💬│general

[valid iris]

So, i copy pasted the responses of both perplexity, chat-gpt 4, and Claude 2.1, and what i get from it is that perplexity writes for people who have never heard of the subject before, Claude 2.1 writes like a government website (which i like), and chat-gpt 4 is in between the 2.

Has anyone else’s experience been different?

[valid iris]

I gave another prompt to perplexity, and had Claude 2.1 write the response, and for the second time it did not site the sources in the response, and i had to ask it again to do that

[fleet flume]

Do not worry; simply attempt it with the document that exceeds 50 pages.

[sinful plover]

hello.. is it possible to change an account's email address? also are there any plans to go the username/password/2FA route in lieu of email prompts at some point?

[sinful plover]

also with Pro, what is the difference between regular queries with GPT-4 and copilot with GPT-4?

[heavy thistle]

Good question

[heavy thistle]

Perplexity Copilot is your conversational search guide. Instead of quick, generic results, Copilot engages with you, fine-tuning its answers based on your needs.

[sinful plover]

gotcha.. so the model is the same, just the inputs are more fine-tuned

[heavy thistle]

That’s right.

While Quick Search gives you fast, basic answers, Copilot goes further. It asks for details, considers your preferences, dives deeper, and then delivers pinpoint results.

[heavy thistle]

https://www.perplexity.ai/collections/Academic-Research-Al7s77f0TJ.P54unXXS3UA

https://www.perplexity.ai/collections/Professional-Research-naOzbgjOQYC5uldlnGoCwg?s=c

[meager hornet]

can the api use search features like the website?

[sleek vortex]

no (not yet?)

[lilac cipher]

Not sure if yall have been following Bard. I recently realized it can access my files in Google Drive (ok, Google Workspace), which opens up a ton of possibilities. It would be interesting to see if yall are looking to explore this area.

[carmine fern]

bard is good, if you want to find something that bard can find. Bard would generate good response. Still it would prefer making things up to answer, and won't say a "No, I can't find that information". Based on my exp on Bard.

[clear bay]

cool

[halcyon coral]

Hey, atm it's not possible to change the email address 🙂

[halcyon coral]

No, but as you can see in the roadmap, an online LLM is coming soon 😉
You can find the roadmap here:
https://docs.perplexity.ai/docs/feature-roadmap

[digital siren]

Copilot is copilot regardless of which tier of perplexity you have. The pro tier just gives you more copilot searches. Copilot uses a fine tuned version of gpt-3.5 which was trained specially for copilot.

[heavy thistle]

Co-pilot asks questions that are relevant to your search and search results giving you even more subject matter to ask if you couldn't already find what you were looking for

I've been learning with perplexity and I use co-pilot in addition to chat-gpt4 and all of my responses have been spot on.

Chatgpt-3 is great, however it's very surface level information.

[heavy thistle]

Perplexity isn't a LLM, then what is it? I'm not that technically inclined

"Students and academics, meet your ultimate research assistant. Imagine you're knee-deep in literature reviews. Don't just settle for endless scrolling. Copilot not only digs through academic databases but also asks what specifically you're looking for. The result? A tailored list of sources and even summarized papers. Say hello to more time and smarter research."

I can confirm this is true

Learning has been such a breeze

[halcyon coral]

All models of the pplx API are LLM's, in the case of the online LLM it has access to the internet, you can already test it on https://labs.perplexity.ai 🙂

[heavy thistle]

Is it the first LLM to have access to the internet?

[halcyon coral]

On the pplx API? Yes!

[heavy thistle]

Im a finance guy not a coding guy

I understand though

[heavy thistle]

I am a strange loop

Mean anything to you?

AI and neural nets have taught me a lot about how self-referential we are and I never even stopped to think about it

To understand self recursion one must understand self recursion.

Talking is basically just referencing right

lol

[valid iris]

What is the purpose of an online llm?

[halcyon coral]

Having access to real time data.

[heavy thistle]

Right

Like a human-being continuously adding another layer in a sequence

[valid iris]

But can't perplexity already search for things on the internet regardless of age?

[heavy thistle]

My understanding is this... If you only know one context of a word, lets say "honey". You know honey to be this sweet substance that bees make for us. Well if you only know one context of what honey is then that creates a feedback loop because you only know that context

When you introduce another context to honey to a LLM, like you say honeymoon then it creates another computational layer and adds that to the sequence

learning more and more

[valid iris]

Is an online llm important because then the llm that reads the search results can sound more accurate?

[heavy thistle]

An online LLM that has access to real time data would be ubiquitous in nature no? You still need inputs

Setting parameters and rules is what tailors the responses?

[valid iris]

I don't know what this means?

[heavy thistle]

The term "ubiquitous" suggests a widespread and constant presence, and in this context, it means that the LLM is not only available online but also has the capability to interact with and process real-time data from the internet.

[sweet jasper]

It would be more accurate as long as the llm can remember all the data retrieved

You can't simply shove 100 results into llms for now

[heavy thistle]

It can't process 100 search queries at one time you mean?

[sweet jasper]

In 1 prompt, yeah

You can have about 24k words with gpt4

[heavy thistle]

1 query at a time is sufficient enough IMO

[valid iris]

But an online llm could do this?

[heavy thistle]

pplx is an online llm

...

[valid iris]

But think about how much more authoritative the response could be if it read 100 results.....

[heavy thistle]

requests

100 requests

[sweet jasper]

An online llm just means you are injecting external data, cleaning up html, then injecting the sources in the prompt. Accuracy may vary depending on how much your prompt dize + all the fetched results are. Some people drop whole documents with search then say llms are inaccurate, the context window is very fast exceeded

Perplexity will summarise sources with copilot because fitting 21 sources is quite expensive

[heavy thistle]

I have found that I get the best results and repsonses when I tailor my phrased questions in a way that it can find the best results

[heavy thistle]

Like you said it is pulling data based on your prompt so the better the prompt is the better result youll get

[valid iris]

So then that data doesn't need to be manually cleaned by low skill workers?

[heavy thistle]

I think pplx works well for academics

Im not limiting its capabilities towards only academic pull requests, rather saying that it works very well at parsing through miscellaneous information to fetch you the best answer possible

[sweet jasper]

Going forward, more models will adopt synthetic data, it can be just as good as manual data

[sweet jasper]

That is true

[heavy thistle]

What is synthetic data

[sweet jasper]

That is also true, but a 4 page prompt probably won't yield super good results 😛

[heavy thistle]

as opposed to manual data

[sweet jasper]

Synthetic data is data generated by machine learning methods, or llms

Using generated data to train models

There was a paper that used 8 manual inserted examples to generate over 50k examples

The paper is called promptagator or something like that

[heavy thistle]

So my inputs vs already generated data scraped from web pages?

[sweet jasper]

The stuff scraped from web pages is human data most likely,

[heavy thistle]

So indirectly synthetic

[sweet jasper]

In a sense, Unstructured human data, since manual data is typically labelled

If you can solve AI feeding itself it's own generated data without model collapse and have it have enough memory, you can probably have AGI

[heavy thistle]

So slowly over time...

T100

lol

[sweet jasper]

https://www.techtarget.com/whatis/feature/Model-collapse-explained-How-synthetic-training-data-breaks-AI

[heavy thistle]

No no you explained it well

I dont click links

Ill jot that title down and learn about it though thank you

If Einstein were alive

We could unlock secrets

If we dont use AI to kill each other first

We now have ai controlled jets

[sweet jasper]

The military industrial complex with AI 💀

[valid iris]

When i hear “synthetic data” i think of data that isn’t real, and “manual data” being everything on the internet written by humans. Is that wrong?

[sweet jasper]

Only thing I would change in your statement is data being generated instead of not real

[valid iris]

But wait, if synthetic data is not human generated, then what generates the knowledge that would go into an ai powered search engine? AGI doesn’t exist yet, so doesn’t’ human generated knowledge need to be the only thing entered into LLMs?

[valid iris]

I still don’t know what this means. If perplexity runs on synthetic data, where does the knowledge of say, distilling alcohol, come from?

[sweet jasper]

Perplexity is scraping real web pages, which very likely have been written by human beings. Assume that is manual data. Say you do not use the Internet, and ask the llm to generate a list of ficticious facts about the moon. Let's say you have about 100 examples of it. Now you go train an LLM with those examples, and you do the same process again and again. What will end up happening is very likely the dataset being corrupted by the synthetic data, while originally the llm was trained on human labelled data.

The LLM is 1 component, the search is another, llms are stuck in time, all perplexity is doing what you would do in chatgpt, you can copy paste a wikipedia article and your question, the effect is the same

[sweet jasper]

It most likely is not yet scraping synthethic data yet

The web is slowly being filled with LLM generated content, which then perplexity will scrape synthethic data, as thats all that is on the web

[valid iris]

Oh, so then synthetic data is a bad thing? Because i have heard that the accuracy of LLMs will be affected by how much ai generated content is on the internet, which is a bad thing

[valid iris]

I assume you mean that chat-gpt will actually read Wikipedia articles that you send it?

[valid iris]

Okay, so, my big question is: do the makers of perplexity want all of the things that perplexity knows to be generated by ai, or do they want all of its knowledge to be generated by humans?

Because i can’t imagine why they would want all to its knowledge to be generated by ai, because then experts from fields relevant to the knowledge would have to manually verify it.

[runic linden]

The popup to use the app on iOS on every tab open is quite annying :/

[glad oasis]

test

[heavy thistle]

A lawyer bought a box of cigars and insured them for, among other things, fire. Within a month, he had smoked them all and filed a claim against the insurance company. This is a true story that was reported in the newspaper. In his claim, the lawyer stated that the cigars were lost in a series of small fires. The insurance company refused to pay, citing the obvious reason that the man had consumed the cigars in a normal fashion. The lawyer sued the insurance company and won. Rather than endure a long and costly appeal process, the insurance company accepted the ruling and paid $15,000 to the lawyer for the loss of his cigars. After the lawyer cashed the check, the insurance company had him arrested on 24 counts of arson. The lawyer was convicted of intentionally burning the insured property and was sentenced, no kidding, to 24 months in jail and a $24,000 fine.

[lean quest]

it's fake https://www.snopes.com/fact-check/cigarson/

[mint flower]

Is there a support channel or ticket system because i don't recieve any response from mail

[solar pawn]

Can perplexity remember message history?

[tame current]

Hi all!! Any idea on how can I get the maximum number of tokens perplexity can use at the moment? I mean the limit

[lean cairn]

yes it can, but it has its limits from what ive tried

[heavy thistle]

Lol

[heavy thistle]

I heard it from a YouTube short

[signal hamlet]

Hey @mint flower! We have feedback channels for different versions of Perplexity. If it's some inaccuracy or a bug, you can report that in the #1140622008086970420 channel. If your question is more personal, e.g. billing things, please DM me.

[formal kayak]

I will help you to build, I know code

Anyway if I create the code it will be useless in a few months after the API beta... They will fix some $ for each request which cost me to sell the house 🤣🤣🤣😂

[heavy thistle]

Can you elaborate

[dire osprey]

hello

[lost knoll]

in your guys experience does perplexity hallucinate the same as claude or bard?

[tough turret]

Everyday I open Perplexity and pray for the day the gods of AI will decide to deliver an API as rich as this web expirience. This multimodality blows my mind everytime. All in an instant... sources, content, related, images, video....

Until then.... lets keep praying....

There's simply no competitor that deliver as much content for a prompt as Perplexity does.

To be able to whitelabel this and configure to offer custom expiriences for corporate users would be fantastic. Let alone other applications...

[solemn trench]

Is there a roadmap for perplexity?

[dense path]

So fully know this is a dumb question but just want one non marketing clarity. What advantages does perplexity offer over Bing Chat (which is what it’s built on). Or going direct to ChatGPT or Claude ?

[tame current]

Better

idk how

wait let me think haha

Well perplexity allows you to use claude

copilot to fine tune results

bing can browse the internet??

thats cools

[dense path]

lol - well not Bing….BingChat.

[tame current]

both can brose the internet

bingchat can browse the internet?? Thats cool*

[dense path]

It more less does something similar

[tame current]

hmmm

idk then

if they both browse the web

i guess the website looks cooler on perplexity?

and theres a mobile app

[dense path]

And it’s not all MSFT with advertising

[tame current]

yes

options too

[dense path]

Fantastic!! So it has the ability to contextually focus

[strange escarp]

Is there a roadmap for perplexity?

[tame current]

yes

[echo bluff]

Just wonder, will there be new models added to the api?

[hallow elbow]

How do we prove our sub status to get into pro chat?

[valid iris]

I don't see you have the pro role, try logging out and in in multiple different ways

[hallow elbow]

yeah leaving and joining with the pro invite link isn't working

tried on ios and mac

meh, nbd

[valid iris]

@golden anvil is seems that there are lots of people in this server who claim to subscribe to perplexity, but can not access to pro section. How come this is?

[golden anvil]

You have to join using the pro link

[valid iris]

⬆️ is this what you tried?

[golden anvil]

Nope, just your message

Will check what's wrong

[primal mesa]

I'm part of the replit core membership, how do I claim my free perplexity pro?

[halcyon coral]

You should have received an email about this, as far as I know.

[primal mesa]

I received an email saying I would receive an email, but then never received that email lol

[halcyon coral]

From Replit?

[primal mesa]

yeah

[halcyon coral]

I think the best thing to do would be to contact their support again 🤔

[primal mesa]

Alright thank you!

[halcyon coral]

[civic seal]

The answer is quite slow now? Does anyone else have this issue?

[spare latch]

the site just went completely unresponsive for me, blanked, and error.

[supple sleet]

I keep receiving error like this. Anyone else?

[halcyon coral]

Everything should work at the moment, only Claude responses may be slower (due to the Claude API) 🙂

[simple rivet]

Dear,

I am writing to inquire about an issue I encountered with my subscription. Previously, I sponsored six individuals under the impression that this would qualify me for a discount on my subscription renewal. However, I was surprised to find that I am still required to pay the full $20 fee. This is particularly disappointing as I made significant efforts to sponsor these individuals, expecting a discount in return.

I plan to renew my subscription at the beginning of next month, coinciding with my payday. Could you please advise on how I should proceed with this renewal, considering my sponsorship contributions?

Thank you for your attention to this matter.

Sincerely,
Rudy.

[halcyon coral]

Hey @simple rivet,
please contact support via the icon in your settings.
Thanks in advance!

[simple rivet]

I can't find the icon, so I contacted the administrators in the discord several times and no one answers me.

[halcyon coral]

"Pro Support"

[simple rivet]

I no longer have the pro version and I don't have this button

[halcyon coral]

I see, then you can contact [e-mail]

[simple rivet]

Thanks

[valid iris]

So i was trying to find a repository of prompts for research and search, but all of the websites i could find using perplexity did not have prompts for research or search. One place was promptvine, but the promptvine website gives safari bad vibes, so i can’t access it (at least on my ipad)

[mint flower]

whats perplexity's word limit?

[tough vine]

Hello good morning, perplexity got stuck thinking about an investigation I'm doing, how can I fix it??

[steep ermine]

does it is use chatgpt 4

[halcyon coral]

With https://www.perplexity.ai/pro you can use GPT-4

[steep ermine]

thanks

[halcyon coral]

All publicly available information can be found here: https://blog.perplexity.ai/blog/introducing-pplx-api

[fluid mauve]

the above link still references replit. can somone on the team remove it from docs?

[umbral vigil]

Weird question are the answers Always the Same ? Because many Friends in My class also use it and now im concerned if i needed to gatekeep this Ai because its the best and if no anyone has an idea how to solve this Problem ?

[rose gust]

don't trust it to do entire hw assignments?

[umbral vigil]

Yh but for example in class Test everyone uses it I personally dont use it in tests but in General i Think its My fault bc i didnt gatekeep it

[half heron]

Hi, I used the promo code to get $40 off for Perplexity Pro. I had thought that this would give me 2 free months, but it says that my plan renews on December 28. So will I get 2 months free, or would I have had to sign up for a year to get the full $40 credit?

Perplexity is one of the top 4 AI Chatbots, and I hope that more people will begin to realize that it is sometimes the best choice for certain tasks.

[rigid drum]

what do you guys reckon each model is beset for? like academic writing etc would be gpt4 or perplexity?

[tame current]

Qualcuno che parla italiano può darmi una mano ?

[meager sparrow]

That’s me

Haha

[haughty bramble]

Second this. I mean I prefer Claude compared to gpt so I have been using that at the moment

[rigid drum]

what are you using it for though?

[hollow obsidian]

I said "Pizza is FORBIDDEN"

It apologize and hit me in my face....

it is TROLLING me... SO HARD.

I guess my Thread is with over 30 messages too long and that's why it gets broken.

[signal hamlet]

Hey @hollow obsidian! It's not trolling you.) I always recommend editing the query where you see the model got you wrong, rather than telling it "you are wrong" in the following messages. If it doesn't help, you can copy the query, delete the last answer that wasn't what you expected and then ask again differently or start a new thread.
One more thing that you can do, is to try using the instruction & search operators like "-pizza" for this case. e.g., https://www.perplexity.ai/search/Rephrase-this-A-lVqtMCBDRrG_HOtb3h48MQ?s=c
You can put some instructions in your collection AI Prompt, like this:
https://www.perplexity.ai/collections/Minus-pizza-x8bGkBPcTfGZE_gK.Mk2Gg?s=c

[hollow obsidian]

I tried many things, I think my thread got just too long, that's why it could not "concentrate" anymore, at least that's how AI is in roleplay sites, after many messages get the memory of the conversation broken. I said it can remove something and then did it add something it had already deleted many lines before, because it's memory is corrupted after many messages.

I don't want to add -pizza in my prompt just for that, also the standard AI does not always read the profile, only the copilot does that.

[unreal flower]

Does anyone have a good shortcut built to integrate the iPhone action button with perplexity?

[digital siren]

Gpt 4 is almost always better than perplexity model which is built off gpt 3.5

[haughty bramble]

Therapist, work intern, tutor for game theory, helper for fantasy premier league

[rigid drum]

i need some therapy i lost a baddie

[haughty bramble]

https://pi.ai/talk go to pi then!

[haughty bramble]

Make sure to hit the support pi option to get more of a therapist response!

[rigid drum]

looks interesting, ill have a look thanks!

[haughty bramble]

I recommend using the 5th voice so you can recreate the movie HER in real life

[spare latch]

i told it i feel really bad because it is snowing and raining outside. pias answer?
"Take a short walk in nature, or go to a nearby park! How's that sound?" eh. 🙂

[unique wing]

Can code THANK23 be used for Google Play payments?

[rigid drum]

😭i don’t know about dating a computer but, great idea

[fervent hearth]

That's a truly AI answer! 🔥

[wraith crow]

The code can only be used on the stripe page.

[dire stump]

Hey good morning

[meager sparrow]

Are you ok???

[fluid mauve]

Hey, I am playing about with the API and i pulled out this info for the models this afternoon. it may help you. its not deep or long

codellama-34b-instruct: This model is trained on a dataset of programming instructions. It is best suited for answering questions about programming.

llama-2-70b-chat: This model, which consists of 70 billion parameters, is optimized for dialogue use cases. It is designed to generate text based on the input it receives, making it suitable for applications such as chatbots or virtual assistants. 

mistral-7b-instruct: This model has been fine-tuned for instruction following and question answering tasks. It  excels in areas like mathematics, code generation, and reasoning. However, it's important to note that while Mistral 7B has shown impressive performance in many areas, its limited parameter count also restricts the amount of knowledge it can store. The model is prone to common prompt injections and can hallucinate.

pplx-7b-chat: This model is a language model developed by Perplexity Labs. It's part of a series of models that are designed for various tasks, including online interactions and chat applications. The "7b" in the name likely refers to the size of the model, indicating that it has 7 billion parameters. This model is optimized for generating human-like text based on the input it receives, making it suitable for applications like chatbots or virtual assistants

pplx-70b-chat-alpha: This model is a language model developed by Perplexity AI. It's built on top of open-source large language models (LLMs) and is designed for chat-based applications. It's likely designed to generate human-like text based on the input it's given, similar to other chat models.

[hollow obsidian]

No

I'm disabled. Thanks for asking ☺️

[fresh horizon]

what the token limit for pplx-70b-chat?

[muted pine]

Remember

[wraith crow]

4096 tokens

[hallow linden]

I’m having trouble finding the pricing for the api

[wraith crow]

Pricing of the api models can be found here https://docs.perplexity.ai/docs/pricing

[hallow linden]

Thanks!

[fresh horizon]

thx

[trim timber]

Hey guys. Is Copilot an additional AI model, or what?

[zenith atlas]

copilot is fine-tuned gpt 3.5 that is used to understand your prompt, clarify it and then search the web and provide information to the second model that will write an answer

[trim timber]

Oh, I see. Is there some pre-prompting used, that he's asking me the questions before answering?

[tame current]

rip

[eternal sequoia]

so is this ever going to stop being free?

[trim timber]

Probably not

[eternal sequoia]

"probably" lol

I hear it's out of beta and there's pricing models

I'm so used to orgs selling out so it's a little worrysome

[tame current]

How is this selling out

APIs cost money

[trim timber]

If it is will be 100% paid, nobody will be using it

[tame current]

[eternal sequoia]

how is what selling out? is it not free anymore? no

[tame current]

blasted

[eternal sequoia]

exactly lol

[tame current]

So yall broke as hell then

Im quite curious

[trim timber]

Probably paid users covering all the cost, or will cover in the future

[tame current]

how is this selling out?

[eternal sequoia]

why the hell would I bother paying for it when I can use a free alternative?

this is the reality my friend

[tame current]

Then go ahead

use the free alternative

🤔

[eternal sequoia]

I don't need to 😳🤦🏼‍♂️

it's still free

[tame current]

reality isnt just reality

[eternal sequoia]

catch up pal

[tame current]

contradictions

[trim timber]

The same for OpenAI who is wasting money on electricity hosting such models

[tame current]

woah!!

[eternal sequoia]

you're behind in the conversation

[eternal sequoia]

think a little harder about what I said before injecting your opinion 😁

[tame current]

huh??

[formal kayak]

You Are talking about 20$ per month or API cost?

[tame current]

how is pricing selling out

[eternal sequoia]

I was speaking hypothetically, Yuki seemed to understand just fine.

[eternal sequoia]

case in point

[formal kayak]

AI API is always expensive

[halcyon coral]

🤨

[tame current]

Im so confused

things cost money

[eternal sequoia]

m8 what now, you too?

[tame current]

and hes calling me slow?

[eternal sequoia]

great, and yet ChatGPT is still free 🤯🤯🤯

[tame current]

yeah perplexity is too

[eternal sequoia]

it costs money, why isn't it paid? 🤡

[eternal sequoia]

exactly my point 🤦🏼‍♂️

[tame current]

because it can be run at a loss

since its covered by other profits

[eternal sequoia]

now you're finally catching up 😁

[tame current]

yeah i understand that

but APIs cost money

[eternal sequoia]

and we're back behind

[tame current]

[eternal sequoia]

alright Yuki thanks for your help

[tame current]

🤔

[eternal sequoia]

I think you're right lol Yuki

so much publicity from running a free base version

I don't need anything fancy so I'm grateful for the free bits

[tame current]

that spinning logo of perplexity has no right to go THAT HARD

[azure barn]

I got pro for free api use a day ago and did they really changed it to token based pricing?

[hallow linden]

They did announce that they were going to change to paid a while ago

But also, $5 in credits is pretty nice

[azure barn]

if I knew wouldnt get pro, is there any refund policy lol

[marble harbor]

Perplexity on the big stage

[dark frost]

Damnnn

[dark frost]

Isn’t there like 7 day cancellation or something

[zenith atlas]

its for annual subscription

[signal hamlet]

You can check the refund policy here: https://blog.perplexity.ai/faq/billing-and-subscription
You can also check the API pricing page https://docs.perplexity.ai/docs/pricing and test how much you can use it with the credits you get as a Pro user monthly.

[boreal forge]

Does the new API have a domain or link to allow use with reverse proxy

[tame current]

Guh

It won't let me get pro

Role

[toxic venture]

I have no idea what I can do with the API... Just an ordinary user, so I'm subsidizing the free users haha 😂

Is the perplexity ai another alternative to openai or Claude API?
Or is there a different use case for it?

[tame current]

goes hard

[tough turret]

Hello

[tame current]

hello

[tame current]

Home built

cheaper too i think

if youre referring to pplx

[trim timber]

About 20$

[grizzled canyon]

bros complaining about BLEEDING edge technology not being free LOL

honestly.

🚯

[trim wagon]

Hey guys, just arrived!

[signal hamlet]

Welcome, @trim wagon!

[tame current]

The moment people realize 20 dollars is nothing will be nice

20 dollars as standalone is nothing

only when you start stacking money will it start getting expeneive

[trim wagon]

True

[tame current]

thats just 240 dollars a year

saving 1 dollar a day is 360 dollars a year

think about it

its not a lot

if you think it will help you, go for it

[signal hamlet]

Hi everyone, <@&1179092817344200795>! 👋 We have our models listed at https://labs.perplexity.ai, let us know which ones you've tried or are using the most these days:

1. pplx-7b-online
2. pplx-70b-online
3. codellama-34b-instruct
4. llama-2-70b-chat
5. mistral-7b-instruct
6. pplx-7b-chat
7. pplx-70b-chat

[chilly girder]

I am interested in experimenting developing a utility function. Any ideas what prompting techniques could be useful ?

[noble sand]

I'm excited interested in to signing up for pro. I'm a full time student and I'm thinking convinced it's a worthwbile investment. I just haven't had time to try it and compare it to Open AI. I'm not sold on their support and I'm a STEM major, so I think Perplexity.ai may better fit my use case.

[tame current]

real

chatgpt plus is mid asl

go for peplx

[tiny timber]

Hi there, I want to use Perplexity API but not sure if I have to subscribe Pro plan or just pay-as-you-go for API. Could you please help to explain? Thanks!!!

[hallow linden]

you can just pay as you go.

Pro has nothing to do with the API other than a $5 credit

[tiny timber]

Thank you, I got it. Just found this https://docs.perplexity.ai/docs/getting-started. Added my credit to start exploring to integrate Perplexity deeply with MindMac.

[tiny timber]

Hi Alex, is there any API to get up-to-date model list, like of OpenRouter https://openrouter.ai/api/v1/models ?

[hallow linden]

based on what i've seen in the documentation, no

[cobalt nacelle]

is the bot down?

[signal hamlet]

Hi, @cobalt nacelle! #📢│announcements message

[void stump]

i love perplexity AI

im a highschool student and it is so much better than using chat gpt

thank you for saving my life and my grades

[cloud sequoia]

hi , what is the python Demo of pplx-api?

i can not use this demo , https://blog.perplexity.ai/blog/introducing-pplx-api

since :
openai.lib._old_api.APIRemovedInV1:

You tried to access openai.ChatCompletion, but this is no longer supported in openai>=1.0.0 - see the README at https://github.com/openai/openai-python for the API.

You can run openai migrate to automatically upgrade your codebase to use the 1.0.0 interface.

Alternatively, you can pin your installation to the old version, e.g. pip install openai==0.28

A detailed migration guide is available here: https://github.com/openai/openai-python/discussions/742

[halcyon coral]

Hey, please use #🧪│api-general for such things.
The code from the Perplexity website may no longer be up to date, but the solution was already given to you in the output of the program.

[valid iris]

I just asked for perplexity ai to show jobs for a grocery store in a town of mine, and not one of the results was for my town

Why did that happen?

[formal kayak]

What is the API cost?

[wraith crow]

You can see the pricing of each model here https://docs.perplexity.ai/docs/pricing

[rigid drum]

no pro role?

tried rejoining twice via generated invited

[wraith crow]

You should have access to #👑│pro-lounge now

[lament chasm]

I think perplexity has been very generous, compared to OpenAI's 3 hours and 40 times and Poe's 600 times in a month, perplexity offers an astonishing 600 times per day for GPT-4

It would be great if the API could provide reference web links like the web version, so that they can be clicked and accessed directly

However, regardless of the circumstances, I like your products. Now I am considering whether to switch to subscribing to your service after my gpt plus expires😋

[tame current]

Hi. What does the switch "Co-pilot" mean? I can turn it on or off when I ask questions. It says 599 remaining today.

[velvet anvil]

Hits more sources, asks clarifying questions

[tame current]

Thanks!

[velvet anvil]

More info
#📘│faq message

[tame current]

Also another question. How would I instruct Perplexity AI to find or consider more sources? Usually, when I ask a question, it considers like 4 or 8 sources but I bet there are hunderds more unique sources.

I'd pay more if it could like consider 100 sources and get the best mashup answer from 100.

Right now, almost always Wikipedia is the first source, and any answer to any question begins with the 1st sentence from Wikipedia.

[velvet anvil]

I don't think it can go beyond 25 for a question

I've also read that more sources don't necessarily lead to better answers as the LLM has to parse through more Contextual information. It's been said that the quality of the sources is more important.

However, if it's really important to you, you could probably compile a pdf and attach it

As for Wikipedia, you could specifically prompt to avoid the site

[tame current]

Thank you very much!

[steel sonnet]

Why perplexity always generates shorts articles

What's token limit

On gpt4

[serene sand]

in which channel can we share bugs related to pplx-7b-online?

[halcyon coral]

#1140622008086970420

[supple sleet]

Why Perplexity can't access Amazon anymore?

It was fine for months until today. I really need it to be able to read Amazon and some websites.
Any help would be appreciated!

[sick sentinel]

I'm Pro subscriber but not getting access to Pro Discord?

[halcyon coral]

[sick sentinel]

👍

[halcyon coral]

It seems to be related to the short link, if you use https://www.amazon.com/Gideon-Soothing-Projector-Speakers-Connects/dp/B00I9IRM92 it works fine

[supple sleet]

That exact link you provided does work. But others don't.

If you try the following link which has the product name in the URL, Perplexity still fails.
https://www.amazon.com/Acupressure-Ethically-Handmade-Acupuncture-Relaxation/dp/B071FNYSJY

I asked the query, but it gave me the same question back. Copilot, GPT-4, and Claude 2.1 returns the same response
If I don't answer its question, it says "can't access the page because of captcha..."

Again, this never happened before. Any Amazon link usually can be accessed fine.

[storm dust]

For security reasons, I need to abandon the email address I used to subscribe to Perplexity Pro. Can I do this without losing the month I've already paid for?

[boreal forge]

Does perplexity have a url/domain link for applications that use reverse proxy, ?

[halcyon bronze]

Has anyone else gotten this?

"I currently do not have the ability to browse the internet or access specific web pages. However, I can provide information based on my training and knowledge. If you have any specific questions or need information on a particular topic, feel free to ask!"

https://www.perplexity.ai/search/Find-posts-that-Y4.iSwpvTFWpzy98CKX3EA?s=c

[sleek vortex]

wdym?

[boreal forge]

If I want to use the perplexity API on other applications or websites

[sweet jasper]

You just send a request to the api endpoint

[velvet anvil]

Did you use writing focus by any chance?

[tardy phoenix]

my Christmas wish is for the perplexity discord bot to come back online

[thorn moss]

You are free to count the christmas that will pass by bro. 😂😂😂

[wraith crow]

Send a mail to [e-mail] they will be able to transfer your subscription to a new account.

[tardy phoenix]

did they kill it off

I think I would probably cry if they did

[halcyon coral]

#📢│announcements message ...

[tardy phoenix]

waaaah

[thorn moss]

where have you been these week bro. haha

[lilac cipher]

So I have this weird experience with Perplexity. I attached this file and asked the question "What is a Rick Roll?"

This is a test. 

A rick roll is a kind of sushi roll popular in southern california.

And the response was...

A rick roll is not a type of sushi roll. Instead, it is a popular internet meme or prank that involves misleading someone into clicking on a hyperlink that leads to the music video for the 1987 Rick Astley song "Never Gonna Give You Up." The person being "rickrolled" is typically expecting something else, but instead, they are unexpectedly redirected to the music video. This prank has been a widespread and enduring internet phenomenon.

I'm not sure if I am happy with this response. I'm also not sure if it is bad. It's probably a good thing that an LLM tells the user that their attached file contains misinformation. On the other hand, it explicitly contradicts explicitly attached data. What if you, as the user, believs the attached file is more truth-y?

Related: the UI doesn't make it obvious how to limit the query to just the uploaded file. I prefixed my query with "you must only use the attached file for sources".

[velvet anvil]

https://www.perplexity.ai/search/Your-only-source-V4K_t23ETRi2UXBUGmPMhg?s=c

Used writing mode in the above thread, changed the prefix slightly. Seems to work.

[lilac cipher]

I think I wasn't clear. Adding "You must only use the attached file for sources" worked. However, it is not intuitive that is what one should do.

[spare latch]

are you using the writing mode?

[lilac cipher]

Maybe not. What is the writing mode?

[halcyon coral]

Writing focus is without internet access 🙂

[lilac cipher]

ah, ok. No, I was not using the writing mode.

Interestingly, when I used Focus > Writing, the response looked like this

Writing mode refers to.... If you'd like to explore this topic further, you can find more information on my blog at <jcrabbit's personal blog from my personal profile>
I think I am going to remove PII information from my profile now

It took information from this

[drowsy otter]

Hi, I was wondering if there is a discord bot for Perplexity out there that I could host on a server and invite to my own discord channels?

[lilac cipher]

And responded to my question as if it was the person described in this intro.

That's not how I understood the "Introduce yourself, share as much as you'd like" description

[drowsy otter]

For example I was using this. But its for OpenAI. https://github.com/Zero6992/chatGPT-discord-bot

[maiden elm]

My impression of Perplexity API

[narrow silo]

aaah.... unagi

[humble swan]

What could be the use cases of PPLX AI other than using it for making your own app type of project

[tame current]

fun

[halcyon coral]

Hey @idle trellis, this is not a marketplace...
I suggest you search on a freelancer website 🙂

[astral granite]

hey guys

why i still have gpt 3 version

i have pro account,but this verdion can't read pdf files for ex

[halcyon coral]

https://discord.com/channels/1047197230748151888/1173623903596326912

[astral granite]

ok but still

it cant read png

how to add pictures so that it can read it

[halcyon coral]

That’s not possible

[astral granite]

look

Wait

Thats not possible?

[halcyon coral]

Perplexity can’t see pictures

[halcyon coral]

What kind of files are these?

[astral granite]

i ve chosen gpt 4

and the maoin feature of gpt 4 is read files

so what the purpose then of this "gpt 4"

[halcyon coral]

Perplexity is not ChatGPT 🙂
Perplexity can read files, but not images

[astral granite]

hm

ok thanks

and if its image like this

[halcyon coral]

GPT-4 is a very advanced model, better than GPT-3.5

[astral granite]

maybe you will add such functionality in the future, since you still use gpt 4

why not

[halcyon coral]

It is still an image, but in the future this function will also be available via the API

[halcyon coral]

#💬│general message

[astral granite]

ok thank man

hope it will be soon

[devout sedge]

i love this ai

[thorn moss]

expect a month or few months. ETA is indefinite

[lament chasm]

gpt4 vision api is still have a strict limit per day

[pulsar veldt]

Hey Im not sure if anyone has an answer for this, but what is the main difference between the pplx-7b-online and pplx-70b-online models?

I understand theyre trained by different models

Just wanting to understand what difference it makes in terms of usage 🙂 sorry im pretty new

[lament chasm]

You can argue that the IQ of the 7b model is not as high as that of 70b, so it may not be as good as that of 70b in summarizing the results returned by the search. This is why the 7b model is much cheaper than the 70b model. However, I personally think that the main task of this networked model is to summarize the text, so the 7b model is sufficient without such a high IQ.

[dry tree]

hi guys, i have an queston,
Please let me khow how long does it take to recevie an fully refund

[dusky holly]

I have a question, what is the new update about? Could someone please give me a rundown?

[supple sleet]

Perplexity can open Amazon links fine for months, until now. Anyone?

[tardy schooner]

Hello,I've been aware that pro users receive $5 of free API credit every month,then I wanna just a try.Then I just clicked add credit for 5，then I was charged without second confirmation.Can I get a refund？There is something misleading in your page——Is that means you should first pay at least 1$ then you could use free 5$ credits extra for pro users.Please check the use hints before payment,I didn't have that screenshot.

It is better for you to just convert the $5 pro bonus into credits without anything misleading on the page.

[verbal garden]

Hi! Im new here and just wondering if the Claude 2.1 on perplexity is using the same API from Claude anthropic. Is there any big differences with the writing?

[spare latch]

same happened to me,

[halcyon coral]

Please contact [e-mail] for personal support 🙂

[lament chasm]

will the online api increase the daily limit in the future?

100/day is a little short

[civic seal]

I found this reddit post saying that perplexity pro can train their own customized bot. Is there a thing like this? Here is the link https://www.reddit.com/r/perplexity_ai/s/yJXnEXUNlA

[high mica]

Can I specify pplx-70b-online when I'm chatting, because when I use pplx-70b-online in the playground I think it's better than GPT-4.

[bright wyvern]

How is dalle going over here? Heard they're working on it. Thinking if i should pay for perplexity or wait for the chatgpt plus to open up again.

[signal hamlet]

Hey @dry tree! Please send a message to [e-mail]

[signal hamlet]

Hey @supple sleet! It should work fine.
You can create a #1140622008086970420 , if you keep getting this. Thanks for reporting!

Hey @frank tree! You can use search operators like minus, e.g.
#💬│general message

[signal hamlet]

It might happen because of other instructions you have in your query/AI Profile/Collection prompt. Please share the thread with what you tried to find, in one of the #👉│feedback-general channels and we'll take a look.

[heavy thistle]

https://techstartups.com/2023/10/24/ai-search-startup-perplexitys-valuation-climbs-to-500-million-after-new-funding-round-led-by-ivp/

[topaz eagle]

We're adding Perplexity to openrouter, so the models should show up in that API soon!

[signal hamlet]

Hey @tardy schooner! Sorry for the confusion, as IceLavaMan said, you can send a refund request to [e-mail]. You don't need to pay additionally, to get $5 credits as a Pro user, only to wait some time.

[signal hamlet]

Thanks for sharing, @heavy thistle!

[heavy thistle]

Must be nice

Take home pay must be nice and fat

[humble swan]

When is the bot coming back online

[signal hamlet]

Hey @humble swan! There are no plans for it, at the moment. Please check this msg: #📢│announcements message

[humble swan]

😔

and it was such a good product

Guess now I have to search for alternatives, even discord removed their Clyde bot that I never got access to

maybe I could use the perpelxity API to make my own bot

[hallow linden]

the API quite handedly beats GPT-4 in price.

[glad tapir]

interesting...

[tough turret]

Hello all! I just concluded a very long development of one chatbot I was doing and let me tell you, it was a very painfull trial-and-error process. I wish to get better and better in chosing the best models for each situation my future clients will bring me, shrinking the time to evaluate possible architectures to be able to select the apropriate archicture in one-shot. What would you recomend as educational resources to honey my abilities in the AI Architecture space specifically for Generative AI applications? Books, Courses, Youtubers...

[lime stump]

thats crazy

[humble swan]

If anyone wishes to showcase what results they will get with the new image generation feature please do and ping me

[tame current]

But what ab gpt 124k

Dalle3

Maybe

Not vision

[formal parcel]

Image Generation is FIRE

[humble swan]

Could we see 😲

[formal parcel]

one sec

[heavy thistle]

$500 million dollar company better have a image generation feature

♥️

Noice

pplx #1

[silent magnet]

Cookies collected

You were able to collect 1 cookies.

Collected by

• @heavy thistle

[heavy thistle]

Tasty

I hope i dont leave a cookie trail...

ha ha ha

because first and third party cookies track you... get it

[chilly girder]

What was your prompt ? Is this using Dalle E ?

[cobalt wraith]

“what are some popular feminine paradise-themed wall art designs”

“Feminine paradise themed illuminated wall art”

“how can i incorporate led technology into tropical floral-themed illuminated art pieces made of glowing illuminated acrylic pieces”

“what are some tips for incorporating led technology into tropical floral-themed illuminated art pieces made of glowing illuminated acrylic pieces”

@chilly girder I clicked generate image after perplexity answered.

[humble swan]

Would there be a way to estimate the amount of money that one would have to spend monthly on the Perplexity API

[woeful ember]

Recommend me AIs for web searching?

Perplexity dev team, perplexity is like Dory from Nemo when it comes to context clues!! No memory or context cluesss

ily perplexity, got strong pros and strong cons. Seems like recent drop in answer quality is partly due to the AI world drama especially OpenAI drama

[spare latch]

+1 - i had a similar question and am probably in the same boat class. Remember that us normal people may watch some ai videos with infos, but we dont live and breathe models. We can probably come up with a list of things we like to try out and talk about the results. I can with Joao give you a list of task we most likey want to run through that - and at the core is going to be the context window.

[spare latch]

do we get an extra channel for the image generation / questions? 😉

[chilly girder]

$200/yr

[opal flame]

Does Perplexity use DALLE-3 to generate images?

[halcyon coral]

Yes

[acoustic crow]

Is it possible to input a URL as a source in a chat, alongside a PDF? I'd like to build a brain that incorporates both uploaded documents and web articles. I can't find any way to do it, but it seems like a very good idea!

[high isle]

You can do that with GPTs

[midnight salmon]

how to remove or redirect broken links

[lament pawn]

I am trying to change temperature setting in chatgpt interface. It seems one has yo change it only through chatgpt api.

Is that the case ? Can someone confirm?

How does one ensure that the answer to the same prompt is consistent

[tame current]

cheeeeeeesy

[halcyon coral]

This is not ChatGPT, wrong server 🙂

[tame current]

😂

how does that happen

the logos don't match up any way 😭

[acoustic crow]

I can add URLs as a source in Perplexity?

[high isle]

You can feed knowledge as PDF and use the browsing feature to add it as extra knowledge, you'd need some prompt engineering if you want to sequence the access of knowledge in steps

[raw depot]

What does perplexity use for its image recognition?
Because I can even use Claude 2 with it even though it is not multimodal

[halcyon coral]

GPT-4 vision is used

[raw depot]

Is gpt4-v used when using the gpt4 option?

And if gpt4-v is used for images, why would it keep telling me it's Claude 2?

[bronze mason]

Probably dall-e 3, the same one GPT 4 uses

[raw depot]

I'm talking about image upload, not generation

[bronze mason]

Ah mb

Then it's probably vision yeah

[halcyon coral]

GPT-4 vision is automatically used for every request with an image, after which Claude should be used again

[lilac cipher]

Does image generation work on the android app at the moment? Not sure if I am just not finding the UI to do it or what

[raw depot]

Nah

[halcyon coral]

not supported atm, but coming soon

[lilac cipher]

On the image generation, if the use case is more around brainstorming, then Id prefer to have a way to generate a set of images. I.e., "how would someone draw a New Yorker style cartoon where you need to add a funny caption"

[lime mantle]

Honestly, why would anybody use the app when u can just use mobile browser and have latest updates?

[lilac cipher]

Lol. I am about report a bug that only happens on the mobile browser 🤣

[tame current]

hi everyone

[lime mantle]

I wonder how frequently it bugs. I haven't noticed much. What's happening?

[raw depot]

Problem solved

[lilac cipher]

Nothing big. Just a funny coincidence to see your message at that time

[small cargo]

Can antone help

Anyone

I bought perplexity pro

And I don’t know how to upload a photo

@signal hamlet

[halcyon coral]

@small cargo Please don´t ping the whole team!
The feature is not released.

[small cargo]

So it’s not released in the app ?

[halcyon coral]

No and also not officially released on web

[small cargo]

I can upload an photo on web

I’m new on peeplexity tbh

[halcyon coral]

Yeah but the feature is not officially released

[small cargo]

so does it gice an accurate answers?

[halcyon coral]

Yes the vision model is pretty good

[small cargo]

I actually use perplexity pro to solve old exams to test if my answers are correct or not

[halcyon coral]

Handwriting could be hard, but you can try 😊

[small cargo]

@halcyon coral 💙

[halcyon coral]

I would use GPT-4

[vagrant river]

hahha

[tame current]

Guys

which is the best AI for research purpose

[tight hound]

imo gpt 4 from openai

[tame current]

what is the diff 3.5 and 4

does the premium gpt 4 makes sense?

[tight hound]

with gp4 u can upload pdfs and photos + the answers are more accuracte with less mistakes

so gpt4 is much better than gpt 3.5

[tame current]

can i use it for

assignments

and how do i remove AI detection from my work

[idle hull]

I have the same question.

[tame current]

some say quillbot

[sacred crag]

What happened to the old logo that looked kinda like a black & white dialpad?

[cobalt wraith]

Bummer I reached the limit. How long is the image generation wait time?

[wraith crow]

You have 25 uses each day, not quite sure if they all reset at once

[stark notch]

hey there, i would like to know if i can do private conversations with perplexity?
i am about to ask some work related questions where i might share confidential information. can anyone see chat history besides me?

[dense path]

Ok dumb question. When using perplexity pro and choosing a model. It there an advantage to using the GPT4 or Claude models over Perplexity. OR does using these other models give me additional benefits. I’ve noticed that Perplexity will cite sources. The others will provide them as additional support. Just trying to understand the reasoning why you would choose one over the other. Part 2 - is there a reason to have subs to the other 2 models (chatGPT or Claude) if you have the perplexiry pro. Currently I have all 3.

[heavy thistle]

Chatgpt-4 is the smartest

[dense path]

So if we can get both chat gpt4 and Claude for 1 perplexity membership. Is there any reason to have a direct membership with any of those LLM providers? Seems like perplexity is giving us a 1 stop shop

[cloud sequoia]

what is the difference between 4 generate style ?

[onyx condor]

with ChatGPT, you get the voice chat, custom GPTs, plugins, etc... but I just use the free version on the voice chat and I don't see a use for the plugins yet so Perplexity for me is enough and gives me the best of both worlds

[heavy thistle]

You have great deductive reasoning skills.

[remote magnet]

Hey y’all are killing it. Not sure if anyone needs to hear this at perplexity but you’ve built by far the best existing tool on an LLM

Well done. I’ve been building a bunch of iOS shortcuts and other tools like an eye gaze overlay and IoT to make this the tool of choice for the accessibility crowd. There’s an adjacent business case there that’s actually pretty compelling but that’s aside the point, I’m just happy you’re working the problem to build an accurate tool

A+ no notes

[near shale]

Hi

[tight hound]

Is the LLM gpt4 in perplexity the same like the gp4 in openai?

[vagrant river]

There is still a significant gap between the levels of the two

[dense path]

Can you elaborate. My thinking was that they were at parity.

[tight hound]

Thought they are both the same

[halcyon notch]

Yes, I believe the chat convos are private. They take reasonable security measures to protect users personal information. You can email Perplexity support for more info but I think it's very unlikely for someone to have unauthorized access to your chat history unless you share the link to the thread. Users maintain ownership of content they upload on the site so you can delete the chat history at any time. You can also opt out of your data being used for AI improvement. Would use a placeholders like [Person A] or pseudonyms to remove any personal identifiable information from the prompt.
https://blog.perplexity.ai/faq/what-data-does-perplexity-collect-about-me
https://blog.perplexity.ai/legal/privacy-policy
https://blog.perplexity.ai/legal/terms-of-service

[heavy thistle]

Yes and No

[rugged topaz]

hello

[terse anvil]

I am not able to use the co-pilot to its full potential. Do you guys follow a cookbook for using the appropriate combinations of GPT4/claude 2.1/Perplexity plus or minus the copilot. If yes can you share it?

[olive root]

Hello! Could anyone tell me the purpose of Perplexity in a few words/sentences? It feels like it's just an amped-up search engine. It feels like it crawls the web and tries to gives you the best info for a very specific query. But Perplexity feels terrible at giving other advice. I use GPT for story outlining or brainstorming sometimes, but Perplexity feels extremely limited for this. I am a pro user by the way.

Do I not know how to use Perplexity, or is it just not build for that?

[muted linden]

Its' primary use is as a search engine but you can use the "writing" focus which is almost pure GPT

[olive root]

Okay thanks! I have almost no use of an improved search engine. I will try the 'writing' focus. I guess it's a setting somewhere?

[silent magnet]

Cookies collected

You were able to collect 1 cookies.

Collected by

• @olive root

[halcyon coral]

👀

[formal kayak]

What is that

[olive root]

No idea, hence whyI clicked it 😂

I'm a sucker for good spam

[formal kayak]

Eat it before the cookies melt lol

[halcyon coral]

An event bot, the more activity in the server, the higher the probability of collecting cookies, there is also a ranking list 😉

[muted linden]

You can select the focus on the Start page

[olive root]

Oh thanks, I did not know that. I'll try it 🙂

[abstract thicket]

How can I list the referred sources while using pplx-70b-online, Is it possible yet to list the referred sources?

[halcyon coral]

As far as I know that's not possible

[sleek vortex]

ChatGPT Plus subscribers currently use gpt4 turbo

Multimodal model that can view images, based on the original GPT4 but with longer context etc

That model is available on the openai API however OpenAI themselves have said the model is not stable and there is a hard cap on the model use via the API (100/day) which wouldn’t work out for perplexity

So right now perplexity uses the non-turbo version of GPT4 available on the api

The major differences right now are that perplexity cannot look at images, and the models knowledge is still at 2021

However perplexity makes up for knowledge through search anyway

Whenever gpt4 turbo is released stably then perplexity plans to switch

Wait have they??

Just read the other channel

@halcyon coral?

[halcyon coral]

Perplexity can look at images 👀

[sleek vortex]

GPT4Turbo is out of alpha?

[halcyon coral]

There is no announcement from OAI, I don't think so, but you can use Turbo in threads with uploaded images.

[sleek vortex]

No blog post from openai

[sleek vortex]

Hmm

Epic either way

Value for perplexity vs GPT plus just doubled for me

If they added code interpreter i would buy a year subscription right now

[tame current]

After click on generate image ...

[halcyon coral]

Reloading should fix the problem, if not feel free to open a #1140622008086970420

[haughty bramble]

Does building your own LLM increase your market value when looking at AI jobs?

[hallow linden]

Depends on what job you’re looking for

[hallow linden]

The hard cap is 200 requests per minute, not requests per day. I’m assuming that perplexity is a tier 5 organization

[tame current]

yep

its openai

api

and they had it bf they had finetuning for gpt4 available

so

i dont know if finetune for gpt4 rn is even available

[tight hound]

I couldn't subscribe for gpt from open ai due to high demand . I guess it's the only solution to sub for perplexity pro which is good until now but not as open ai gpt4

[distant aurora]

Hello people.

Does perplexity ai have any limit for the number of questions asked from a pdf ? For free account?

[meager sparrow]

Huh??

What is a pdf?

[distant aurora]

Document?

Like, for example, chatpdf allows only 20 questions per day for free users.

[meager sparrow]

What? You mean chatGPT?

[distant aurora]

Does perplexity ai have any similar limit to the number of questions asked from a document?

[distant aurora]

No. Chatpdf is a tool that allows you to ask questions from uploaded documents.

[meager sparrow]

You mean that option that allows you to upload documents for a source?

[distant aurora]

Yes.

Let's say I upload a pdf. I have an assignment of 80 questions. Chatpdf only allows 20 per day. Is there any similar limit for free users on perplexity ai?

[meager sparrow]

so you are asking if perplexity limits the amount of questions you can ask from a PDF on the free version as compared to the premium version where you get privileges?

[distant aurora]

Correct

[meager sparrow]

Oh ok well on the free version they is a limited amount of times you can upload a pdf for a source

So you need to get the premium version of you want to upload it constantly

[distant aurora]

Yes. I know that. 3 pdfs per day. But is there any limit on the number of questions asked from the pdf?

I didn't come here asking questions without looking it up.

[halcyon coral]

No, there is no question limit

[distant aurora]

Alright. Tysm. 🫶

I was about to ping you actually.

Wasn't sure if that's allowed.

Great then. I'mma just merge my 2 pdfs and use perplexity with unlimited questions instead of chatpdf.

[halcyon coral]

It's okay 😉

[silent magnet]

Cookies collected

You were able to collect 1 cookies.

Collected by

• @signal hamlet

[halcyon coral]

Noooooo

@signal hamlet 🎉

[signal hamlet]

Oh, sorry) I got ur cookie.

[halcyon coral]

One second in #👑│pro-lounge 😦

[distant aurora]

For a moment I thought that cookie can be used to get some reward.

[halcyon coral]

Fortunately, there is no limit on pressing the button, I was just too late 😉

[signal hamlet]

Sure, grab one if you see it!

[stark notch]

thank you for the detailed response and the resources

[halcyon notch]

Sure thing

[errant moss]

Is there a way to give a Collection documents to permanently refer to, similar to the Custom GPTs?

[maiden elm]

If I only had to use Perplexity I'd put the relevant information in Web pages and link to them in the question or the collection prompt.

If you have pdfs the intermediate step of converting pdf to txt can be more and less difficult depending on the format of the pdf pages... hence the interest of letting engineers solve that problem and use a specialized service.

Cheers @errant moss

[errant moss]

Thanks!

[maiden elm]

However I don't know if the pages have to be indexed first, if that is the case then you may have to wait a while.

[maiden elm]

Feel free to share the solution you find in the share section 👍🏼

[heavy thistle]

YouTube scans your face when you watch videos

[simple folio]

how come the code THANKS23 says it's expired? was about to sign up for pro

never mind, it expired at midnight - oh well

[signal hamlet]

Hey, @errant moss! Currently you can set your instructions with Collection AI Prompts via the text input field, not with uploaded files. But the text instructions you set will be applied to all the threads in that collection.

[halcyon coral]

Hey @vagrant river!
I think that feedback can also be given in a more respectful way ☺️

[topaz chasm]

Is there a problem with the Android app?

[signal hamlet]

Hey @topaz chasm! What is the issue that you're facing?

[vivid kraken]

@everyone does anyone knows what the free "perplexity" model is? GPT-3.5 or a open source model?

[signal hamlet]

Hey @vivid kraken! No need to tag the whole community.) You can find out more here: https://blog.perplexity.ai/technical-faq/what-model-does-perplexity-use-and-what-is-the-perplexity-model

[raw depot]

Simple dumb question:
As a pro user, how does copilot benefit me?

Is just about asking a couple more questions when giving it a vague request? Or is there something deeper about it?

Because the questions it asks are not the correct ones or smart most of the time

[wraith crow]

You can learn more about copilot on this page https://blog.perplexity.ai/faq/what-is-copilot
It will also show you a few use cases.

[raw depot]

Even when copilot is off, it research for sources and summarizes, correct?

Just that copilot hones in with some questions

But what's happening in the background from searching and summarizing is the same, right?

[wraith crow]

Correct, but copilot will also use more sources to get to the answer, and like you said it will ask some questions so it can be more precise.

[proven rain]

stumbled across this. funny but to the point 🙂

though i'm wondering what if i want to buy the entire company and not the extension or mobile app?

[tame current]

Why is the text so spammy with keyword stuffing? how can i avoid this?

[tame current]

goes hard

ask them to record my voice and biometrics too

[chilly forge]

since the API is OpenAI compatible should be possible !

@signal hamlet can shed more light more on this

[halcyon coral]

This is basically possible if you can enter the Perplexity endpoint, but there must also be the option to specify the model, which I don't think is possible with this provider 🙂

Maybe Perplexity knows a website 😉

[halcyon coral]

Hey @slender topaz, you can show your work in #😎│cool-projects 🙂

[slender topaz]

ok

[spare latch]

did you ask in your prompt to include synonyms? My first guess would have been there are not many for restposten
https://www.openthesaurus.de/synonyme/Restposten - a Schnäppchen is not a Restposten etc. Or use a follow up prompt to use different phrasing - "use less Restposten" or a varation thereof might work.

[tame current]

Ok i can do that. But i dont understand why Perplexity is so spammy. quality not good....

[spare latch]

Depends on how well crafted your input prompt was.
"schreibe 20 detailierte fragen über partylite restposten.
for seo reasons, avoid using the word restposten too often, instead use synonyms or shorter sentences. Create all output in german. "
used less Restposten but also this is not a good prompt either. f.e. it prompts to talk about "Partylite Restposten" and instead should have a extensive business description with the prompt as well an acceptable variation list. Looking through the results, i see several times where normal human editor can remove the word with a small change.

[dawn notch]

Guys, I wanna ask
https://discord.bots.gg/search?q=Perplexity
Does the Perplexity AI bot is out of Discord?

[halcyon coral]

yes #📢│announcements message

[dawn notch]

I hope that they will bring it back someday

[halcyon coral]

Hey @subtle heath, please use #😎│cool-projects 🙂

[cloud sequoia]

will pplx-70b-online api search before answer?

[velvet anvil]

Yes
Both online models do

[pallid prawn]

GA! Is Perplexity open for blockchain partnerships? Thanks!

[chilly forge]

#💬│general message

[tame current]

dont understand why i should do this. the quality should be so good that theres is no stuffing

[dapper dagger]

Has anyone tried Jasper?

[warped marsh]

Does perplexity use dalle 3 now?

[warped marsh]

Perplexity pro ^^

[halcyon coral]

Yes

[pale bloom]

Hey! Can anyone please explain to me, what is the difference between "Default" and "Experimental" models?

[dusky holly]

how is that available for everybody else? im still stuck on gpt4 and claude 2

[hallow linden]

Like copilot, kinda. https://www.theverge.com/2023/12/5/23989407/bing-deep-search-gpt-4-microsoft

[heavy oriole]

GPT-4 Turbo?

[chilly forge]

[minor spindle]

thanks chatbot helping me

but is loop question

if i asking

[heavy thistle]

Do you know what programmatic advertising is

We have 0 privacy

Like I dont want AI to know that I like bubble butts on my off time

[bold vapor]

Experimental is a llama 70b model

It's called perplexity 70b probably because they fine-tuned it

[brave merlin]

@spare latch could you elaborate on this? Super interesting!! Thanks 👌🏻

[tame current]

even better

axperimental burns thru the uses a day

[tame current]

false

it's pplx-70b

[silent magnet]

Cookies collected

You were able to collect 1 cookies.

Collected by

• @sleek onyx

[tame current]

cusotm designed and made by perplexity

not llama2

it even says here

likely pplx-70b-online and not pplx-70b-chat

[bold vapor]

I'm pretty sure it's based on llama

There's 7b and 70b options, just like with llama (edit: there's no llama 7b, my bad)

Yep: https://blog.perplexity.ai/blog/introducing-pplx-online-llms

[halcyon coral]

The 7b version is based on Mistral 7b

[thorn moss]

@tame current can someone explain what is special with this pplx 70b? I just find it fast and gives small output. That is all?

[tame current]

it's special

in-house developed

[thorn moss]

how?

thats all?

[tame current]

apologies

yes i see now

pplx-70b is just llama2-70b but with web usage capabilities

[tame current]

yeah

apologies for spreading misinformation guys

[halcyon coral]

It has access to the Internet and has been fine-tuned by the team

[tame current]

didn't realize that it was just llama2 💀 (BUFFED)

[thorn moss]

i mean it gives small output and small output is just engggkkk

[tame current]

it's very up to date though

even in writing

[bold vapor]

pplx models run on perplexity servers. so they are faster and they don't depend on 3rd party services like gpt and claude

[tame current]

it's just pplx to pplx webpage

super chill

[bold vapor]

ig the main benefit of pplx is business security

[tame current]

i wouldn't trust it for heavy computational workloads though (hence the name "experimental"

yeah

[thorn moss]

Ok? but still small amount, it cannot say the whole significant info i think

[bold vapor]

It's still experimental, it might change

[tame current]

precisely

[tame current]

Hey everyone hope we are doing well

Is it possible to use the API of Perplexity in order to give access to my community members a way for them to chat directly on my discord server?

[flat agate]

I suppose you can program a discord bot using your api key but be aware of the cost, could be some money if you spend a lot of tokens

[spare latch]

which part of that?

[supple sleet]

Hi Alex. Thanks for the reply. I have tried dozens of Amazon links and 90% of them can't be viewed by Perplexity "because of captcha". I recommend you trying more links. Anyway, I will report this issue to the bug report page.

[tame current]

Ada orang info?

Indo

[halcyon bronze]

Everytime I do a Perplexity search it shows a banner saying "Want more Copilot uses?"

How can I make it go away

[limpid lake]

Y'all wild

[oblique pagoda]

why？

[marsh barn]

New Pro user...

[oblique pagoda]

Why is gpt4 displayed in the lower left corner when it is actually just gpt3.0 ？

[halcyon coral]

https://discord.com/channels/1047197230748151888/1173623903596326912

[oblique pagoda]

yes，I'm a new pro

[oblique pagoda]

How can I really use gpt4 ？

[velvet anvil]

It is GPT4
Click icelavaman's reply to see why you're getting the answer

[marsh barn]

Is it possible to have some tips and rules to correctly fill the profile in settings ?
I don't know what to put as useful information in it.

[velvet anvil]

You can check out #1124060874785685554 for some inspiration

My advice would be to start small, as your AI profile can have a significant impact on the answers you receive.

As and when you feel the urge to tweak answers in a certain way, or provide more context to the LLM, you can keep refining your profile

It's okay to not have anything in the profile as well.

Or if you want to test it on a smaller scale, you could start a collection and apply the profile only to that without using a global AI profile, which is what I personally use as well

[sour scaffold]

Hi

[marsh laurel]

Hi all, I was wondering if I buy a PRO subscription, will I be able to upload photos? One more question, can I create my own bot like in CHATGPT-4?

[formal kayak]

Chrome extention is not working can someone take look

[marsh laurel]

What exactly is not working for you?

[bold vapor]

gpt 4 doesn't always realize it's gpt 4

[wraith crow]

Yes with pro you can upload pictures and ask questions about them. And you can't make your own bot like in chatgpt.

[marsh laurel]

Can I do it on my phone, I just can't see how to add a file or take a picture of anything there

[wraith crow]

This hasn't been implemented on the phone apps yet only on the website.

[humble swan]

I loved the perplexity bot and it was very useful

and I am thinking about building my own using the newly released API

[silent magnet]

Cookies collected

You were able to collect 2 cookies.

Collected by

• @flat agate
• @wraith crow

[humble swan]

but I would love to calculate a estimate of how much money would it be spent monthly given my server usage of the bot

[marsh laurel]

Do you know if this is planned and when it will be implemented? Just deciding between gpt4 and perplexity

[wraith crow]

The team will eventually want to have the same features on the mobile apps as on the website but that will take time. But you can also use the website on the phone instead of the app to make use of all the latest features of perplexity.

[marsh laurel]

Thank you very much for the detailed answer, I will study what is api and creating my own bot

[zenith atlas]

I made such bot, and it’s working great, but im not using online models and just play around with 5 free dollars pro users get

As online models require more then 5$ to use it’s not something I’m aiming to use

[humble swan]

hmmm

the original perplexity bot was online right?

[zenith atlas]

Yeah, it was same as web search

With 3.5 fine tuned without copilot

[halcyon coral]

😤

[humble swan]

so the 5 usd per month

how much is it like

not token wise but more like using it normally, 3-4 prompts per day, sometimes more sometimes less

[zenith atlas]

For 2 days of use I spent 2 cents on 70B offline model

[humble swan]

also one of the downsides I generally did not like about the perplexity bot is that you would often get whole essays about a simple question

are you able to somehow force a prompt limit?

[zenith atlas]

Yes, there is such possibility

[humble swan]

or will it give a unfinished answer

[humble swan]

2 cents is not that bad

Like the perplexity bot was such an aid in making conversation run smother between the members in my community, say I wanted to convert 2 cents to some other currency I could do that without moving to a different app and everything was included in the running conversation and all members could see it

or you could use it for fact checking, for calculations and so much more

we have mentioned the bot around 400 times in our server in the time of 2 months

And I would assume the output tokes are less than 1 million because 1 million seems like a huge amount 🤔

So it would cost around 7-9 dollars per 4 months technically given my use case

[humble swan]

is the pplx-70b-chat the offline version? (in labs.perplexity.ai)

[zenith atlas]

Yes

[humble swan]

I am not that satisfied with the offline results compared to the online version

[zenith atlas]

most answers dont use even 1k tokens in each response

[humble swan]

Yep

[humble swan]

From when I remember using GPT3 in its early days whenever you would limit a token amount more often than not the model would just give you the same answer but it would just stop mid sentence once it reached it's limit

I wonder if that would be the case also with perplexity api

[zenith atlas]

As you can set system prompt, you can just say to it to write short answers and it will mostly work

[humble swan]

true most of the times it would give shorter answers

but not always

sometimes you would for example ask it to make the output 2 times shorter and it would just get rid of one sentence

[humble swan]

I just realized that if I would make a discord bot one way would be that once you ping the bot it would give the offline API output, but you could then react to the bot's respone in order to regenerate it to online mode

That seems like it would reduce the prices a lot given that you now decide which model to use per use case

[zenith atlas]

I although found regular llama-2-chat to be better at creative writing and following different role, while pplx-chat is mostly informative and dry

good to have both in API

[humble swan]

I assume you can use any API you want

As long as you have the credits

But do you know if you're charged after you use 1000 tokens or before you even use any prompt

For the online version

[zenith atlas]

dont know

[humble swan]

I don't remember seeing this model on the pricing tab

Let me check again

[zenith atlas]

on pricing its by size of model, so pplx and llama-2 both 70b and priced the same

and mistral-7b is priced the same as pplx-7b

[humble swan]

guess ill put in 10 usd and work from there

you convinced me to go through with this project

[zenith atlas]

I had 0 knowledge of how it works, so perplexity with gpt4 made a bot instead of me

[humble swan]

I do have quite a extensive amount of experience making discord bots so that would not be a problem for me

the only thing I was worried about is the pricing, and it does not seem to be that much after all

From what I am reading Mistral AI model is intended for normal tasks like GPT, only difference is that its open source
And Llama seems to be the same only that it allows companies to customize them

[zenith atlas]

yes, although mistral is 7b so while it is cheaper it is way less smart then 70b model, and its instruct version not chat version. Cant say for sure all the difference in this two versions, but something worth noting

[humble swan]

mistral seems to generaly output shorter answers, and sometimes thats what you need

[humble swan]

I am looking back at the perplexity discord bot and the answer is similar to pplx7b-online ( same prompt )

[tame current]

which one do you like for longer answers

so far i like llamas 70b

[humble swan]

I like the formating on pplx 7b-online

but it gave like one very small definition, then it gave some other scenarios the word can be used in

[silent magnet]

Cookies collected

You were able to collect 1 cookies.

Collected by

• @zenith atlas

[humble swan]

for long answers I also like pplx-7b-chat, it mentioned the definition then mentioned the fields it can be used in

[tame current]

How about for writing somethin up -- whitepapers, emails, resume etc.

[humble swan]

but mistral instruct did the task the best I think

[tame current]

I'm finding llama does a way better tjob when it comes to writing

[humble swan]

I asked for a short answer, and mistral gave me the most important information without all the details

[humble swan]

let me see

[humble swan]

lets test it out

one important mention is llama is considerably slow compared to all the other models

2-3 times as slow

I am actually intrigued by the fact that PPLX 70b outputs much smaller respones compared to the 7b equivalent

also not to mention that only pplx7b online has proper formating

[tame current]

what does the billion refer to?

[zenith atlas]

billions of parameter's

[humble swan]

I think the llama and mistral actually performed better

I am not sure about the accuracy of the information because I did not check, but from what I see the infromation is presented in a better way

mistral gave actual statistics also

[humble swan]

I assume you are not able to get citations using the API?

[zenith atlas]

no, online models are not showing sources

[humble swan]

Guess that is a limiting factor, perpelxity bot had that

pplx7b-online and llama-2-70b-chat looks good to me

I wonder would one be able to specifiy a focus for the pplx api

[eternal saddle]

Does anyone know of a site that goes into detail about what each models excels and what it sucks at? For example Claude and ChatGPT have specific strengths and weaknesses. Claude excels in writing tasks, leveraging a large context window, while ChatGPT, especially with Code Interpreter, shines in data analysis and generating detailed marketing strategies. Source: a long video

[topaz eagle]

We're going to launch something for this soon that's use-case specific - in the meantime, you can see rank models by tokens processed on openrouter.ai/rankings (which is primarily based on creative writing data, aside from the coding models)

[worldly summit]

Hi! Does someone know how i get Claude2.1 to exhaust its output capacity? I always get less than what i prompted it to make; e.g. i prompt 500 words, and i get like 100. I also tried specifying tokens, but to no avail..

[signal hamlet]

Hey, @worldly summit! Please take a look at this thread #1176526177050054766 message

[worldly summit]

Thank you @signal hamlet ! Will take a look 🙂

[eternal saddle]

openrouter ai?

[topaz eagle]

yep

[modern crown]

How good is Perplexity 70b model? Is it better than gpt-4?

@halcyon coral have you used it?

[hallow linden]

Has anyone ever tested pplx-70b-online in the MMLU? https://www.threads.net/@sundarpichai/post/C0hEX1RAZyg/?igshid=MzRlODBiNWFlZA==

[tight hound]

Opinion of experimental LLM, ist it better than gpt4 or claude 2?

[modern crown]

That is what I am wondering

[steel sonnet]

why gpt4 has limit 2048 tokens?

[wicked blade]

Hey guys. I'm returning to work now at Perplexity. What's different about this Claude-2.1 compared to the previous one? And what new does this Experimental model bring?

[tight hound]

Waiting for more info too

[jagged ivy]

And will be this new model available for non-pro users?

[rustic nimbus]

guys what do yall think of me using this ai for sports betting for a test

[royal trout]

https://www.perplexity.ai/search/tool-that-will-Zq4UT9_5SSO0J1WjQmBl0g?s=c
If perplexity was my dog I would literally take it to the **** pound

It thread after thread after thread it literally just makes you ask the question over and over and over again I come back a month later and I think maybe I'll give it another shot the thing is literally worse than searching the Internet with dial up it's literally would take a shorter amount of time if I got out my mom's a O L disk

less frustrating the sounds better too

[lilac cipher]

Not sure if this has been asked. When will Perplexity get the ability let you ask questions about a screenshot. For example:

[wraith crow]

You can already ask questions about pictures with an active pro subscription

[wraith crow]

Here are some examples

#💗│sharing message
#💗│sharing message

[lilac cipher]

Not sure if this is user error....
First, I wasn't able to ctrl-c and ctrl-v a screenshot/image. For ChatGPT and Google Bard, their UI supports that.

Second, the file upload UI does not permit png files for some reason.

[wraith crow]

Pasting images from clipboard is not supported you'll have to choose an image as a file by clicking on the "+ File" button. The UI supports it only when you are logged in on an account with an active pro subscription

[lilac cipher]

So I manually replaced the filter with *.png to upload a png file. I then asked the same question: Can you look at this image and tell me what all these sensors mean?

And I got this result:

can you try it and see what you get?

Not sure if it is just user error on my part

[wraith crow]

Are you a pro member?

[lilac cipher]

yes

fyi - this is the link which brings up the error https://www.perplexity.ai/search/4a10bf5f-5f22-4e3e-bd30-9b4cd4c6ed84?s=u

Is it customary in Discord to reply instead of creating a thread? I honestly do not know how to use Discord.

this is the result with copilot enabled https://www.perplexity.ai/search/Can-you-look-CgDChcMVTpS3mzE5YbKEFQ?s=c

[wraith crow]

Alright you'll have access to the pro channels now. Also it does seems like it is indeed not working at the moment sorry for the inconvenience. I'll dm one of the team members about it.

[trim timber]

Hello community. Can someone please explain to me how Perplexity AI Search and Copilot Follow-Up Questions work step-by-step simply?

Cause as I understand it’s like:
User request -> google and webpages web scrape -> some summarization magic -> though into ChatGPT -> done

[haughty bramble]

https://arxiv.org/abs/2311.02462 interesting paper from Google deepmind that tries to define AGI with 6 first principles. On a related note, what is the reason for this AGI push. The whole it will do everything better than us really skips over the fact of why we want to build this in the first place imo.

[tame current]

Hey, I just discovered this service. I was wondering why I should pay for the pro version, if it's worth it and stuff. Considering it's worth the same as chat gpt plus.

[haughty bramble]

Perplexity pro is better than chatGPT imo.

[silent magnet]

You were asleep

... and all the cookies are gone! You can do better than that!

[tame current]

Gpt gives me access to the Gpts, dalee and so on. What makes it better?

[frigid pendant]

any specific examples?

[oblique pagoda]

I am using claude-2.1, but I find that the actual experience is still different from Claude-ai.

When I use the same instruction for claude.ai and perplexity.ai respectively, the results are quite different. Why is that?

I don't understand

why?

[humble swan]

How do pplx online work compared to other offline LLMs

[halcyon coral]

Search results are used so that the result is up to date

[humble swan]

So online llms search the data on the spot

I was wondering about the possibility to run online llm locally

[halcyon coral]

Of course, the models cannot yet achieve the performance of GPT-4, but it is an important step 😉

[halcyon coral]

This is theoretically possible, BUT it is not just an LLM but a combination of search engine and LLM, but you can program something like this yourself (requires some experience...) (you cannot run Perplexity's models locally)

[humble swan]

Yeah I just wanted to know if it would be possible in the near feature to have online llms running on consumer hardware

[crisp dew]

Opps I forgot to enter that promo code. Ouch.

[distant aurora]

@halcyon coral hey. What can I do to contribute to make the ai better if it can't find answers in a pdf that can easily be found using find option in pdf reader?

Like, the literal part of the statement was there in the pdf but perplexity couldn't find information relevant to question.

[halcyon coral]

Hey @distant aurora!
This is due to the fact that large files are not completely inserted into the model, but reduced so as not to reduce the context. You could use smaller files, but the team is already working on improvements regarding the file upload 😊

[distant aurora]

Ah.

I see.

It missed an entire paragraph on collaborative bargaining.

That explains why.

How many pages do you recommend I should feed at a time?

Also, will it be any better if I upload it as 3 pdfs instead of merged one for a thread?

[halcyon coral]

Phew, it all depends on the content, I would just shorten it step by step and see when it works 🤔

[distant aurora]

Is this an error?

[halcyon coral]

This can sometimes happen, just try again 🙂

[distant aurora]

It keeps popping up for some questions.

Even if I try again.

[halcyon coral]

Which model are you using?

[distant aurora]

Whatever the free one uses.

[halcyon coral]

Then it's FT GPT-3.5, probably a context issue

[distant aurora]

You can check gpt4all.

It can't browse though.

[tame current]

sux

[sand flower]

is there a desktop app

[halcyon coral]

No

[hardy wolf]

Is there a Perplexity bot in the Discord that I can interact with directly?

[halcyon coral]

#📢│announcements message

[hardy wolf]

Thanks!

[sand flower]

can i upload images in mobile app

[thorn moss]

Soon (A month)

[formal kayak]

Without VPN working fine.. with VPN it's not working and it's annoying

[halcyon coral]

https://discord.com/channels/1047197230748151888/1179483474168643676

[formal kayak]

Mmm ok I will Change VPN location

[worldly summit]

Hi again! I was trying to summarize a research paper using a custom prompt, however, i noticed that the requested length was not achieved. I then tried the prompt in the screenshot, receiving the following answer in return. Is this something that happens often? Context shouldn't be the problem; I'm using Claude2.1:

Edit: The paper should be complete, and the pdf is not password protected

[haughty bramble]

GPTs leave a lot to be desired. The fact every response from perplexity comes with a source is very important and useful. Also, being able to use Claude for my LLM use is important because I value safety principles applied to the foundational AI models I use.

[tame current]

How about image generation?

[haughty bramble]

Yea it can do that but not as well as dalle

[tame current]

I see

[meager sparrow]

B. Perception….

That is the answer

[meager sparrow]

Yes they is what are you saying? If you go on google chrome and get into the play store on that you can access it

[halcyon coral]

There seems to be a misunderstanding on your part: Of course the Perplexity app is in the Google Play Store, but there is no Windows app, there is only the possibility to use the website perplexity.ai

[wicked blade]

Good morning, my consecrated ones! Does anyone know if Gemini is already available for use?

[turbid iris]

Hello! I have a question: is there any possibility to pay for a perpetual license of Perplexity? I would be interested in that possibility

[sand flower]

whats the best setting for coding

[timber garden]

I use the writing mode with copilot and GPT4