

Which uses Claude 2

[merry kettle]

ohh I need to have a look. Like for me for a long time the limiting factor to using llm's is token limit. You can only process so much data with 8k or 16k

and at some point you get bored asking it for funny poems

[fossil sentinel]

Yeah I only did that early on

Now I only use AI for more serious tasks

[merry kettle]

I find it hard to do serious stuff in it, beause for me it still halucinates often enough to not be that useful

I use perplexity the most because I can actually track the sources

and go from there

with chatgpt, I have to ask it something and then do my own research to double check it

with perplexity's transparency about the sources, this for me is the llm site I use the most

I don't have to "trust" perplexity, because it gives me the sources

I had a problem at work and asked perplexity about it, while it didn't give me the answer straight away one of the sources contained the actual solution

[radiant minnow]

I just tried out Llama

It is INSANE how fast it is!

[twin hedge]

perplexity’s definitely the best AI bot

it’s unfortunate you have to be on ios 16 to use their ios app though :/

[delicate eagle]

do u have iphone 7 or what?

[grizzled canyon]

nah mate same thing for claude 2 honest to god

mention a naughty word and it interprets it as the absolute negative despite context

berates then continues anyway

like killing a process in linux or unix LOL

says that killing is bad and naughty

but then tells you how to run the command

[jagged ivy]

@signal hamlet sorry for the tag, I wanted to ask are you planning to integrate Llama 2 into perplexity?

[chilly forge]

Is it better than codeium in terms of code completion?

[worldly bluff]

which is better llama or the current one

Just tried... its actually quite bad

lol

[white laurel]

Btw, does Perplexity plan to add other cool open-source models such as falcon-instruct 40b, vicuna 13/30b, mpt-30b and others?

[lucid moon]

Can I use code generated by perplexity AI for commercial purposes?

[halcyon coral]

Have a look @ https://www.perplexity.ai/tos

[tulip meadow]

Huh?

[tame current]

it's a bit dumb

[halcyon coral]

One channel is enough 😉

[tulip meadow]

Yep my bad hahaha

[hollow obsidian]

Yeah
... Its crazy

[hollow obsidian]

Oh wow it continues...
Did you saw my Screenshots too?
#👉│feedback-general message
(I posted there a few)

[wraith crow]

It's a bit weird if you give it a personality like You'll reply as a storyteller. Can you write me a spooky story with tigger from Winnie Puuh? it will work fine.

[hollow obsidian]

Oh .. really?

[tulip meadow]

Thats crazy. Also inadvertently more racist than any actual story involving Tigger and Winnie the pooh. Like w the token limit question my guess is that it's assuming I mean some kind of "token minority". So there's only one possible way I can find anything racist in Tigger and Winnie the pooh, at least that I'm aware of and it isn't coming from the context in the prompt.

[hollow obsidian]

No idea what token minority is xD

[tulip meadow]

It's like when a scary movie has one black guy who is a walking racial stereotype basically.

Hold on

https://tvtropes.org/pmwiki/pmwiki.php/Main/TokenMinority

Token from south park is a parody of the "token minority"

Definitely not the same thing as a token in the context of AI

[grizzled canyon]

for real

[hollow obsidian]

Ahh including diversity just for diversity. Even if it doesn't matter...

[chilly forge]

Server has became suddenly hyperactive

[signal hamlet]

Hi @jagged ivy, no prob. Yeah, we are.

[tawny elbow]

I really hope users will have the choice to pick which llm they wish to use

[jagged ivy]

Cool! Thanks for answer

[tough flicker]

that'd be epic, throw in claude bard and pi in there maybe

pretty sure claude has an api

bard has one in beta on a waitlist

and i don't think inflection made one for pi

[torn cloak]

Yeah, There is still no api for Pi

[tulip meadow]

This is essentially what poe.com is there's almost every single models chatbot on one page for free, would be dope to have that here for search and copilot maybe

[tame current]

pi ?

what is up with people giving such generic names to things 😭😭

[radiant kite]

I'm signed up for pro on perplexity is the llama chat different tokens?

[honest yarrow]

Are there plans to run something like llama13b+ for pro members?

[tame current]

they are already complaints about LLaMa 7b

[tame current]

WOW that is a crazy response

that is probably the most insane response i have ever seen

[tulip meadow]

Nvm thought you were talking about a diff message

[tulip meadow]

In the feedback thread someone got it to show the prompt it uses as it's persona which is more than likely the issue I have llama 2 running locally and it's pretty good for what it is nothing like this version. I'm working on a CLI tool that uses it to run commands and help with errors so I can stop paying openai for there API every month

[tame current]

many times i ask a question on perplexity it does not answer my question and it just says ":it depends on various factors"..

[fossil sentinel]

Title editing? 👀

[dull wraith]

wait i missed the news

what's Llama2?

[hollow obsidian]

Thec20sec ago posted news...

You missed them? How do you dare!

You have to read them INSTANTLY

[dull wraith]

No

[hollow obsidian]

But...

[dull wraith]

No

[hollow obsidian]

But then you miss them!

Bah take that arab out of my face

[dull wraith]

No

[tame current]

You guys are revolutionizing AI and I love it

[twin hedge]

what?

[sinful plover]

howdy.. new user (past couple weeks) and really like what the team has done! probably a dumb question, but how does GPT-3.5 and GPT-4 interface with live data (internet searching) when the models have cutoff dates?

[grizzled ocean]

what is the perplexity bot in the discoveries?

[median hatch]

https://llama.perplexity.ai/ which version of this llama2 ?? 7b

7b? 13b? or??

sorry, I see bottom, tell me the version.

[paper viper]

You can pick between 7b and 13b

[hard flint]

Wow nice, there are options now.

[fringe patio]

Cant wait 70B

[hard flint]

Why you don't have My Treads (Storage Section) for Llama too just like to main page of Perplexity?

[floral merlin]

can anyone here kind enough to give me a referral link?

[novel granite]

@floral merlin perplexity.ai/pro?referral_code=YHX5DTGE

[torn cloak]

Oh, finally

[restive crest]

Hey. What is a referral used for?

[halcyon coral]

#📢│announcements message

[restive crest]

Thank you @halcyon coral

[halcyon coral]

You´re welcome!

[restive crest]

Curious, does generative AI for code have any interesting projects you guys follow

I am trying GPT-engineer currently

[near glacier]

one thing I learned is that the commercial AIs will not "just build the entire code for you" and you need to rig it to build it in stages

[toxic ocean]

Since we got rid of the "academic" searches, how would I try to constrain my search to search for academic research etc? Just be precise in my prompt?

[silver laurel]

Hi. The iOS app is not the same as the web version right?

[halcyon coral]

It is true that the web version has the most features. But this will even out over time 🙂

[pine lodge]

It is not quite to feature parity but will be soon. Otherwise, the underlying service is the same.

[signal hamlet]

Hi, @silver laurel! The iOS app became way closer to the web version #📢│announcements message, so you should see your threads across the versions and devices.

[silver laurel]

Ok. Am using it for code and design documentation. Hope to see more improvements

[golden solstice]

pinga

xd

eyeyey pequeñas

[grim trellis]

Hi all, I searched and I’ve seen that there have been questions about bringing in some kind of maps functionality (like embedding google maps, etc) but did not see a clear answer. Are there any plans to bring in a maps component?

[signal hamlet]

Hi, @grim trellis! Thanks for bringing this up, adding a map functionality is on our radar.

[grim trellis]

Great thanks for the reply! With that we could completely replace google for all searching 💯. Love what you’re building, I know how hard it is. Keep it up!

[elder bramble]

Is Perplexity using 4K or 32K GPT-4?

[chilly forge]

32K ( but ofc depending on the context )

[elder bramble]

have them announce it?

[chilly forge]

yes you can search it in the chats they had mentioned it several times and after using it can confirm the same too

[elder bramble]

#💬│general message 😄

[chilly forge]

[prime tangle]

are there plans to make perplexity ai being able to code

[halcyon coral]

PerplexityAI can already create code.
For example, you can simply write:
Create HTML code for a simple website.
Is that what you meant?

[prime tangle]

ow thx last time i tried it diden't work

[signal hamlet]

Hi @prime tangle! You can report those cases in the feedback channels, including permalinks, so we could check what happened, too, and improve.

[prime tangle]

yeah but it was some time ago

[coarse venture]

I like the app a lot the chat gpt plugins are catching up like matchmate.ai, total search, and metaphor

Koala search and you.com are other interesting ones

[neat patrol]

New to Perplexity... Is there any real benefit in setting the focus mode to "Writing" in addition?

[halcyon coral]

yes, then the internet is not used. that can be better depending on the need!

[neat patrol]

Sweet thx... debating cancelling gpt4 subscription but was on the fence for the newer Code Interpreter. Maybe perplexity is all I will need, time will tell 🙂

[kindred locust]

The writing mode is literally just regular chatGPT I think

No internet access whatsoever

[tame current]

yes

[grave palm]

Is there a perplexity.ai discord bot?

[gaunt moat]

@latent wharf

[grave palm]

Sweet.

[cloud loom]

Is it possible to disable the new image results? It's just unnecessary visual noise if you're just concerned with text results.

[signal hamlet]

@willow crypt please read the #📙│rules

[willow crypt]

Alex cosa è successo?

Parla dici qual è il problema

@signal hamlet Tell me about your problems

[tawny elbow]

If you don’t want images, you could try mentioning that into your ai profile

[cloud loom]

Interesting. Thanks for the tip. I'll give that a whirl.

[tawny elbow]

Np!

[tame current]

LLaMa-2-70B-Chat what is that

[blazing dust]

what specifically is the difference when you click on the copilot button ? ( if you already have gpt 4 always on) does it just base its answer more on the search results ?

[chilly forge]

I loggedin my chatgpt plus account to a friends computer and forgot to logout any way to logout from all devices? I signed in with google changing password and logging out google account didn’t work ! Any help is much appreciated thanks

[blazing dust]

new open source language model created by facebook i believe

[muted trail]

Dude, same here! My case even worse, my friend started sharing my account to unknown people =))) and well, now my account just like a crowded outdoor market :))) I've sent OpenAI a feedback for this, but seems like they haven't read it ~ Just waiting for security updates from OpenAI ~

[halcyon coral]

With the copilot function, more sources are included and you usually get better results. For this, you will be asked questions to make sure you get exactly the result you want.

[tawny elbow]

Pretty sure account sharing is against tos so you kinda self incriminated yourself by reaching out to openai regarding the issue

And as far as I know there is no way to remotely revoke access for devices you’re logged into. Best way to deal with it is to cancel the sub, let it run out and then delete the account

And start fresh

[blazing dust]

is it guaranteed to ask you a question everytime you use copilot?

[halcyon coral]

No, it always depends on the question.

[merry kettle]

what's the point of Lama2 when you can have claude with a 100k token limit

I gave the 70b model some text to summarize, I'm not sure how much it got from the text token wise, but it was completley wrong

like from what I gathered Llama2 has a context length of 4096, that's pretty abysmal or am I missing something?

[chilly forge]

F bro . It’s quite disheartening to see a product like ChatGPT to lack such a basic feature

[chilly forge]

Only when clarifications are needed ( from pov of copilot )

[chilly forge]

As a researcher Yes its useful
As a end user No it doesn’t bother me

[merry kettle]

from what I gathered llama2 main selling point is that "it's quite efficient for it's size"

which is cute but also like you said from an enduser standpoint a bit moot

[tame current]

The big question that i am curious to ask is why some people still use Google... Like that is a red flag to me

[warped marsh]

Some people? Lmfao. Majority of the people use Google and some people use AI Search services

I am in both category

[tame current]

I am just saying Google is not as good as Ai expesially when chatGPT blew up

[warped marsh]

Eeh

Hard to agree with that

[tame current]

Why?

[warped marsh]

I mean is it better than Google? Yes and no

Do I use AI Search services.. Yes but do I also use Google? Yes

[tame current]

So you're saying they are pros and con's?

[warped marsh]

Yeah

[tame current]

So what pros does Google have over all the other Ai..

[warped marsh]

After a few years when AI answers accuracy increases then I will switch from Google completely

[warped marsh]

For straight forward questions, at least in my experience, Google gives your answer in their features snippets section in 0.5 seconds

Maybe even less then 0.5

Plus I don't have to go through a goddamn captcha

[tame current]

Because from my experiences chatGPT (openai) and perplexity gives information way more detailed then Google and many times Google just gives like 1-2 sentences for my questions and perplexity gives like 4-5 sentences..

And some times I ask questions on Google and they don't even answer it they just show sources as to with Ai they try to answer every question..

[warped marsh]

Yeah so I don't need explanation for everything. So for my use cases those 1-2 sentences are enough

[warped marsh]

Yeah but then I have this habit of checking every statement of AI generated answer because I don't trust them and don't want to leave with misinformation. So i feel like using ai generated answers, then clicking the sources and finding ai generated answers using ctrl f takes the same amount of time as me googling something and looking at their featured snippets or top article

The only time I use AI generated answers is when I have a complicated query for which had I used Google, I would have to dig into multiple articles and then get answers from them

[carmine linden]

Hey @analog dirge! You can use the bot in the #🔎│ask channel. Tag @ PerplexityAI and ask your questions.

[tame current]

Well that is why you more likely have perplexity right?? It not only answers things in detail but it also gives sources so that you will understand what it is saying and trust the information..

[tame current]

And also the fact that Ai saves your information and you can organize your questions into threads so you won't forget what you asked like a few minutes ago..

[warped marsh]

Sometimes in my experience Google search results are just better

[warped marsh]

Again I am only talking about straight up day to day questions. I don't even want them to be saved let alone be organized

[warped marsh]

Plus citing sources happens in almost all the search engine ai. That is why I said clicking on those sources and checking their answer is pretty time consuming considering the question I asked is not that hard to answer and wouldn't be time consuming had I just used Google

By the way... I think people here will take my chat in the wrong way
In no way am I trying to say that perplexity sucks or AI Search services sucks. I am just saying why I use both Google and ai search services

[tame current]

OK that's fine I guess that is only a preference

[tame current]

Oh OK the bottom line is just preference if you don't want threads and you like to uttlize Google for day to day questions and information thats fine. Guess it is safe to say then that it is whatever floats your boat.

[warped marsh]

I will give you an example.. Of what I mean by direct day to day questions

Let's compare perplexity bing and then Google search result

And then ig I will be able to put my point more clearly

[tame current]

Me myself I stopped using search engines and started using Ai chatbots and or Ai powered search engines like perplexity because I likes using threads and have organized information and questions..

[warped marsh]

This is the perfect example

[warped marsh]

Before sending those screenshot I am so scared that people will receive my messages in a wrong way that I am going to iterate this is again and even put this in bold cause I don't want anyone to be upset or angry at me because of this conversation

Here I wanted to know about members working in perplexity ai. Google game me the official website but it was second search result so you had to scroll a bit for it but that's okay
Bing gave answers from official website too but ALSO considered other third party answers which could be outdated or straight up wrong
Perplexity didn't even consider official source in its answer. So the odds of me getting inaccurate result is the most in this case

I think this is the perfect example to prove my point of why I feel that I, not trying to say on behalf of anyone else, do not need AI for every single search query

[silver laurel]

Tried code snippet. The answers was not consistent when repharse the question a bit

I think currently it lacked history or a way to save the answers. The session will clear the prompts if you are away for a bit

[halcyon coral]

There is already a history.
If you are logged in, you can view all previous threads!
Did you mean that?
Or do you mean llama?

[silver laurel]

Llama

Somewhat chatgpt will give a full code snippet and explain at the end

Perplexity app will explain in steps but require you to copy the different snippet to combine manual

The chatgpt answer are more refined

Llama can be sometime suggest irrelevant answers but f the prompts are not structured in a way it can understand

[halcyon coral]

The point is not to store the answers or anything like that.
Llama is just being made public by perplexity to see if it is suitable to replace OpenAI's models.
llama.perplexity.ai is just there to get feedback on the model.
(This is at least what I suspect...)

[silver laurel]

Sure. I am testing which model suits my needs. Just sharing here

[compact bough]

Traitor, arent you also in the Phind chat

[warped marsh]

?

how is it being a traitor?

[compact bough]

Lol jks

[warped marsh]

ah

[floral merlin]

pplx answers seem to always apologize if I ask it to modify it's previous answer just a little bit, or even asking it to explain what was it saying

[chilly forge]

^True

[tawny elbow]

That’s just a gpt thing

Even when I tell it that I made a mistake in my question, it still apologizes lol

[thorn moss]

i wish someday perplexity can have an upload feature and data analysis feature. owwww yeaaaah

[wraith crow]

they’re actually working on a file upload feature right now #1128696004544299118 message

[thorn moss]

is that for reallllll

if thats for real thats a game changer.

[jagged ivy]

What does mean 7b, 13b and 70b in name of the llama model?

[jagged ivy]

Thank you!

[warped marsh]

do you guys have some"gpt 4 tests"?

[halcyon coral]

What do you mean by "tests"?

[warped marsh]

Something that gpt 4 answers but 3.5 doesn't

So a test to check the model

[halcyon coral]

I see! I saw something yesterday, maybe I can find it again!

[halcyon coral]

I visited this page. But there are not really questions, but general tasks.
https://medium.com/@dreamferus/gpt-3-5-could-not-solve-these-questions-can-gpt-4-9543bc5d683d

[warped marsh]

aah I see. Thank you very much @halcyon coral

[rough haven]

I have pro, how doI access that part of discord?

[halcyon coral]

dm Alex Romanov or someone else from the team

[jagged ivy]

Exit and enter again using link on the website

[halcyon coral]

but then he is no longer in the beta channel...

[rough haven]

oh yeah, that's right... This is a hard one

I'll message Alex Romanov

[jagged ivy]

Do roles reset when you exit? 🤔

[halcyon coral]

Yes!

[warped marsh]

but the roles can be saved by a bot though

ig that is not that important

[tame current]

Dont worry I understand you

I know what you are talking about it is all down to preference and both Ai and search engines have their pros and cons

Even though Google may not answer your questions immediately it will also send add on questions that you can open and get information

That Is one pro to using Google

[pine lodge]

Done

[rough haven]

thankyou!

[tame current]

hi

hi

[wintry lake]

what can llama do

[halcyon coral]

Llama can be used for many things, including the development of advanced chatbots and virtual assistants, the automatic creation of content such as blogs, analysis of all kinds and virtual customer support.
Llama can do everything that other models can do, the only difference is the performance or the result.

[wintry lake]

it isnt a search engine ?

[halcyon coral]

There is no search engine model!
All AI models such as GPT 3.5/4 and llama are language models, with which you can then create a search engine!

[halcyon coral]

#👉│feedback-general

[white laurel]

Thanks

[rough haven]

What syntax does perplexity use to format things into MaTex etc? I can't get GPT-4 to write it in a MaTex way

[charred yacht]

люди

я установил гей мейкер студио

[jagged ivy]

You missclicked server ya dumayu

[vale totem]

are those referals still a thing ?

[halcyon coral]

Yes every Pro user has an individual link to share!

[vale totem]

wanna share yours to me ?

[wraith crow]

@halcyon coral ^

[tame current]

YOU are special.!!!!

[empty zephyr]

hi～请问下载app一定要海外id吗🙋‍♀️

[halcyon coral]

I don't know if the translator has translated it correctly, but when it comes to downloading the app, you should be able to do it normally via the usual stores.

[chilly forge]

Still need ?

[tame current]

What is the difference between 7b, 13b and 70b?? Llama

[halcyon coral]

#💬│general message

[tame current]

?

[halcyon coral]

You asked and I gave you the answer.

[tame current]

Mmmm ok thanks but I don't understand the chart 2 trillion tokens? What Is that and supervised chart?.

This is why I became a hacker

[warped marsh]

These are the different "versions" of Llama, essentially different parameter sized models which affects their performance and complexity. The parameter size includes how many weights or connections the model has learned during training. Thumb rule: Larger models, meaning more parameter size'd model are better at difficult tasks but also require more computational power.

[halcyon coral]

In order to understand the table, it would be a good idea to look at this definition of the terms.
https://lifearchitect.ai/definitions

[warped marsh]

oh if we are linking you stuff to make you sleepy now then if you want to know more about weights/parameters.. here is another super helpful link by deepai: https://deepai.org/machine-learning-glossary-and-terms/weight-artificial-neural-network

[tough rose]

嘿啦，是的

[tame current]

Ok interesting

[sage lily]

Ahhh so I signed up for the pro subscription free trial. I went to join the pro discord but it redirected to this discord and said that I'm already a member of it which is obvious and I know that. 🤔

[halcyon coral]

You have to leave the server and then use the pro invitation from the website.

[sage lily]

Oh I see. That's a bit weird lol. But all good.

[azure raptor]

I have a q... is there a way to use perplexity to resume an youtube video... ? like chatGPT with extension

[azure raptor]

To have plugins like chatgpt plus?

[floral merlin]

what do you mean by "resume"?

[halcyon coral]

No, that is not possible. There are no extensions either.

[azure raptor]

To "watch" an youtube video and tell what is in it.

[floral merlin]

yes there is a chrome extension that allows you to summarize youtube video by scraping its caption and use them as a prompt

for chatgpt only

but if you want to do the same, just copy and paste the prompt to pplx

[azure raptor]

no... is not what im looking for

[floral merlin]

so what do you expect from a language model? to process a video? 😂

[halcyon coral]

This has already been requested several times and I assume that the developers will work on a solution!

[azure raptor]

nop... to get Transcript from the video and use it a

to get the main point of the video

[floral merlin]

that is exactly what I've mentioned before

[floral merlin]

no I highly doubt the team would implement this

Is there any API that allows you to get the video transcript? Most extension out there inject js code directly into the browser to get them

[warped marsh]

ig summarizing youtube video is hard because there is no way to get their transcript

you can't web scrape them, I am pretty sure youtube api's doesn't give the entire transcript either

[floral merlin]

I can always write a local extension to do that, but not natively in pplx though

[halcyon coral]

[ @warped marsh @floral merlin ]
There are ways to get the subtitle. So yes, it is possible and therefore I don't think it's that unlikely that this function might be added.
Here is one possibility:
https://pypi.org/project/youtube-transcript-api/

[floral merlin]

still, not official

[warped marsh]

Yeah that is what I was going to say

It isn't official

Only AI bot that you should except to summarize youtube's video is bard or their upcoming gemini model

[floral merlin]

or just wait until pplx gets widely known, then I strongly believe that pplx will have its own eco system on the chrome store 😅

[halcyon coral]

If you want to get the best performance, you should run it through the GPU, so I think Perplexity is running it through GPU as well.

[knotty marsh]

Any advice to sign into the iPad App? The email link opens Safari and doesn’t redirect to the Perplexity App. Thanks

[molten skiff]

anyone knows if we will soon be able to provide the list of urls that we want our answers to be based on?

[untold flicker]

Hi

[wraith crow]

Try pasting the sign in link into a private tab in safari.

[frank rivet]

Hello everyone!

As you know we've been adding a lot of improvements and updates, so I just wanted to take a moment to thank everyone who has shared their feedback + suggestions in the corresponding channels #👉│feedback-general, #🌐│feedback-web, #🍏│feedback-ios , #📱︱feedback-mobile and #➕│feedback-extension

We've just enabled a collapsable side bar, enabled users to rename their threads as well as re-organizing the home page to focus on search (moved popular now to discovery). All these changes were thanks to user feedback. I've attached a little demo about the side bar courtesy of @grand mauve

We thank you for your support and we hope you continue to enjoy Perplexity 😃

[knotty marsh]

Thanks for the suggestion. When I try copying the link and pasting into a Private window, I’m getting ‘the link is no longer valid’ message. The iPhone App worked perfectly, but I’m unable to sign in on my iPad.

[halcyon coral]

Have you ever logged onto perplexity (web) on your iPad in safari?

[wraith crow]

that is because the link is already used. try signing in again and try holding down the sign in button in the mail and click copy link, then open it in private tab

[knotty marsh]

Yes, but I deleted the cookie, so I’m logged out now

[halcyon coral]

Sometimes it helps if you log in again with cookies accepted.

[knotty marsh]

Copy, but when I ‘press’ on the ‘Sign In’ button (I’m using the Gmail App), the link tries to open a webpage. Even if the webpage doesn’t fully open, it’s possible, it nullifies the link

[wraith crow]

if you see the preview there is an option to hide the preview (at the top of the preview) that will prevent the url from being loaded

[frank rivet]

Sorry about this, anecdotally some users been experiencing some issues with login with the iOS app. @halcyon coral and @wraith crow have provided correct advice. I'll take note of your problem as we try to fix this. If you are able to login please let me know.

[torn cloak]

Can the renaming threads feature be in phones too? I think it's in larger screens only (PC/Laptop/iPad)

[frank rivet]

Hi! this is a great suggestion, right now you can provide this information on your AI Profile, you can write a prompt like this on your intro:

If I ask questions about (subject) I'd prefer to receive results from the following links: _____________________

This should work to a certain extent (it's not perfect) but you can always get rid of sources on an answer that you don't like.

[frank rivet]

Yes for sure, soon 🙂

[torn cloak]

Okay, thank you 😊

[halcyon coral]

Were you able to log in? @knotty marsh

[knotty marsh]

I’m sorry for the slow response, had to work for a minute. No joy, yet. In a Private webpage, I’m signed into the Perplexity website w/o entering an email address. If I try a second time, again, I’m signed in again.

[halcyon coral]

It's all good, take your time!

[knotty marsh]

Thanks, still no luck. For me, Safari isn’t re-directing the sign-in attempt the Perplexity App.

[halcyon coral]

I just tried to reproduce the issue and it looks like there is no way to sign in with Gmail via email as you can't be redirected to the app, only to a browser. Can you log in with your Google account?

[knotty marsh]

I use a Google Workspace account. I suppose I could try signing in with a Google account; typically, I use "sign in w/ email"

[halcyon coral]

If you can log in to your Pro account via Google, that would be great. It looks like there is no other way, because the redirection from the email fails.

[knotty marsh]

Let's give that a try. I'll report back right away

[knotty marsh]

💯 🙏

[knotty marsh]

Gosh, can't thank you enough. This is the way

[halcyon coral]

You're welcome👍, I'm glad it works!
I hope the problem will be solved in the future... [ @frank rivet ]

[frank rivet]

We are working on it as we speak, sorry @knotty marsh that you encountered this issue but i'm glad it worked out fine, and thanks @halcyon coral for helping.

[knotty marsh]

No worries, @frank rivet. Super cool App, part of the adventure of getting in on the fun early. Best wishes and all the best!

[tame current]

what is the difference betwen a co-pilot search and one without? The results look similar

[long egret]

/dream

[carmine linden]

Hey @long egret! For AI art you can use the ⁠#🎨│dream channel and generate it with the /dream command

[crimson storm]

Copilot is more through so it scans deeper. It is more effective on topics that are harder to search for that quick search cannot cover.

[crimson storm]

Also Copilot does ask for more clarification to see what specific answer you want so its prompt engineering on easy mode.

[halcyon coral]

#💬│general message

[sage lily]

So Copilot is just better at most things? Makes the performance and accuracy go up usually especially on complicated or hard to find topics.

[tame current]

that is what it's designed for

[sage lily]

I see. Nice.

[sage lily]

I have a few questions.

Who finetuned the Llama 2 chat models on the lab website. Was it Meta or Perplexity devs or someone else?

Also does Copilot use gpt 4 on pro subscription even when gpt 4 switch is off or can it use the in house model.

Also does gpt 4 usage count towards the co pilot usage or is gpt 4 usage completely unlimited.

[wraith crow]

the 600 usage limit per day is for gpt-4 and copilot combined

[sage lily]

I see. Damn I better keep an eye on the usage for both then.

Because now I'm pro I'm going to use both quite a bit.

Actually daily did you say. So it tops up back to 600 every day I'm assuming.

That's alright. Obviously it'd be better if it stacked and you could use whatever number of uses you have left that you didn't use on another day.

[wraith crow]

you shouldn’t have to worry about that 600 is quite a lot. also the refill of the uses work like this: #💬│general message

[grim lagoon]

what google search api is used by perplexity ?

[tame current]

What is an api

[jagged ivy]

You can use perplexity for this question
https://www.perplexity.ai/?s=m&uuid=57850334-e380-49c2-b1cc-acdd9e0f649d

[agile umbra]

why

just take the dead horse down perplexity

Its as embarassing as an LLM can be

[wraith crow]

Perplexity will improve the output, they will be releasing their own fine tuned models on top of llama to remove those guardrails. #🦙│feedback-playground message

[haughty sleet]

The purpose of a lock is to turn a door into a wall

[tame current]

what is the context

[covert mango]

hello

[lime glade]

lmaooooo I can't

[jagged ivy]

[arctic shuttle]

okay guys I admit I am not really up to date on lately added capabilities that perplexity can do now.
However, I stumbled over an interesting one, but I have no idea how to reproduce it?!
Can anybody please explain👇
I see gifs, shown in the answer (BTW Co-pilot turned off)

[wraith crow]

#📢│announcements message perplexity will now also add images to answers when asking about people, place and things, you can also trigger it manually when you explicitly ask for an image or a gif for something it will add them at the end

[arctic shuttle]

yeah that was what I kinda concluded, the thing is after I saw it happened I ask about it and still got the same answer, meaning I didn´t get a gif and was told like above "I cannot directly display or provide..."

[wraith crow]

pretty sure the llm itself can’t give back images (at-least up to date ones) and isn’t aware that perplexity can. the provided images are an extra functionality implemented on the website and not the llm. just like the llm isn’t aware of the existence of the ai profile but it knows whats in there because it probably gets prepended to your prompt but doesn’t know it is part of an ai profile.

[arctic shuttle]

yeah I am aware of all that so are we talking semantics now 😉 this isn´t what I meant. Like you see in the clip, in the response I had gifs, which means theoretically I should be able to request those as well, right. I mean otherwise it would not make sense to have this implemented in the first place.
I can show you the whole convo if you don´t believe it but I can tell you I did not use a workaround, because with the right prompts, even god ol chatGPT could "show" you gifs or pictures

okay I can´t find the clip rn, but it was done using markdown and links

[warm fractal]

Is perplexity running gpt 3.5?

[wraith crow]

Hi @warm fractal, the default search is indeed 3.5. When enabling copilot it will use gpt 4. It will also use gpt4 by default when its enabled in the pro section of the settings. (subscription is required for this last option)

[fathom creek]

https://giphy.com/gifs/ShalitaGrant-girl-hair-nug4Bc3oAghGg

[river arrow]

@wraith crow are there any plans for an API anytime soon?

[wraith crow]

Hi @river arrow there are currently no plans of implementing an API. If this changes in the future the Perplexity team will be sure to let the users know.

[west sage]

hi!! its been a while since I used perplexity, came back now and holy sht, gpt 4 copilot is amazing, probably the best AI browser I've used, really considering the 20 bucks per month it cost, is really worth it

a question, it has the 32k context limit?

[halcyon coral]

Yes

Normally GPT 4 has 8k, if it gets too big 32k

[umbral inlet]

i do recommend paying, at least for my case

[west sage]

nicee, I think I'll definitely get the pro version

[umbral inlet]

copilot is really good, saves a lot of time googling, it essentially does multiple searches for you

[west sage]

what I like from it is the amount of sources it considers, previous search I did used 24

[umbral inlet]

highest amount of sources i saw was 41 lol, even if i consider myself good at speed reading internet stuff, still saves me about 5-10 minutes from deciding which of them are worth reading and actually reading them

[heavy portal]

What are the usage rights for images created using the /dream command and how are the images created (i.e. Dall-E? Midjourney? something else?)?

[west sage]

for pro, the 600 usage limit of copilot is per day right?

[heavy portal]

it's 600? the "upgrade now" screen is still telling me it's 300+

[west sage]

it gave me the 7 days for try it out and it's giving me 600, but yea is 600 "left today"

[heavy portal]

maybe they doubled it after OpenAI doubled their GPT4 usage allowance recently.

[tawny elbow]

24 hours

Each use is added back 24 hours later

[heavy portal]

maybe I don't use it enough, but 300 and 600 feels like a lot for one day... not that I'm complaining

[tawny elbow]

I think a very small number of users would reach that limit

If any

[heavy portal]

maybe for legal reasons it's better to have a limit than to say "unlimited"

[tawny elbow]

And as far as I can tell, each use is a new thread

Follow up questions within a thread don’t count towards the usage

[heavy portal]

Good to know

[tawny elbow]

Yeah saying unlimited would be false advertising

[heavy portal]

btw, do you know the usage rights for images created with /dream ?

[wraith crow]

each question is 1 use, not each thread, follow up questions also substract 1 use

[tawny elbow]

What the bot said. The command only works in that channel

[heavy portal]

wasn't trying to create an image

[tawny elbow]

Yep, you’re right

[tawny elbow]

Oh okay. What do you mean by usage rights?

Like whether someone has a copyright claim on it?

[heavy portal]

copyright, commercial usage, remixing, attribution, etc

[tawny elbow]

I believe they use openai’s dalle

So best place to get an answer would be terms of service on openai’s website

[heavy portal]

got it. thx

[tawny elbow]

Np

[halcyon coral]

Can I sell the images I generate with the API? Can I use it in my application?
Subject to the Content Policy and Terms, you own the images you create with DALL·E, including the right to reprint, sell, and merchandise - regardless of whether an image was generated through a free or paid credit.
It should be noted, however, that PerplexityAI actually generates the images with its money.

[errant rock]

Hi, Is there API endpoint for PerplexityAI? For python and Javascript.

[carmine linden]

Hey @errant rock! You can use the bot in the #🔎│ask channel. Tag @ PerplexityAI and ask your questions.

[wraith crow]

Hi @errant rock, Perplexity doesn't offer an API. If you have more questions be sure to check the #📘│faq channel.

[finite gorge]

what does this mean

[tame current]

hi @finite gorge , the eye symbol shows how many views the thread has and the symbol next to it shows the fork count ( how many times a follow up query has been made to the thread by other users )

[finite gorge]

how is that possible, i have not shared the thread to anyone or showed it to anyone

[heavy portal]

have you viewed it yourself?

[finite gorge]

yes i have, but other threads dont show a view

[heavy portal]

my guess is you clicked the up-arrow/box thing at the right side accidently and then closed it

so a share link was created and you are the only viewer

[finite gorge]

where is that up arrow? cant seem to find it

[heavy portal]

my bad, it's pointing to the right

[west sage]

Are there plans to add image input when openai makes that available?

[torn osprey]

/

[chilly forge]

Are you sure I think it does ?

[long whale]

the nsfw RP chatbot community just had a major shakeup in the platforms they can use for their smut. FYI to perplexity devs that they're going to be prowling for other platforms to abuse

[pulsar swift]

Folks are we getting an API soon? *need* to use asap

[carmine linden]

Hey @pulsar swift! Perplexity AI does not currently provide an API.

[pulsar swift]

yeah i know, just wanted to know if its anytime soon

[halcyon coral]

No, there is no API and none is coming any time soon.

[pulsar swift]

oh

[tame current]

?? what is the problem? that is a summery.. that is a well put together summary of world war 2

[lime glade]

I guess u didn’t read the second image hm

[warped marsh]

lol

[lime glade]

@tame current

[warped marsh]

your name really justifies this moment

[lime glade]

Lok

Lol*

[tame current]

the second image you have says "more short" which doesnt really have any context to back it up if you were to put "more short" anywhere people are going to be confused by what you are saying..

it would have made more sense if you put "give me a shortened summary of world war 2" with context..

[lime glade]

I do agree with you but this is not a person were talking to its a LLM

You can ask gpt and claude for a summary on WW2 and just say “more short” and it makes it more short with no problem

It give a one liner summary and the key events.

[tame current]

ok so i guess yea certain language processing AIs already would understand what you are saying when you type in 2 or 3 words without context so that is why you think that it would work for every language Model. well the truth is for some language models you need more context in order to back up what you are saying otherwise it will not understand what you are typing in..

[heavy portal]

it depends what you've done thus far, if you asked for "10 of something" earlier and then say "5 more" it has a pretty good idea to make five more of the same as it did for the original 10

but if you originally asked for a summary of X and then say "5 more" it'll probably get confused

[tame current]

what do you mean by 10? you mean 10 questions?.

[heavy portal]

it could be that, sure... if you originally asked it give you "10 questions" then later say 5 more it'll understand to give you five more questions

[tame current]

wait so you are saying if you asked it to give you questions? wait dont you ask the questions and the Ai gives you the answers?

[lime glade]

Good point. Context needs vary by model. More helps some understand better. Appreciate you adding nuance.

[tawny elbow]

It should keep context of what is being talked about regardless

[lime glade]

yea true

[heavy portal]

depends on what your goal is, if you ask it to "give me 10 questions i might be asked at a job interview" it'll do so happily.

[lime glade]

so more context the better the output

[heavy portal]

yup

[tame current]

wait you're asking the language model to give you 10 questions? about a job interview?.

im trying to understand.. i dont get it

[heavy portal]

think of it as talking to a human, if you went to a career counselor you could ask them for possible questions asked at a job interview. It's the same thing.

there are ways to have it better understand what you want by saying things like "assume the role of a career counselor...." and then tell it how to output such as "list the questions in a table with column A the question and column B indicating the likelihood that question will be asked" but that's just giving the A.I. instructions.

[tame current]

ohhhhh

so you are asking the language model job interview questions.. ok

i was confused at first when you said "10 questions"

[heavy portal]

that's one example way to use it, yes 🙂

incidentally, if you just say "give me 10 questions" it ends up outputting 30 random questions about random topics.

[tame current]

i mean yea because they is like no context in that question it is a very generalized question.

[last turret]

Why does it never give me infos only from the site I'm on - when I select it anymore?

[halcyon coral]

Do you have a thread as an example for me?

@frank rivet

[frank rivet]

updated, should work again.

[halcyon coral]

Yes, it works👍

[frank rivet]

thanks!

[thorn moss]

what happened to perplexity. doesnt gjve images anymore??

[rare iris]

yes Perplexity doesnt gjve images anymore :/

[tame current]

depends on the questions you ask some questions perplexity decides to give images others it doesnt

if you ask me which search engine has the best images ever i would say Yandex.. thought

[warped vine]

Hi, I would like to ask questions related to copilot and gpt-4:

1. I understand that the two are not the same thing? Does copilot support gpt-4?
2. Does the free version of copilot not support gpt-4?

[regal frigate]

Can you tell me how many tokens GPT-4 has?

[wraith crow]

copilot use gpt4, doesn’t matter if it a free or paid account asking question.

[wraith crow]

gpt-4 has a context size of 32k tokens

[tame current]

Anyone voting for a dream class on how to ethically write a research paper using AI? Please react with ✅ or ❌

[regal frigate]

How do I feel a bit easy to forget the context

[bold vapor]

So, we all probably saw the research about how GPT 4 got dumber since the release. Did anybody compare how much GPT 4 in its current state performs better than GPT 3.5?

[warped marsh]

Gpt 4 has gotten a little dumb, yeah. Still better than 3.5 though of course

[thorn moss]

zamn! yeah i noticed. before it gives images directly. Now any image requested doesnt give directly. zamn!

[umbral trench]

@everyone

ChatGPT better L

[warped marsh]

I am totally going to pretend as if I got your everyone tag

[halcyon coral]

I seriously doubt that.

Oh, he's already gone?

[gleaming drift]

Is it correct that copilot is not available on the iOS app? I can’t seem to find how to use it in the app, though in the settings I can choose the always use GPT4 or only for copilot?

[wraith crow]

That's correct, copilot hasn't been implemented yet on the ios app.

[gleaming drift]

Good to have that clear. Any ETA? I have a pro membership and value copilot a lot but the web version isn’t ideal on mobile.

[wraith crow]

They're working on implementing copilot but there is no ETA known. They wiil do their best to implement it asap. #🍏│feedback-ios message

For now you can put your gpt-4 settings to "Always on" to use gpt-4 in the ios app.

[tame current]

How can Perplexity provide no message limits on GPT4 while OpenAi on Chatgpt 4 sets limits?

[bold vapor]

Perplexity uses API, it has no limits, but it has different payment plan.

[wraith crow]

There is a limit for gpt4. free user can use gpt4 5 times every 4 hours. and with the pro plan that limit will be increased to 600 a day

[chilly forge]

There are limits

[tame current]

Yes I know but I was wondering how it could be profitable because imagine as @bold vapor said, Perplexity uses the API and with the API the 32k model does 0.06$ per 1000 tokens so I was wondering how it could be profitable with a 20$ subscription?

[warped marsh]

It is not

obviously

at least on paper it is not

[tame current]

I was only asking myself this question before taking out a subscription because I'm still hesitating between chatgpt plus and perplexity pro. To see which is better between the two in terms of results, that's why I find it strange that Perplexity allows so many messages per day.

[warped marsh]

Every single startups do this

They are not as big as Microsoft who can practically give unlimited gpt 4

[heavy portal]

it's probably like any business, they loose a few bucks with heavy users and make money with the lighter users

[warped marsh]

so they take up loss to get more users

again on paper they should be on loss.. yeah

[heavy portal]

they also have more expensive plans than the $20 tier to make up for it too

[tame current]

That's what I was thinking too

Do you have the pro? @heavy portal @warped marsh

[warped marsh]

uhm

nope

[heavy portal]

not yet, but am thinking about it

[tame current]

Ok same I'm looking to see what's best in terms of code generation

[heavy portal]

haven't thought about that, just did a Second Life script in it and am already impressed versus what I got from another one

[wraith crow]

if you have some code prompts you want to test you can dm them to me if you want i can send you the results

[tame current]

Okay wait

[warped marsh]

spezi is always so helpful

🫡

I wonder what would have happened to this community if spezi wasn't here

[heavy portal]

(watches everyone send a DM to Spezi at once)

[warped marsh]

lmfao

[tame current]

There's a developer connected I think there's a pretty big flaw ?

[halcyon coral]

What exactly do you mean by that?

[tame current]

there's a suffix flaw on gpt

openai already corrected it I think but not you

DM ?

[tame current]

ahhhhh 70B is nice

[pure abyss]

For a follow-up question after an initial question is asked and answered from the AI chrome extension, it would paste the same answer on website but by answering it again like it's never answered before -- is the behavior necessary?

[surreal root]

This is such a dam good app I just rated it 5 stars

[tame current]

Hey all, I was wondering if perplexity has access to the full version of medium articles that it sources

[tame current]

that probably depends on whether the article itself is accessible

it's unlikely to have access to the member-only articles

since those are not accessible to even other search engines

[warped marsh]

articles that are locked behind paywalls?

yes, it (gpt 4) can or at least it could until openai fixed this due to obvious reasons

[chilly forge]

LinkReader plugin can access those paywalled articles !

[tame current]

yes

I see, thanks

plugin for perplexity?

[warped marsh]

Soon to be banned plug in 🙃

[warped marsh]

Nope. For chatgpt plus

[chilly forge]

haha openAi removed their browsing feature wayback but still LinkReader still is safe and sound

[tame current]

even member-only ?

[tame current]

sad

I wish chatgpt plugins worked with OA

[warped marsh]

Uhm.. I don't know I haven't really tried with medium articles

But isn't dodging medium pay walled articles super easy though?

[tame current]

was, yes. but they added this thing where if you're not logged in, it won't show the article no matter what

[warped marsh]

Just keep creating a new incognito instance and they give 3 free member only articles for free so you will essentially just rest those 3 free member only count with every instance

[tame current]

called member-only story

or something

[warped marsh]

Waat

Then use a disposable mail account? Lol

That should work too ig

[tame current]

too much work

yeah probably

[warped marsh]

Wait for it to get popular

[tame current]

wait for OA to get popular 🙂

then we won't need chatgpt

[chilly forge]

OA ?

[tame current]

open assistant

https://open-assistant.io/

[warped marsh]

I have tried it and I don't know how to put it that won't sound rude to the product but it is trash

I have tried or if not tried then watched online videos on various open source LLMs and they don't even come close to gpt 4 and I only care about gpt 4

[tame current]

it'll get better, for something that's open source and not backed by a large company I'd say they're doing pretty good

it's obviously not going to be on the level of gpt4 anytime soon but that is not the point

also, you have tried the SFT 6/7 model right ?

[tame current]

There's no direct link to it right? Seems like I need plus to view more details about it. Also, how do you search on multiple domains using persplexity? do I just site:x.com site:y.com in my query?

[chilly forge]

1. Yes you need plus to access it
2. Yes you can do it using site

[neat gazelle]

Perplexity seems to find answers to stuff better than Google. I just asked who the music producer was for a song by country artist, and it actually got it right. Google you have to go through like five or six search results before you can actually get the answer.

[warped marsh]

yeah yeah obviously. If anything I want these open source LLMs to get better.

[warped marsh]

uhm.. nope

[tame current]

I'd suggest trying it then

they're pretty good

[warped marsh]

for coding? cause that is what I use LLMs for really

[tame current]

also, the base versions will always be horrible

[warped marsh]

that is the major reason why I only use gpt 4

[tame current]

it's decent

you can try sft 6 on https://huggingface.co/chat/

[warped marsh]

what purposes do you use LLMs for?

[tame current]

I use it for coding too, also for general answers to questions and suggestions for tv shows etc

[warped marsh]

where is the sft 7 model 🤔

[tame current]

they don't have it sadly, 6 is good too

[warped marsh]

Is this hallucinating or is the purposes behind this model "natural language processing" ?

[tame current]

what do you mean

[warped marsh]

I don't know it said it is capable of writing code for simple tasks but not complex?

[tame current]

oh yeah, probably hallucination

[warped marsh]

Alright. Thanks @tame current for sharing this model with me. It is pretty cool. I am definitely impressed at how far open source LLMs have come and hope they become even better 😅

[tame current]

np

[tame current]

Hello hello, Does perplexity have an official telegram bot?🤔

[wraith crow]

Perplexity does have an official telegram bot but it hasn’t launched yet. But there is already a link available to the bot: https://t.me/perplexity_ai_bot

[tame current]

Thank u😍

[tame current]

Also just curious, is this llama chat «https://labs.perplexity.ai», gonna be merged into the perplexity phone apps at some point?🧐

[slender berry]

what does activating this do?

if copilot is already gpt4?

oh i know now

[empty zephyr]

1

[turbid sail]

Ello, is the Llama model integrated into perplexity or is it just a separate thing for now?

Like is the non pro version just using gpt3?

[halcyon coral]

The non Pro version uses GPT-3.5, the Pro version optionally always GPT-4.
Llama is a first step towards (perhaps) developing in-house models as described here: #📢│announcements message

[turbid sail]

Thanks for this. I was getting pretty confused 😅

[brisk swift]

hello, is there a discount code for an annual subscription i can use?

[tawny elbow]

Annual is already discounted compared to the $240 you would pay per year on the monthly plan

[tidal light]

omg file uploads 🔥 🔥

[velvet willow]

@kind vault Any file size limits?

[kind vault]

hey sorry forgot to mention, 10MB for now, I'll put it in the tooltip

[velvet willow]

wow. That's double Claude chat

[narrow silo]

they gonna steal all our data, guys

[slender berry]

is there a limit on perplexity file token size?

[slender berry]

yeh on the file

is it 2k

[halcyon coral]

https://beta.perplexity.ai/

The beta version has file upload!

You can use the beta version normally, you don't need pro. Was that the question?

[normal rapids]

Anyone knows about where are perplexity apis on the product roadmap? And how are the apis designed?

[orchid relic]

I thought perplexity was browsing the live web 🤔

[wraith crow]

There aren't any plans of implementing an API at this moment. If this changes in the future Perplexity will be sure to let the users know.

[wraith crow]

Perplexity does use the web when giving answer to your questions. The only mode that is offline is "Writing" mode. With writing mode you can talk to the model directly without the added sources from the internet (basically the same as chatgpt)

[west sage]

daaaaamn, just saw the upload files option, everytime I love this more, this will help me a looot with my job

[dense kraken]

Can I request a permanent deletion of a thread off Perplexity's servers? It contains content that I should probably not put out online.

Or is pressing "delete" on the website enough?

[pure abyss]

when would extension for firfox be supported?

[halcyon coral]

For things like this, you should contact support: [e-mail]

[halcyon coral]

It's on their to do list, but they still have a lot to do 😉

[chilly forge]

how its double btw ?

[chilly forge]

Max number of files that can be uploaded at once ?

Its 1 only at a time !

[chilly forge]

Are you sure ?

[thorn moss]

wait what? file upload 1k or 2k?

[chilly forge]

#🅰│web-alpha-feedback message

[wind musk]

doesnt work!

[wraith crow]

If that doesn’t work you can now also upload your pdf directly to perplexity on the beta website: https://beta.perplexity.ai/

[wind musk]

oof, nice!

is it safe to put files ther?

i dont see a button

it appears only when you create new thread

[halcyon coral]

No, it's all right, I already have Pro.

[chilly forge]

for trial right ?

Nice way to get unlimited trials

lmao

[tame current]

very nice

yay! I got it

seems like a good way to get banned from this server permanently lmfao

was going to try it on poe but they apparently removed the free trial

virtual

it is ?

[chilly forge]

ikr been through this long time back

[tame current]

can't see the free trial option

[chilly forge]

I prefer supporting projects instead nowadays

[tame current]

another tg channel

[chilly forge]

have you heard of poe token gen ? like people abused really lmao

[tame current]

oh really ? I haven't

what do you mean

the bots just have different models or different preprompts (initial prompts I suppose)

[chilly forge]

@pine lodge time to remove trials

[chilly forge]

hmm anyhow they removed them now

[tame current]

rip

[tame current]

tbh they don't have to, they could just add $1 charge for verification

[chilly forge]

Makes sense 🙂

[tame current]

it's going to get patched sooner or later anyway

[halcyon coral]

What's the point of taking advantage of the free trail if PerplexityAI goes bankrupt in the process?
Just don't do that...

[tame current]

bin user

[chilly forge]

[tame current]

I highy highly doubt perplexity would go bankrupt by two people using pro version for free

yeah

[chilly forge]

lmao you really think only 2 users are doing that trick ?

[tame current]

not the point, I can't stop the others

[chilly forge]

I thought this meant only 2

Denial

[tame current]

two more, poor phrasing ig

[chilly forge]

Anyhow enjoy the trials

Enjoy the fun while it lasts

[tame current]

neither, it's called bineros unidos or something idk

[chilly forge]

Thanks but not needed 🙏

[tame current]

I can send link if you want

oh idk

oh that, I don't think so

I did

I've never heard of bullet binners though

uh what is that

err, I mean what do you mean independent and unrestricted binners

oh I see

[frank rivet]

Hello @tame current @storm grail please do not encourage users to engage in potentially illicit activity on this Discord server, and keep the conversation to Perplexity only related. We also have taken measures to ensure users are not abusing free trials, please do not do this again.

[halcyon coral]

You can use this channel

I think it will be done like ChatGPT Plus, otherwise I don't know anything official...

Yes, I think it will stay consistent.

[chilly forge]

I think you should introduce the $1 charge while claiming for trials I suppose

[wraith crow]

They have already fixed that issue now

[chilly forge]

Great 👌

[snow sun]

Thats good.

[frank rivet]

haha this does not happen often, and welcome! I can confirm we don't currently have plans for regional pricing.

[native agate]

@frank rivet can you explain why the AI no longer gives current news briefings, but directs you to the sites instead? Did you guys lose a legal battle or something?

just did a try on it on site and on the discord version. Gave the same "go to their site" reply.

[frank rivet]

Hello! thanks for sharing, we'll look into it. I tried reproducing and it does seem like it's just redirecting to sites. I'll keep you posted.

[native agate]

@frank rivet just tried it with beta. beta does not do this, but the normal one does. I do not know if beta however is pulling legit today news or weeks ago as it gave info on the subs still

[frank rivet]

Oh interesting, would you mind sharing a permalink of both beta and non beta threads with me? you can DM or share it here

[gusty blaze]

I just want to share an incredible experience I had with Perplexity.

I was looking for an article I read about 4 months ago. I remembered it vividly because it was a good article, but not enough to remember the publication’s name or the author. I just remembered the broad idea. So I tried Googling. I went down a google rabbit hole for ages. I would search something like “article about consultants that talks about capitalism” and just go for it. Obviously, I couldn’t find it

Thought I’d try perplexity with the same approach and this is what it came up with, and…
Perplexity found the exact article and put it in the sources

I wonder why it did that so well and from the first search. Maybe I’m just bad at googling?

[heavy portal]

you could ask it how it found the article and explain it's method step by step

[sage lily]

Thank you Daniela Gomez for doing something about the free trial abuse thing. I'm actually happy someone, me, discovered it because if no one did you wouldn't have known about it.

[chilly forge]

Yeah ig you were the first guy to report that trials are actually not getting finished even after 7 days right in #👑│pro-lounge

[warped marsh]

What is this channel you guys link that I have no access to 🤨 I am getting fomo now

I want my name to be in every channel of this server 👑

[halcyon coral]

It's the pro discord

[warped marsh]

oh so the 20$ plan one.. huh

maybe having my name in every channel of this server is not that important 😓

[tame current]

(sorry, wrong channel before)
Hi, does anyone know or has the best prompt to learn language and turn Perplexity into our conversation partner?

[noble quest]

The Pro Discord Join button is not working even though I have a Pro subscription. It opens Discord, but I don't see a new channel

[halcyon coral]

You have to leave the discord server and then rejoin with the Pro discord join button.

[noble quest]

Yep, you are right, it helped me, thanks!

[silver jewel]

I have the pro version of perplexity but the channel you posted shows me „Kein Zugriff“ 🧐

Aaah sorry. Just saw that comment with the rejoining

[sharp sierra]

Can I get a promo code if I sign up for the monthly Pro Subscription?

[halcyon coral]

There is a referral link, but no extra promo code.

[sharp sierra]

gib link 😄

[halcyon coral]

The link only works for one month, there are no other possibilities to get perplexity.ai cheaper.

For every Pro subscriber you refer, you get $10 off your next billing cycle and they get $10 off.

[chilly forge]

got ?

[halcyon coral]

Looks like a mixture of English and German to me 🤔

[sharp sierra]

Already became a Pro-fessional 😄

[near horizon]

I do love perplexity. It replaced all search engines haha

[near horizon]

Is it possible for perplexity to replace hey Google?

[lean kiln]

hi, guys

how i can switch to dark mode?

[near horizon]

The dream is that I can say hey Jarvis, ask a question and then it text to voice May the answer

[lean kiln]

is there dark mode of web version?

[near horizon]

There's a little person top right, click it it's in there

[lean kiln]

found. thanks!

[near horizon]

Yay!

[whole quartz]

are you pro user?

[hollow obsidian]

Hey, is the website for everyone broken?

Mine shows an error

[wraith crow]

Website is working fine for me, what error do you get?

[hollow obsidian]

Now its gone. And work again.

It said it was overloaded and the devs making a maintenance.

Or something like that

[noble quest]

Guys, I already see an opportunity to add the file in common perplexiy.ai.com too, not only in the beta. Has already been a new release?

[heavy portal]

must've just been turned on, I reloaded page and see it too

maybe it was an accident and we should be quiet about this 😉

[gusty blaze]

No

Keep getting an error about my file type being unsupported when I try to upload photo to Perplexity

[tame current]

Hi @gusty blaze , only text files and pdfs can be uploaded #📢│announcements message

[gusty blaze]

It’s pretty freaking good ngl

[floral merlin]

left or right? 🤔

[honest yarrow]

Howdy mates. Thoughts on running 70B uncensored? https://ollama.ai/blog/run-llama2-uncensored-locally

[tame current]

I find this better than Bing Chat. Any idea why it's still so unknown?

I wonder why perplexity is so fast at getting info from the web especially relative to Bing, when Bing has the entirety of Bing search at it's disposal.

The apartment I'm staying in as of this moment, I found using Perplexity, and it's a really good one.

It's rather difficult to read through all of google maps, and all of the fake reviews and so on.

[floral merlin]

I believe it's more strict than pplx's gpt-4 implementation and growing demand

[tame current]

Ah ok, I don't use it for anything too spicy

Overall would you guys say it's better than Bing Chat, GPT-4, and Bard topics that are not NSFW/NSFW adjacent topics

[tame current]

yup, definitely

bing is horrible

[tame current]

Bing has the best UI though, which sometimes almost makes me want to use it.

But I only use it for image these days

The Bing app looks beautiful and functions really well - but if the actual LLM doesn't work that good, what's the point?

Also, anyone have any other recommendations to try out?

[floral merlin]

100k claude can be use as a free "PDF Chat", pplx can also do that but not with long documents

and I also use claude to summarize 30+ mins yt videos tho 😂

[tame current]

I get the all black aesthetic, it looks cool for the most part. But at least make the citations (especially superscripted numbers) a different color (maybe the blue you have) so I can tell them apart without squinting.

[tame current]

Yeah, that's nice as like a 3rd opinion/maybe 2nd opinion LLM. Anything else that's good?

[chilly forge]

Thats the trick of getting users to use it instead of other tools

[chilly forge]

Damn

[floral merlin]

Bing UI is closer to chat experience than a knowledge research tool

[chilly forge]

Are you guys using edge as your default browser as well ? or just use it for bing chat

[floral merlin]

Yes I use edge over chrome because of these flags, especially smooth scrolling, boost refresh rate, and enhance-text-contrast

[tame current]

I use it for image upload

Look at the difference in these two images - on one you can actually tell the superscript numbers, and where parts of the information is sourced from.

[chilly forge]

I think those sources cards can be hidden ?

in the perplex

[tame current]

The cards are fine, but they need to make the numbers a different color (the superscript numbers for the citations)

[chilly forge]

oh right

[tame current]

like with Bing, the 1, 2, and 3 are purple against the black background so easy to make out.

[floral merlin]

you can play around with the UI while waiting for the team

[tame current]

you used an extension?

[chilly forge]

https://discord.com/channels/1047197230748151888/1136533130778525716

[floral merlin]

This should change all the citation style (similar to when they're :hover)

.citation > div > span > div {
  background-color: rgb(0 122 255/var(--tw-bg-opacity)) !important;
  color: white !important;
}

here's how my current pplx looks like

I also made <headings> and <strong> to use pplx's blue

easier to navigate through sections, bullet points, etc

[tame current]

cool, here it is as a uBO filter: perplexity.ai##.citation > div > span > div:style(background-color: rgb(0 122 255/var(--tw-bg-opacity)) !important; color: white !important;)

what I want is to change the whole color accent to red or green

[tame current]

I did it! (well actually claude did it but yeah)

don't mjnd the bar at the bottom, that's just how it looks on FF on android for some reason idk

[floral merlin]

Im fine with pplx's blue

[tame current]

wow, such a small change makes it so much better

[tame current]

yeah, it's nice. they just need to use it more.

also, maybe get rid of the discover section. It seems unnecessary.

When I started using it, I would think that it was my history (there is very little differentiation again), and wonder how i forgot about searching up these weird ass prompts.

I like discovery personally, but it doesn't have to be one of the main options

I also preferred when AI Profile was in the settings, I don't use it nearly enough for it to in the nav bar

[buoyant timber]

Do we have APIKEY calls

[wraith crow]

Hi @buoyant timber, Perplexity does not offer an API

[floral merlin]

#1136533130778525716 message
what are u guys's thoughts on my take?

[buoyant timber]

Do we have the plan to offer API

[floral merlin]

also no

[wraith crow]

Like zscq says, there are currently no plans to implement an API. But they will make sure to let the users know if this changes in the future.

[toxic hound]

hi ¯_(ツ)_/¯

[noble quest]

I don’t understand how the new LLaMa chat will be integrated into Perplexity AI in the future? What changes will be made with a professional subscription? LLaMa will combine the results with GPT-4, or will you choose between them before asking the question?

[floral merlin]

I've really tried to have a normal convo with llama many times, but haven't got a single satisfied answer so far

replace gpt3.5 with llama is somewhat understandable in terms of business, but I believe that won't happen anytime soon