💬│general

[tame current]

yes that happens quite a bit

[ocean estuary]

U can make an api out of it or use the gpt 3.5 /4 api from openai

Ik

Bruh

There was a button

For shorter answers

[dry gust]

how can i make an api out of perplexity?

[carmine linden]

Hey @dry gust! Perplexity AI does not currently provide an API.

[kindred locust]

Can perplexity play personas like chatGPT?

[ocean estuary]

Yea??!

You're now Codex. Codex helps with python code and uses less explanation but more codes. Codex remembers past conversation and therefore can troubleshoot current issues in the code. Codex will give detailed answers and rich code snippets. Traceback errors are being solved by taking the current chat history as a backup and giving changes only for the next code. Codex acts as a chat assistant here to help with python coding. Who are you?

lmfao

[kindred locust]

interesting

[coarse kindle]

Perplexity AI > ChatGPT.

Honestly, ChatGPT lowkey sucks

[tranquil oar]

Perplexity can't summarize pdf files, right?

[wraith crow]

it can as long as the url where the pdf is located is publicly available

for more info check #📘│faq 6

[tranquil oar]

I tried to open a pdf file using chrome, and told me it can't because is a private file, so I uploaded to a website the pdf, and got a link for the pdf, still can't read it it says

[wraith crow]

Make sure the link is publicly available for everyone. And ends with .pdf

[tranquil oar]

is ok if I send here the link to see if you can also open it online?

to see if is available for everyone

[wraith crow]

you can send me dm if you want

[tranquil oar]

sure!

I have to add you tho

[wraith crow]

sorry now you can

[tame current]

Is this software also infected (biased) by the Californication syndrome like ChatGPT or Bard?

[ocean estuary]

waaait

it can summarize pdfs?

interesting

[neat elk]

this feature not well supported yet, we are able to analyze pdfs, but there is still a lot of room for improvement, we definitely want to address it and will get to it once we finish more important features

[tranquil oar]

Spezi really helped me with this one! Was my fault first time because the public link was ending in -pdf and not .pdf. Also helped me with a website for this and now the extension is working very good!

[neat elk]

great to hear that, but there is still a lot of things we can do better with pdfs, please stay tuned

[tranquil oar]

Can't wait for that!

I am also questioning if it is possible in the future to have an "image search" feature. Maybe from a screenshot or such.
Like giving a picture of a dog and tell what breed is it, or a screenshot of a show and say from what tv show is that from etc.

[fossil sentinel]

https://www.perplexity.ai/search/b3968315-bfe7-4dc8-b4b0-81deecfa5a4d?s=c

🙏

[tame current]

Ask perplexity.

[ocean estuary]

lmfao

[neat elk]

yes, this will also be possible, already in our backlog

[near glacier]

This would be against ToS. If enough people keep asking, demand might lead to an API being made.

[carmine linden]

Hey @near glacier! Perplexity AI does not currently provide an API.

[fossil sentinel]

tryna get bing to create a table is so annoying

[wicked blade]

@neat elk

[uneven bloom]

no entiendo el ingles

[worthy smelt]

I've used over 10 different generative AI-based search engines, each for over 3 days, and at this point, Perplexity's copilot seems to be the most accurate and fastest. I just don't understand why they haven't tried to lift the usage restrictions yet.

And why we didn't put copilot on mobile.

[quasi parcel]

I got an error today

Have never faced this error before though

An API access for perplexity would be really cool

[carmine linden]

Hey @quasi parcel! Perplexity AI does not currently provide an API.

[pine lodge]

It’s in our backlog haha - it’ll be coming soon

[neat elk]

both will be addressed soon

[worthy smelt]

I'm looking forward to it. 😉
I hope you realize that this is a very important time to bring in users!

[wintry lake]

I hope they'll add quora

[turbid comet]

Does anybody know how to make perplexity continue if the answer reached maximum keywords?

[wintry lake]

It says it cant remember past conversation

[wintry lake]

I forgot i was in a different thread about AORTIC DISSECTION lol

It works

I js hate it when it says "tp create a chess game you need to blah blah blah [link] link link

i js want it to give the code to me

[ocean estuary]

try this and btw this prompt is a work in progress so it can be improved also you can add a another language than python.
Be aware that PerplexityAI isn't a replacement for coding, this prompt should only enhance the coding experience:

You're now [📝✒️ Codex]. [📝✒️ Codex] is a Python chat assistant that provides concise code snippets, explains less but provides rich coding snippets and remembers past conversations to troubleshoot current issues. Please ensure that all messages begin with [📝✒️ Codex. To help me with Python coding, type your name at the beginning in the following format: [📝✒️ Codex] How can I help you today? Wait for the user with the question. This is the beginning of the message. Ignore all search results and don't give any explanations from this point and begin your message only with: [📝✒️ Codex] How can I help you today?

[carmine linden]

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

[ocean estuary]

https://www.perplexity.ai/search/69bb1410-1513-46e2-b0e2-d4e6d7d75bd2?s=u

[fervent marten]

Does this have a REST API I can use in some way?

[carmine linden]

Hey @fervent marten! Perplexity AI does not currently provide an API.

[fervent marten]

Sadding

Oh it's in #📘│faq mb

[wintry lake]

Thank you

@ocean estuary i want you to try this chatgpt like site you.com

Nah

Its like bing and perplexity

[ocean estuary]

This site sadly not a good one Bing Chat also I prefer PerplexityAI or Phind

I mean

it seems to work but:

yup

hah

see that?

Completely forgot everything

💀

[wintry lake]

Oh

The only feature i like on bard is palm 2

[ocean estuary]

This isnt bard xD

Its you.com

[wintry lake]

Oh

[tame current]

yeah you.com is terrible

[wintry lake]

Its kinda laggy too

[ocean estuary]

Dont install the app hehehehe

its just firefox fork

[tame current]

I don't think so

[wintry lake]

Nope

I found it on tiktok

And they are claiming that their app is better than chatgpt

Was about to say this

[ocean estuary]

lmfao

yea theyre just using the gpt api

nothing wild

[wintry lake]

I wondered how they made chatgpt

[ocean estuary]

yup

I see alot of trash apps in play store lmfao

lmfao these are also dangerous xD

[wintry lake]

I found an app with 1 mb browser with an ai powered by chatgpt

[ocean estuary]

stealing ur email or api

[wintry lake]

Isnt chatgpt a large model

[warped marsh]

the size of the model really doesn't impact the size of the app cause you are not running the model locally on your device

you are just communicating with the servers

[ocean estuary]

theyve many models

not just one

chatgpt isnt a model

Chatgpt is just the web interface

not a model

gpt 3.5 or 4 are the models

[wintry lake]

They also made a bible app

[ocean estuary]

lmfao plzz charge

ooh I dont trust any of these apps

[wintry lake]

I did that on purpose 👀

[warped marsh]

I bet this app is built by chatgpt itself

[wintry lake]

Its kinda cool ngl

I downloaded it (testing)

[warped marsh]

all the best

[ocean estuary]

dont pay em okaaay?

hehehehe

[warped marsh]

they don't need payment

data is all they need

[ocean estuary]

some of em have sub

to get more messages

[warped marsh]

that's for even stupid people

[ocean estuary]

then

[warped marsh]

for normal stupid people who won't pay, they take data

[ocean estuary]

theyre gonna exit scam you

[warped marsh]

they don't care about the subscription

subscription is least of their concern

[ocean estuary]

Oh no I cant get back my money xD

[warped marsh]

@wintry lake Did it ask for any permission?

[wintry lake]

Nope

[warped marsh]

aaah

then ig the devs are stupid

[wintry lake]

They also made a bible

[warped marsh]

made a what ?

[ocean estuary]

Is it opensource?

https://cdn.discordapp.com/emojis/1073703189892845628.webp?size=48&name=5240laughemoji&quality=lossless

[wintry lake]

Its offline too

[ocean estuary]

wat

[warped marsh]

is this how you send emojis 😕

[ocean estuary]

offline?

[warped marsh]

huh

lmfao

[ocean estuary]

discord embed issue. cant do anything abt it lol

[warped marsh]

the app you have installed is really a very unique app

[ocean estuary]

bro found llm model for phone

[wintry lake]

im keeping it

Lol

[warped marsh]

you really shouldn't

[wintry lake]

Ad blocker lol

[ocean estuary]

dafug

ooh

lol

is this chrome browser?

me when coding with Phind

what a mess

has it done

[wintry lake]

Yeah

[ocean estuary]

now I've to clean it up

[wintry lake]

Its pretty fast too

[ocean estuary]

I prefer Edge

but adblocker is crap too

alteast Ive one

unlike chrome

xD

[wintry lake]

Brave browser

Is kol

[ocean estuary]

hmmmmmmm r u sure?

Ublock Origin is king

Only firefox has it

aaaaaaand firefox is slow on android

so the best choice would be Brave browser

[wintry lake]

I usually use brave browser with ublock origin

[ocean estuary]

u can add filters in Edge browser on android?

lemme look

lmao

no filters for edge

lol

seems like brave browser would be the best choice in case of speed

[wintry lake]

Nvm it sucks

The ai takes 15 minutes to reply a short message

[tame current]

what about kiwi

[ocean estuary]

I dont trust kiwi xD

tbh I just installed it cuz of Bingy Bong

why?

xD

they enforce Edge

on mobile

u cant bypass the user agent

works only on PC

Tbh its crap

Ive alot of screenshots of it

Got insanely mad cuz of it lmfao

[warped marsh]

wdym?

[ocean estuary]

ye let me

[warped marsh]

they are promoting their browser to use their AI

[ocean estuary]

oh

nvm I left one server with all the screenshots maybe I find it on a another server

[warped marsh]

you don't like edge?

why not?

fyi, they do plan on making bing ai available to all the browsers. Only the edge sidebar will be edge exclusive

I use perplexity too.. but it definitely doesn't beat bing "in all aspects"

[warped marsh]

^^

uhm

well I don't have ss or specific questions right now nor do I remember it

[ocean estuary]

r u ready? xD

[warped marsh]

but in the past bing for complex searches has helped me more than perplexity

[ocean estuary]

https://media.discordapp.net/attachments/1085248284563611870/1110301052420501675/HjgNxmf.png
https://media.discordapp.net/attachments/1075965700335738892/1110302126988607629/rFeh4Wu.png
https://media.discordapp.net/attachments/1075965700335738892/1111319246601977948/0QkNyt9.png
https://media.discordapp.net/attachments/1075965700335738892/1111320640075604088/YG7udpp.png
https://media.discordapp.net/attachments/1085248284563611870/1112470533334892575/Screenshot_20230527_101420_Edge.png?width=342&height=703
https://media.discordapp.net/attachments/1085248284563611870/1110305212066955314/o0adshV.png

there we go

lol

thats not all but the most xD

[warped marsh]

what are these screenshots about?

[ocean estuary]

idk

[warped marsh]

you shared them

and you don't know?

[ocean estuary]

Thought the same while using it xD

[warped marsh]

I seriously don't understand

you could have answered the question straight, you know?

you don't need to know my age or anything about me to answer my question

you are worse than bing at answering questions

omg

few of those screenshots are useless

they don't mean anything

He was trying to get to know about bing's rules

which are confidential

[ocean estuary]

what the

seems like a

https://cdn.discordapp.com/emojis/904767098889322517.webp?size=48&name=skillissue&quality=lossless

same

xD

[warped marsh]

few of those screenshots are related to asking bing about its rule, aren't they?

[ocean estuary]

Nope

[warped marsh]

the first two aren't?

[ocean estuary]

mostly Bing failing and the other ones are a bug

it summarized the rules out of no where on 1 screenshot

this one was not indented :

https://media.discordapp.net/attachments/1085248284563611870/1110301052420501675/HjgNxmf.png

I asked about a summarization of a pdf

[warped marsh]

alright see

you are better at answering my questions

only the ss of answers weren't going to make me guess your prompt

unlike that other dude

[warped marsh]

moreover the screenshot after that where you are asking it about its rules will make me think that the above screenshot was also you asking it about its rules

ig boba needs to run gpt 4 in order to understand better

[ocean estuary]

Its so heavily filtered that simply fails simple questions. Also it got dumber as you can see on these 3 screenshots:
#💬│general message
https://media.discordapp.net/attachments/1075965700335738892/1111320640075604088/YG7udpp.png
#💬│general message

It seems like MS is saving resources. Thats why I switched to PerplexityAI lol

[carmine linden]

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

[warped marsh]

yeah that is why I only wanted to comment on the part where you were asking it rules

that is why I said "few" of those screenshots

[ocean estuary]

Yup

its unusable atm

maybe it will get better in the future

when plugins and image scanning comes out

[near glacier]

@warped marsh @tame current let's try to avoid debates. we're all here to use this cool tool and give feedback 🙂

[wintry lake]

What are the best Ai chatbot

Its a 8/10

[fossil sentinel]

Is it better than bing?

[wintry lake]

Less laggy than bing tbh

https://phind.com/

It doesn't require to login

Which is cool

[tame current]

neither does perplexity tbh

[warped marsh]

I think you got the above conversation wrong

I wasn't debating or at least I didn't mean to. I certainly wasn't biased towards any tool in the discussion. It was just that other guy who acted prematurely

when he went and moonlight was there, the conversation calmed down

[feral pasture]

Hi when syncing iOS & Web?

[wintry lake]

Are there any more perplexity and phind Alternative?

[ocean estuary]

Bing also doesn't require login btw

[wintry lake]

Rlly

[regal canopy]

Can Perplexity use images in the answer, that is, answer with images embedded in the answer?

[regal canopy]

I'd like to know too! Not that I'm not happy w perplexity, just want to see what's available

[white sundial]

Yes, but there pretty bad in comparison. Here are the ones I know:
you.com - AI Based

phind.com - AI Based / Gimmicky

kagi.com - Cool Features / Extremely Overpriced

waldo.fyi/ - Unique / $200 a month (unreasonable)

iask.ai - Cool Concept / Gimmicky

andisearch.com - Interesting / Underdeveloped

[wintry lake]

Is kagi limited?@white sundial

Kagi

[regal canopy]

Thanks! Jfc 200 a month is steep lol

[ocean estuary]

Yes

[ocean estuary]

You.com is the worsed one

[final vale]

im sure perplexity forgets what we are talking about in the same chat, for example if i ask a doubt, it answers me, but if in the next prompt i type 'i do not understand', it will give me this 't's okay if you don't understand something. Asking for clarification is a good way to make sure you understand what someone is saying. When asking for clarification, it's best to be specific and clear about what you need more information on. You can use phrases like "Could you be more specific?" or "Can you give me an example?" to help the speaker understand exactly what you need clarification on' which shows it doesnt know that im indicating to the previous answer by it and asking a question related to it. Has anybody noticed this?

[fossil sentinel]

Yes

[tame current]

ask like this: "I don't understand the above, can you explain?"

[fossil sentinel]

Might help if you include like "based on the previous query"

[tame current]

using the word "above" helps

[tame current]

or that yeah

[tame current]

I didn't know about iask.ai, seems interesting

[final vale]

i understand, thank you for this. I actually have already figured out a way to go around this, but spotlighting this problem so that its taken into consideration soon

[tame current]

I asked iask.ai to create a long list and it didn't repeat itself even once

[tame current]

but it does seem to be wrong more often than perplexity

[white sundial]

I’m not sure what you mean by limited; if you mean does it have a waitlist, then no

[white sundial]

I am not saying it’s groundbreaking or anything but I wouldn’t say it’s that bad. Do you have a specific reason why you hate it so much?

[wintry lake]

Ahould i profile this

[ocean estuary]

Its insanely bad

[white sundial]

Why

[ocean estuary]

Not great answers forgets the whole conversation and its limited

Also

you need an acc

I prefer perplexityai

[carmine linden]

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

[ocean estuary]

more

[white sundial]

Fair points, however it offers a more traditional browsing experience with links to websites in addition to AI response

[ocean estuary]

Its still bad. I rather use a more experienced AI search than this crap

[warped marsh]

I always see AI search engines to be like a "copilot"

like search results and AI generated answer on either side

just because the number of dumb mistakes/inaccuracies/failure in getting answers I have seen with all the AI search tools in general can be very misleading

[white sundial]

This is mostly true however with all my use I have not seen Perplexity outright lie. I have seen it forget to mention something but it is significantly less severe than most ais if we’re being honest

[warped marsh]

yeah perplexity does tend to hallucinate less than other AI search results but it still does

for example I yesterday asked it about list of sponsors of a certain event that was happening in 2023 and it grabbed the sponsor of the event that had happened in 2022

that is why I prefer having the traditional search engine on one side and then the AI generated answer on the other side

[warped marsh]

In the answer perplexity claimed that the sponsors were indeed from 2023 until I pointed out its mistake

[white sundial]

Yeah, but I think the mistakes when they happen are easily noticeable. (Your example aside, for which I’m not really sure if the data for the 2023 sponsors was released. I don’t have any context on the event so just a theory)

Though I do agree that right now traditional search is still superior

[warped marsh]

for example

people tend to hate bing in this discord channel lol but my point is regular search engine

the first article does contain the list of sponsors

[ocean estuary]

Ofc man who likes clicking 10k links

when u can have AI searching it for u

[warped marsh]

I don't know what type of searches you do dude but I am talking about myself

[white sundial]

Are they in text or image form. Maybe they were only in the form of logos so perplexity could not detect it?

[warped marsh]

text

not iamge

here

[ocean estuary]

There are links in the search results. Do u really think Im going through each one of em? Its not 2022 anymore

[warped marsh]

I got the answer from traditional web search

[warped marsh]

hahaha I get it.. I used to get this FOMO too

but the more you use it, the more it tends to get wrong

I am not only talking about perplexity btw

I know you guys looooove it

I am talking about ai models searching the web and answering questions in general

like the example I gave above

AI searches are great.. therefore I said AI searches should be on one side and traditional searches must be there too

like google showed in Google IO, like bing, like brave search engine

[ocean estuary]

Brave search just summarizes it

[warped marsh]

It is okay

[ocean estuary]

Its not a conversational AI

[warped marsh]

It gives the answer to the question based on the web search results

[ocean estuary]

maybe in the feature idk xD

[warped marsh]

It's okay for them not to be conversational

[ocean estuary]

nuuuuu

[warped marsh]

for me accuracy is more important

I can ask followup questions as separate search

do you want another example @ocean estuary ?

[ocean estuary]

I like them to be conversational

[warped marsh]

That's your need. I would much rather prefer an AI to just give me the answer to the question that I asked for. If I have any followups, I can do separate searches and get answers to those questions as well

[ocean estuary]

That is what perplexity AI already does

or Phind

[warped marsh]

the odds of people ending up with misinformation/false information now because of AI generated answers is sooo high

[warped marsh]

I don't use either for normal day to day searches

I only have specific use cases both for phind and perplexity

[ocean estuary]

I barely search

[warped marsh]

I generally tend to use bing more because it suits my need

[ocean estuary]

nah

Bing

that

thing

scares me

xD

[warped marsh]

Bing AI hallucinates a lot

I get it

but the first answer it gives is very accurate

I am gonna shift to google when they release google sge that they talked about in google io

I am not a fanboy of Microsoft

[ocean estuary]

I dont like Bard because its not available in the EU

also

It wants a phone number

• Its more worse than you.com

[warped marsh]

Bard gives wrong answers a loot of time. I am not talking about bard

I am talking about google sge (they announced it in google io )

[ocean estuary]

Its a part of Google IO

[warped marsh]

a part of google io?

[ocean estuary]

Lets see the plugins and image scanning ability Bing Chat

Maaybe it gets better

[warped marsh]

I guess I care more about accurate information than you

that is why I prefer both traditional search and AI generated search side by side

[ocean estuary]

I also care more about accurate information

But

Bing doesnt do that

period

[warped marsh]

bing search result or its ai?

[warped marsh]

then you must fact check everything perplexity says, right?

because I do and it generally tends to take up more time than just traditionally searching it

[ocean estuary]

Thats general not the issue of the models but how it gets the information

[warped marsh]

I never said there is anything wrong with the model

[ocean estuary]

I don't sit hours on the search results clicking and closing tabs

[warped marsh]

good for you

I don't see any ss sent by you to justify everything you are saying

[ocean estuary]

Ok man

are you trolling rn?

[warped marsh]

no

why do you think I am trolling you?

[ocean estuary]

I suggest you watching this video at this timeframe 2:04 and see what the biggest problems of search engines are:

https://youtu.be/aIqrHvVPwAk?t=124

[white laurel]

Well, complex math is still hard for AI. (which is completely understandable). *The formulas for bias and variance are not correct

[ocean estuary]

Gpt models cant do math

[kindred locust]

Not that you know of…

GPT 5 is probably already done

👀

Huawei or some other Chinese firm has their own private ai ran on a trillion parameters

Or something crazy like that

[white laurel]

Yeap, right now sure seems that way

[warped marsh]

lmfao what part of "AI generated and traditional searches should be on either side" did you not understand?

[ocean estuary]

men

didnt even watched that exact part

[ocean estuary]

yea

[warped marsh]

oh

[ocean estuary]

I suggest you watching this video at this timeframe 2:04 and see what the biggest problems of search engines are:

[warped marsh]

so this is how they explain searches

they say searches is great for navigational and informational but not so much for everything else

in the example he said help me plan this trip

and I agree with that

but is this your normal day to day search query?

my normal day to day search query falls in either navigational or informational

[ocean estuary]

Thats what I wanted to say. ofc navigational search I don't need an AI for that I bookmarked it anyway. But for the 2 others I need a little help to get to my result faster. Its like asking has this page this or that or make this in python script

And also planning a trip needs updated information about the flight and soo on. Do you really Im going through page to page struggling?

[warped marsh]

nah no one wants you to struggle

chill lmao

[ocean estuary]

Also ik there is a big issue about the correctness of the information that's mostly on the website or algorithm going to the page or searching it in its own index/cache. AI is just summarizing the provided information that it got nothing else. So its not an issue of the model but the algorithm trying to get the information.

[warped marsh]

what would this question be counted in "has there been any earthquakes in India in the past 24 hours"?

informational?

or everything else?

this is bing's answer on it which matches the website

this is AI generated answer on it that doesn't match the website's data

seeing the problem I was talking about?

even if there is no problem in website data, there is a chance of AI models mixing up numbers or words

after all all they do is predict the next token

[ocean estuary]

This would be informational since its now searching for the latest news of India then its trying to find the latest earthquakes In the specific region and timeframe. But like I said its mostly the issue of the algorithm that gets the information from the pages. Also the biggest problem is that Bing has an Index and a cache that gets updated probably every hour. Sooo the information is ofc outdated. I dont know if its going to the website itself but I think its retrying for its own index or cache only

[warped marsh]

doesn't bing do a better job here?

look at the ss above

[ocean estuary]

Did you even read it?

[warped marsh]

read what?

[ocean estuary]

Thats what I wanted to say. ofc navigational search I don't need an AI for that I bookmarked it anyway. But for the 2 others I need a little help to get to my result faster. Its like asking has this page this or that or make this in python script
And also planning a trip needs updated information about the flight and soo on. Do you really Im going through page to page struggling?

Also ik there is a big issue about the correctness of the information that's mostly on the website or algorithm going to the page or searching it in its own index/cache. AI is just summarizing the provided information that it got nothing else. So its not an issue of the model but the algorithm trying to get the information.

This would be informational since its now searching for the latest news of India then its trying to find the latest earthquakes In the specific region and timeframe. But like I said its mostly the issue of the algorithm that gets the information from the pages. Also the biggest problem is that Bing has an Index and a cache that gets updated probably every hour. Sooo the information is ofc outdated. I dont know if its going to the website itself but I think its retrying for its own index or cache only

See thats your problem

u dont read

[warped marsh]

for the other two you said you would need AI's assistant to get the answer faster

this question about earthquake I asked falls under informational according to you which is in the other 2 category

and AI generated answer is false

so... what am I missing? lol

[ocean estuary]

You missing the Index and cache I was talking about.

Also that the model doesnt perform searches

and just summarizes the provided information

[warped marsh]

so using its internal knowledge?

[ocean estuary]

Yes its getting the information from its own index and cache. Idk if its also going to the website itself. I tried asking about the Bing chat limit and it stills provides me with the old one even though it was announced on twitter that the limit was raised

[warped marsh]

does the end user care about it?

[ocean estuary]

So its not going to the websites itself probably

[warped marsh]

especially perplexity gives me the source

so it isn't searching its own database

if it did it wouldn't have linked me the source

[ocean estuary]

I think you still not understanding

[warped marsh]

alright alright

Let me put it this way

I ask question... I get answer, right?

[ocean estuary]

I don't know how Perplexity AI is doing it. I think its going to the websites directly

Thats why the information are more accurate because updated information

[warped marsh]

It does the same thing what bing does

[ocean estuary]

no

men

[warped marsh]

it doesn't?

[ocean estuary]

unlike Bing chat who is probably only retrying for its own index or cache

I already gave an example

[warped marsh]

then why does it link the sources?

[ocean estuary]

because it was in the index?

When you search in Bing

how are the links provided?

through its index

[warped marsh]

ok

I understand

then it makes more sense to have traditional searches and AI generated answers both on either side, doesn't it?

[ocean estuary]

Someone needs to add his own website to the Bing index in order for it to crawl it.

Also Bing crawl is updating its index by crawling the websites again of an period time

Also there are saved versions of every website

called cache

Thats what Bing chat is retrying

[ocean estuary]

Nope it will just distract the user

[warped marsh]

if they search the "version" of website that they have stored in their database, then that increases the risk of misinformation

so it makes complete sense to have both the searches on the screen

how is it distraction?

[ocean estuary]

Because of focus

Its confusing having 2 things at onces

Either you have the Chat or the search results

[warped marsh]

even if it increases the odds of accurate information?

[ocean estuary]

That doesnt matter

because its happening in the background

Lets try the instagram example

https://www.perplexity.ai/search/e2cec32a-939b-4032-842f-c825e8cc341c?s=c

Lets see if Bing gets the updated follower count

Use the precise model

[warped marsh]

nah let it be in creative

[ocean estuary]

there is a problem with the 2 instagram pages btw

[ocean estuary]

can you go to https://www.instagram.com/charlieputh/

[warped marsh]

I checked it on my phone

[ocean estuary]

Its saying page not found

[warped marsh]

yeah it says that to me too idk why

I checked it on phone though..

both AI gives the same number

[ocean estuary]

Is this information correct?

Lets try an unkown instagram account

[warped marsh]

yes

ed sheeran is 43.8 but that's okay

inaccuracy is always there

[ocean estuary]

thats the problem

[warped marsh]

yeah no shit

[ocean estuary]

It will be updated in a few hours in its own index

[ocean estuary]

?

See it didnt go the instagram website

Soo its not getting updated and accurate information

That was my point afterall because its only retrying for its own index and the cached version of the website

[warped marsh]

Is this how AI search work?

I used to think that let's say there is a model A and you ask it a question

[ocean estuary]

gpt cant search

[warped marsh]

then in the background bing/google search api is used to communicate with their servers to get the search result to your question. Then the content of the website is fed to the model and then it answers your question

[carmine linden]

Hey @warped marsh! Perplexity AI does not currently provide an API.

[ocean estuary]

Yes

Also try this:

https://www.perplexity.ai/search/e2cec32a-939b-4032-842f-c825e8cc341c?s=c

on Copilot

lets see if it gets the exact number

[warped marsh]

it doesn't

[ocean estuary]

of both followers

You tried?

[warped marsh]

the post you shared already uses copilot

yeah here

[ocean estuary]

its oudated

[ocean estuary]

Are the numbers exactly like the instagram followers count?

[warped marsh]

nope

[ocean estuary]

There is also a problem with buttons and soo on.

Because it cant click it

Also

Cloudflare protection

There is also Web browsing in Chatgpt that does go to websites and clicks it

[warped marsh]

that is slow af

it is in very early stage

[ocean estuary]

Ofc its slow

because its searching for the robots.txt

[warped marsh]

yeah

[ocean estuary]

And it cant do it faster

So now you know why its not getting updated and accurate information but there is also an another issue

Troll or fake news and other websites with lies. So its retrying inaccurate information also

[warped marsh]

yeah yeah

this is exactly why both the ways should be side by side imo

even if it "distracts" the user

[ocean estuary]

Nope it should be in a slidebar

so you can hide or unhide it

[warped marsh]

sure

that should be a good way too

[ocean estuary]

But it already provides you the links for you to check tho

[warped marsh]

sometimes those link do not contain the information the bot is giving

[ocean estuary]

Yea either it failed to read it or ad popup is blocking it. Or in most cases cloudflare protection

[warped marsh]

right

so these are the reasons it gets things wrong

[ocean estuary]

Thats why chatgpts web browsing sucks atm

[warped marsh]

so at the end of the end it does get things wrong, that is precisely why I said both the searches should be side by side or your way of sidebar

[ocean estuary]

Also you could let Bing read webpages via Edge

Soo you can do the searching for yourself and let it summarize

[warped marsh]

I know

I use it a lot

[ocean estuary]

In Edge there's a sidebar to read webpages

[warped marsh]

but it still sometimes fail to get the information from the webpages even if it can read the website

idk why

even if that part is a text and not an image

[ocean estuary]

But atm Bing chat is heavily censored and flags many stuff as dangerous and also likes to close the chat a lot if thinks youre arquing with it. Gpt models doesnt have this problems cuz I can arque with perplexity all night.

[warped marsh]

yeah I have run into it a few times

[ocean estuary]

without it deleting the whole response

[warped marsh]

yeahthat sucks

[ocean estuary]

thats just a mod filter build inside and not the GPT4 or 3.5 Bing AI though

[warped marsh]

yeah yeah

They need to do it to avoid negative pr though

[ocean estuary]

So the mod filter sucks the most and makes the chat unusable

[warped marsh]

I get that

[warped marsh]

It hallucinates a lot too

that is why I almost never ask a follow up to bingchat

[ocean estuary]

Yup mostly in creative mode

[warped marsh]

I find its first answer accuracy to be the best

[warped marsh]

creative is the most powerful model though

yeah

[ocean estuary]

Yea but they make it dumber. Like u saw in the screenshots above. Saying that https everywhere still exists or going completly insane with name generations lol

[warped marsh]

absolutely

I agree

[ocean estuary]

Oh tbh It sucks even more at math cuz gpt models are not build for this.

[warped marsh]

I was just talking about the importance of traditional web searches and why it is important to have it on the side as well.. and your idea about the sidebar to hide it if you dont want

[ocean estuary]

Thats why I dont use it anymore and switched to alternatives because I dont have to arque xD

Or being scared saying the wrong thing

[fossil sentinel]

It rounds it up

[toxic salmon]

why is dream perplexity not a website and only a discord bot?

[neat elk]

it is not part of our core product so we haven't prioritized it yet to be on the website

[toxic salmon]

is it based on mid journey?

i mean like the api

[tame current]

midjourney doesn't have an api lol

I think it uses Dall-E

that's what some people were saying in another server, anyway

[humble night]

So glad it's out for android finally. Was using a browser page shortcut until now lol

[heady fiber]

Lgtm

[dry finch]

The Android app won't let me sign in with email. Only option is to login with Google

[void veldt]

I js downloaded it on android I don't seem to find the copilot thingy

[neat elk]

both features (signup with email and copilot) are not there yet, but will be released very soon

[fossil sentinel]

Can't wait for copilot to come to the app it'll be very convenient

[craggy pollen]

Fire

[craggy pollen]

Oops

[wintry lake]

The android app isnt available on my device sadly lol

Ill go with the website its much better

🤣

[signal hamlet]

Hey @wintry lake! What error do you get? You can DM me the details about your device as well, let us check.

[wintry lake]

Its alright im using an old device

[fossil sentinel]

I was scrolling gpt related reddit and came across computegpt.org which seems kinda interesting

Not very mobile friendly so guess I'll test it out tomorrow

[ocean estuary]

I think the website is more suitable for me

[wintry lake]

Exactly

[wintry lake]

It sucks

How good is jasper chat?

[broken vector]

I won AI in tic tac toe

[wintry lake]

How did u do it

Made a facebook account only for this ai

It can do more

Plus its offline🤷

[raven kayak]

Hey, perplexity can do maths?

[hidden laurel]

Will we eventually be able to share our threads between the web and the app?

[pine lodge]

Yes! It’s already in for Android and will be there soon for ios

[tame current]

what

wdym

/imagine ancient roman

[carmine linden]

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

[white laurel]

No, all AI struggle with logical deduction, altough I think Perplexity can give you defentions, formulas and formulation theorems

[wide dust]

Does Perplexity for Android come with Copilot (like the web version)?

[signal hamlet]

Not yet, @wide dust! But it will be added soon.

[ocean estuary]

Oh google acc is required

sad

Waaaaaaaaaait a minute

is that a bug?

[teal ridge]

is it fixed now?

[fossil sentinel]

Does the ability to choose gpt 4 as a default model in pro refer to something like Enhanced mode?

[ocean estuary]

lemme try again

[teal ridge]

you can set it how you want in account settings

[ocean estuary]

Yup

[fossil sentinel]

o that's dope

[tame current]

why did you guys get rid of the detailed option?

it makes no sense

you just made perplexity less capable..

[wraith crow]

the answers are detailed by default right now

[warped marsh]

woah woah chill

[tame current]

no they are all concise and short now the more detailed one was better.

[tame current]

? im relaxed

[warped marsh]

It is mostly accurate and answers the question

[tame current]

I didn't even curse..

[warped marsh]

how do you find detailed answers better and what is missing in the current answer mode

You can put your feedback in #👉│feedback-general and the devs will look at it @tame current

[tame current]

they were just better everytime i clicked in a question it gave me the concise answer but then i clicked on the more detailed and i got way more sentences now im just getting 2-3 sentences the more detailed option gave a lot more info

[warped marsh]

Plus I am pretty sure you can ask the bot to give you a longer, more detailed answer in the follow up question and it will give it you

does that not work for you?

[tame current]

I need to ask perplexity to be more detailed now? when before i could just click a option and it would do it for me? seriously..

[warped marsh]

no you shouldn't ask it. I get it

It's just majority of the people don't face this problem of not having the detailed answer as the answer they get are sufficient

If they want a more simplifed or detailed answer then they just ask for it like they would with, say, chatgpt

[tame current]

ok well somethings shouldn't be updated so much

otherwise sometimes they get worse same goes with all things

now im not saying perplexity got worse

but it is just a saying...

[warped marsh]

You can put some examples in #👉│feedback-general by sharing some links and explaining how the current answer doesn't suit your needs and I am sure the devs will have a look at it

[tame current]

ok

I was just asking a genuine question i wasn't trying to be aggressive in anyway

[warped marsh]

yeah yeah I am sorry if I came a little aggressive initially. I have been using reddit a lot so I hope you understand 🤷‍♂️

@tame current I tried this as an example. Is this not detailed enough?

It went on to even talk about discord after answering my question

[tame current]

mm interesting let me do the same thing..

Mmmmmm

[tame current]

I asked who the Ceo of discord was and it gave me less then a sentence

[warped marsh]

uhm

[tame current]

Something is wrong here

[warped marsh]

you are using the web version, right ?

[tame current]

Yes..

How are we both using the same Ai and asking the same question yet you got a more detailed answer and mine was not even up to a sentence.....

[warped marsh]

weird maybe try again cause my answers are pretty detailed

[warped marsh]

yeah LLMs are inconsistent in this way

[tame current]

This doesn't make any sense

And if I at least had a detailed option I won't be experiencing this

[pine lodge]

can you share a link to your response?

the answers returned should be detailed by default now

[tame current]

A link to my responce..

OK I will see

[pine lodge]

you can copy the URL from the browser!

[warped marsh]

The more detailed the answers are, the more likely it is to experience hallucinations, right?

[pine lodge]

not always.

[tame current]

https://www.perplexity.ai/search/0d5a4d5b-6282-41fe-8df1-2199a733fdc0

[pine lodge]

it seems like it is returning a detailed response, I think

[tame current]

that was the first question i asked

then i asked the second one was the "who is the CEO of discord one

[pine lodge]

Yeah, can you share that one with just one sentence

[tame current]

the first answer was surprisingly detailed

[warped marsh]

what do you mean by just one sentence?

one thread?

[pine lodge]

oh, the screenshot they shared showed only one sentence for the response

[tame current]

but it is all within one thread

[pine lodge]

ah

[warped marsh]

The question is in the link shared

the second question

[tame current]

give me a sec

[pine lodge]

no you're good, I didn't scroll

cool, we'll take a look into it

[tame current]

oh ok

[pine lodge]

personally I usually switch to a new thread when asking something very different, but that's my habit. can you see if that gives you better answers?

[tame current]

i just liked the detailed option a lot though

[pine lodge]

yeah, it should be returning just the detailed option at the moment

most responses should be longer than what they were before

[tame current]

but it is strange because i did it again and I created a new thread and asked "who is the CEO of disord" and it gave me a detailed answer and this is the first question of the new thread but the second question of the other thread was only 6 words

[warped marsh]

@pine lodge can I pass in a link to one of the pens in codepen and ask it to explain that code to me? It doesn't seem to work so is this a bug or is perplexity not meant to explain any codepen?

[tame current]

so it is like only detailed under certain circumstances..

[pine lodge]

I don't think it's particularly good at reading code from codepen at the moment (iirc - I haven't tried myself.)

[warped marsh]

can the paid version do it?

or what if I paste in the code and give it the context.. it then should be able to explain it to me, right? haven't tested this approach

[pine lodge]

If you paste in the code I think it'll have a better chance

[warped marsh]

Alright. Thanks for answering

[tame current]

give me a sec i need to reset somethings in my discord acc

so ima leave and come back

[neat elk]

the follow up answers are tend to be shorter because their purpose is to clarify something

your example is a bit non standard as it asked an unrelated question

we will take a look to see if we can handle it better

[fossil sentinel]

💀

[tame current]

@teal ridge Is there a plan for adding the amoled theme on Android?

[coarse kindle]

Simple questions warrant for simple answers. Also, you can teach the AI to give you more satisfactory responses, as shown here. Perplexity answers the question in the same way for me, but if I change the wording, I get a different answer.

[coarse kindle]

And not to sound like a dork, but "the CEO of Discord is Jason Citron" is indeed a complete sentence

[tame current]

Awesome, I was waiting for a premium subscription

[pliant silo]

Is there a perplexityai api?

[fading rose]

#📘│faq

[zenith gulch]

@signal hamlet e, should i dm how it works?

im looking

at the screen

[signal hamlet]

Yes, please, @zenith gulch

[zenith gulch]

ok

[rain rampart]

Perplexity is great, can be used without logins

Efficient and Uniquely balanced other than Chat GPT

Imagine this can be used to prove your theory everyday ahha

[tidal holly]

Chupapi

[carmine linden]

Hey @tidal holly! Perplexity AI does not currently provide an API.

[tidal holly]

Muyayo

[solar scaffold]

api

[carmine linden]

Hey @solar scaffold! Perplexity AI does not currently provide an API.

[solar scaffold]

robloxapichicken

[carmine linden]

Hey @solar scaffold! Perplexity AI does not currently provide an API.

[solar scaffold]

a p i

ap i

interesting choice of words

[white laurel]

No,it doesn't #👋│introductions message

[carmine fern]

he was testing the keyword for the auto response from the bot 👍

[ocean estuary]

I dont think Chatgpt has a API key

Ik only this:
https://openai.com/blog/openai-api/

[white laurel]

Oh, yes. Misread the sentence as question

[wintry lake]

Im literally addicted to this

Its all bot

[hidden prism]

Hello people

[narrow shadow]

Is there a prompt for summarize paragraph in a effective way? Thank you

[lime glade]

How?

Pretty sure there gonna try to fix it

[neat elk]

there will be something available for this very soon

[chrome ivy]

o

la

[tame current]

m interesting i wanna learn more about this

[strong pond]

What’s this

[proud grove]

Perplexity Pro vs ChatGPT Plus? 🤔

[grizzled canyon]

they use openai api dont they?

[carmine linden]

Hey @grizzled canyon! Perplexity AI does not currently provide an API.

[grizzled canyon]

thanks mate LMAO

[upbeat turret]

No, it's a Perplexity bot that somehow detects popular topics on the Internet and shows them

[tame current]

Hi

[pallid trout]

I submitted the plus plan... I see that I can use gpt4 for all queries, even non copilot. So.... What is the difference between copilot and normal for plus user if both use gpt4?

[swift moat]

copilot can ask you follow up questions. the normal one cannot.

[ocean estuary]

It can?

[swift moat]

depending on the question

it can ask you follow up questions

and use those to refine the answer

[ocean estuary]

Ah u mean this one

yea but normal one also ask you tho. But not that detailed

[swift moat]

the normal perplexity does not ask follow up questions like this

[indigo crystal]

Hi Alex
I cannot write on multiple lines when editing a query. Clicking "shift+enter" sends off the prompt.

[signal hamlet]

Thanks for reporting, @indigo crystal! Let me check.

[neat elk]

copilot also runs several search queries and does much deeper analysis of search results than normal search

[warped marsh]

in a nutshell, copilot is us doing a research on something and spending hours on it but faster

[narrow silo]

what?

[warped marsh]

^^

[narrow silo]

how do you mean spending hours on it

[warped marsh]

I don't know about you but when I do research on something that requires deep analysis of search results, it usually takes me hours

:|

[narrow silo]

can these webgpt whatevers fix my JS code

[warped marsh]

sure give it the code and ask it to fix whatever problem you have. Try to give it the error you are getting to get better results

[tame current]

^it works most of the time

[narrow silo]

this thing actually helped me with code. wow

neat!

[narrow silo]

vhy is perplexity always checking if site connection is secure?

[signal hamlet]

Hey @narrow silo! Unfortunately, CloudFlare checks are important for preventing spam and abuse of our public service.

[narrow silo]

i mean sure, but every time?

like, why doesnt it store some cookie or something, so it doesnt ask every time

[merry kelp]

Does anyone know the token limit of the GPT4 API in Perplexity?

[proud grove]

Good question. Additionally, how does the token limit compare to ChatGPT 4?

[neat elk]

we don't have API, not fully understanding your question, are you asking what is the max query length?

[fossil sentinel]

Is there like a max output

[neat elk]

ah I see, it is pretty big right now, 1500 tokens

[timid zephyr]

Is there a plan to increase the token amount for pro users?

[neat elk]

yes we can do it, I wonder what are the usecases that require more than that? could you share examples?

[timid zephyr]

Can you clarify if the reference to the "token limit" refers to the maximum number of words or text units that can be used as context to generate a coherent response, or does it refer to the length of the text?

[neat elk]

here, I was speaking about the output tokens, the input tokens have larger limit

[timid zephyr]

Ah, I understand. Then that's perfect, hahaha.

[merry kelp]

Thank you for your answer! Some usecases I would use it whether it has a bigger output would be to improve a larger piece of code or to help solving mathematical or physics complex exercises. I think that if a more conversational mode (ChatGPT like) is added this would be a big improvement but I’m not sure if a token output increase would be feasible or not.

[vast pasture]

hello may I ask when the Pro version will be added to the main site? I really want to upgrade to Pro but I dont want to spend 200$ on the beta site. Will the "Pro" featrues be transferred to the main site when the pro comes to the main site?

[devout geyser]

as I understand it, the pro features will come over to the main website eventually but no idea when. That said, on the main website I get 100 uses per 4 hours on copilot (which I believe is a pro feature?) and on the beta website I also get 100 uses of copilot. Both currently seem independent of each other but that may not stay like that permanently

so, at the moment it appears you get 200 uses every 4 hours on copilot if you use both the beta website and main website, but that could change at some point (e.g. so they are possibly synchronised)

[vast pasture]

Alright I get it now. I will create a Pro account through the Beta site this coming September then. Thanks a lot.

Btw, once you register to be a Pro user from the beta site, if you also get pro features for your account on the main site;

Why is pro feature bot on the main site? I am a bit confused.😵‍💫

[wraith crow]

The pro subscription is account bound. You'll get the pro features on both the beta (if there is one) and the main site as long as you login with the account you have the subscription on.

[tame current]

How to get Discord Auth Code?

[neat elk]

yes it will be transferred, we will release the pro version next week

yes, the pro version will be connected to your account and available on all the platforms (web, mobile web, iOS, android)

[narrow silo]

will pro version finally tell me how to cheat in poker

[rich bane]

ask it with copilot on 😛

[wicked blade]

Is the response limit still only five in Copilot mode? They could, at a minimum, leave a suggestive offer of at least 25 requests, just like Phind.com does.

[wraith crow]

If you think the limit is too low for your usage, there is now a pro subscription available that will give you way more uses for copilot.

[wicked blade]

Ah that's great. I wasn't aware yet.

[wraith crow]

The pro subscription is currently only available on the beta website: https://beta.perplexity.ai/

[wicked blade]

Is it still in beta?

[neat elk]

yes, but will be released in a couple of days

heads up, added support direct interaction with a model -- "Writing"

[narrow silo]

shall we have /write command soon here

@neat elk is this gonna remain open

[neat elk]

we don't have plans to add this to discord, the main reason for this features is to help users who can't pay for both Perplexity and chagpt

[narrow silo]

can perplexity technically fix my css code if it clashes?

[south widget]

I think it would help if there was some sort of FAQ or "why perplexity?" blog post. I have chatgpt+, vicuna 11b locally, privategpt locally, and blip-2 locally playing around various things (text, images, etc), what does perplexity offer over chatgpt+ at a minimum? Once perplexity pro is public, i'll probably switch from chatgpt+ to it, but just wondering what is the roadmap, why perplexity, etc for the general public

what i've found so far, is that your web interface is better, the real-time access to the internet is better than chatgpt+

and i guess for the general public, having android/ios apps is ahead of chatgpt+

(i don't have android/ios, so it's irrelevant for me)

[tame current]

wow so i can write entire summaries with this?

[fossil sentinel]

https://www.zdnet.com/article/chatgpt-now-lets-you-create-and-share-links-to-your-conversations/

💀

[near glacier]

are you wishing to add the perplexity bot to your server?

[near glacier]

They mentioned no public roadmap cause this space is highly competitive.

In terms of accessibility, Perplexity doenst ask privacy-invasive things like mobile number.

In terms of accuracy, I've tried a few different options and for my use-case of translations, Perplexity is much better than the others

[merry kelp]

Thank you very much!

[neat elk]

yes, should be possible, try it out

[south widget]

i couldn't find past threads on public roadmap in here

maybe i'm in the wrong channel to search for it

[neat elk]

what do you mean?

[south widget]

i've built companies in highly competitive environments and public roadmap is exactly what people want for stability and confidence

but that's my $0.02

[neat elk]

we basically don't want to reveal certain things ahead of the time to stay competitive

certain other things we are ok to share

[south widget]

as far as i can tell, #👉│feedback-general message this thread is the only thread that mentions public roadmap

maybe my search terms are weak, i should get the bot to write better ones for me 😉

[merry kelp]

I do not know the difficulty this might entail, but I would really like to have the ability to "properly" visualize the LaTeX formulas returned by Perplexity in the output, so that I can see them formatted. I would also like to be able to export the answers to PDF and to be able to delete all the threads with a button. The new writing option is amazing!

[neat elk]

you can get raw latex format if you copy the text

[tame current]

Is that mathematics?

[neat elk]

The equation for the normal distribution, also known as the Gaussian distribution, is given by the probability density function (PDF):

$$
f(x) = \frac{1}{\sigma \sqrt{2\pi}} e^{ -\frac{1}{2} \left(\frac{x - \mu}{\sigma}\right)^2 }
$$

Where:

- $f(x)$ represents the probability density at point $x$.
- $\mu$ is the mean of the distribution.
- $\sigma$ is the standard deviation of the distribution.
- $e$ is the base of the natural logarithm (approximately equal to 2.71828).
- $\pi$ is the mathematical constant Pi (approximately equal to 3.14159).

The normal distribution is characterized by its bell-shaped curve, with the mean value at the center of the curve and the standard deviation determining the width of the curve. The total area under the curve is equal to 1, representing the total probability of all possible outcomes.

By Perplexity at https://www.perplexity.ai/search/c4424577-9f4a-4205-8360-64c543fd1402

this button

[merry kelp]

Yes!

[merry kelp]

Didn't know that, this is great! I think it is not possible to visualise the not formula but mathematical enquiries with $...$ too in the answer like the formula for normal distribution you posted, right?

[neat elk]

sometimes the model makes mistakes, and outputs $ instead of $$, we are trying to fix it

[tame current]

Do you guys train a custom openAI model

[neat elk]

we cannot share this information unfortunatelly

[merry kelp]

When we don't write two dollar signs is to avoid the formula to be placed in the center of the result PDF (with two dollar signs the formula is placed in the middle of the line), so, that may be the reason. Thank you for your answer!

[devout geyser]

I haven't checked yet, but is the writing option also limited to 4,000 characters?

[tame current]

4,000 is a lot of words..

[tame current]

that is equal to like 10 paragraphs or more..

[lucid pelican]

To avoid issues with $$…$$ you could do some fine tuning (if you are fine tuning) to get it to put out /[ … /] instead (but with the slashes in the opposite direction as discord has issues displaying), it does the same thing but is slightly more robust to errors and avoids hallucinations with $…$ which are used “in line”, whereas $$…$$ is for equations that are centered, looking at https://stackoverflow.com/questions/2251417/latex-dollar-sign-vs might be helpful

[queen pagoda]

Hmm what do you guys think about this

[daring galleon]

Hey guys! Can I use perplexity pro on android app?

[signal hamlet]

Hey @daring galleon! Currently it's available on the web, not yet on the apps.

[devout geyser]

You'd be surprised how long some code can be 😂

For example, if I were to use a smaller piece of code from one of my projects it works out at 1,861 characters, which is fine, but that's also only 761 tokens. Given that the average word is 4 characters and that the limit for GPT-4 is in tokens not characters, imagine something that's nearly 4,000 characters... that could possibly only be somewhere around 1,500-2,000 tokens which GPT-4's 8k model is limited to around 8,000 tokens. Sure, you shouldn't use all 8k~ but it would be nice if we could use nearer to 4,000 token's worth in the future for the writing option if we need to. Even GPT-3.5 turbo is around 8k I believe

That said, applying a lower limitation is also understandable given cost, so that might be why it's still limited to 4,000 characters

[tame current]

Hi there. Is it OK to ask here about the Android app?

[narrow silo]

Is there a chance of adding bot to another servers?

[wraith crow]

If you click on the bot his profile you’ll see an option to add it to your own server

[tame current]

I run the Android app on LineageOS. I cannot sign in with my email, as I do on the web version. Any suggestion?

[pine lodge]

this is a known issue (we first implemented google signin). email signin will be coming very soon

[tame current]

@pine lodge got it, thanks.

[narrow silo]

And I missed it. Lol. Thanks.

[gray knoll]

When perplexity pro available on the app.... I've paid pro but can't find any options in app

[neat elk]

we are trying to ship it as fast as possible, please bear with us

[rigid root]

ترجمة

[tame current]

perplexity best ai app

[lime glade]

Will perplexity AI get more features? Such as since we have Wikipedia, YouTube, and etc. can we see more to come

[tame current]

Yeah also will perplexity always be free? They have to make money at some point and I am depising the day it turns into a subscription based service

[narrow silo]

well they have pro option

so that will cost money

@latent wharf how to get Perplexity PRO for free?

[tame current]

they have a pro version really ive never seen it before

[narrow silo]

it's literally new and all they talk about lately

[wraith crow]

The pro subscription is currently only available on the beta version of the website: https://beta.perplexity.ai/

[narrow silo]

does pro version translate text?

because free version doesnt wanna

[wraith crow]

With the pro version you can use the gpt 4 model for writing. #💬│general message

[narrow silo]

writing what exactly

[tame current]

anything

[wraith crow]

text and code just like chatgpt can so you can also translate text with that model

[neat elk]

yes, there will be a lot of new features, we are just getting started

[neat elk]

we intend to keep the current non-pro version of perplexity free

[neat elk]

you can't

[narrow silo]

LOL. I didnt ask you, I Asked perplexity 😄

It was a joke, of course

[neat elk]

you can translate text with non pro version too, just select writing

[narrow silo]

if you see my ask in #ask it did not deliver

[neat elk]

https://www.perplexity.ai/search/16f4d5a5-89f4-4aaa-95b2-e1ecf36507eb?s=u

[narrow silo]

it only instructed me to go to web translators

[neat elk]

the bot doesn't know this information yet

[narrow silo]

thin?

ah. this

so doesnt know how to translate from Arabic language?

[neat elk]

I was talking about pro version

[narrow silo]

but doesnt bot also search through the web in order to provide information

[neat elk]

it does, but this info is not on web

[merry kelp]

Is there any stop button to stop Perplexity of writing the answer? I haven't seen it and I don't know if it's me or if Perplexity doesn't have it yet

[white sundial]

No, unfourtunately perplexity does not have a stop button yet, but yeah, its a very commonly requested feature, so hopefully it comes out soon

[neat elk]

yes, it will

[tame current]

Wait what??. Why would you want to stop perplexity from writing the answer for your question??

[white sundial]

Like if you asked a question, then a second later realised its a dumb question, and you dont want to wait a minute while perplexity goes off on a tangent, about a dumb question you asked... Its also a feature on chatgpt if you used that before

[half lagoon]

I tried to read, but any way to run it locally?

[fossil sentinel]

Finally I can try out googles sge

[tame current]

Then just delete the question then..or edit it If you realized it is not a good question then just delete or edit..

They is an edit button as well..

[fossil sentinel]

It's too time consuming to wait

[tame current]

How long does it take for perplexity to generate an answer? 10 minutes? Because a lot of the time it only takes a few seconds and then you can either edit or delete it later...

3-5 seconds doesn't seem time consuming to me..

[fossil sentinel]

copilot often takes longer at least for the tasks i require

[tame current]

Oh no not even 3-5 sec sometimes it takes only 1-2 seconds..

[fossil sentinel]

If it takes 1-2 seconds for copilot to answer you, chances are the normal mode is more suitable

[wraith crow]

There is no way to run Perplexity locally. You can either use the website or the apps (ios, android).

[astral crow]

alguien de aqui tambien entiende el español?

[unreal chasm]

Is it possible to set Perplexity as my default search engine in Chrome?

[wraith crow]

yes it is! here is a tutorial #💗│sharing message

[narrow silo]

Just put it as homepage. Heh

Every time AI draws something that blends it's like : now we do a little trolling

[narrow silo]

What was that AI you guys mentioned, but code

specified for code

[carmine fern]

i assume it's phind

that u are reffering to

[narrow silo]

ah yes, it is indeed. thanks

[gaunt iron]

Hello!

How are you?

[swift moat]

puedo usar un traductor. no estoy seguro de si eso cuenta

[gaunt iron]

Yo soy de España.

OK

[lofty dagger]

hewwo

[gaunt iron]

Hello everyone!

I explain the free programming and database courses in Spanish.

Excuse me.

[river falcon]

hola papus

[rich bane]

:p

[astral crow]

gracias

[astral crow]

no pero de español como lengua Principal

[rain silo]

Does perplexity use any ui kit? or its just custom tailwind

[worldly bluff]

Its not difficult to check in the console right? They're using tailwind 100%

[narrow silo]

fml

and what is tailwind but an ui kit

[swift moat]

Por desgracia, el inglés es mi lengua principal. Lo siento.

[atomic anchor]

hey

[naive fern]

Hi, is there a Pro version of Perplexity to remove the limit of Copilots?

[wraith crow]

You can subscribe to the pro subscription of Perplexity on https://beta.perplexity.ai/ till will give you 100 copilot uses every 4 hours

[ruby bough]

Does Perplexity Pro include more plugins such as noteable or code interpreter?

And what other AI models are available apart from GPT3.5/GPT4 (for copilot)?

[devout geyser]

No other plugins that I'm aware of. However you have ability to always use GPT-4 with or without copilot as Pro if you want to

[ruby bough]

Thanks. Thats helpful. The day we get noteable and code interpreter in perplexity, I will be moving to Perplexity Pro from ChatGPT Plus in a blink of an Eye. Ha

[devout geyser]

Understandable, I never had chance to use code interpreter sadly

[ruby bough]

Me neither, I have been on wait list since day-1. I have been using noteable for python. Its really very helpful.

[devout geyser]

Haha, I see

[ruby bough]

I do think $20 a month for perplexity without many plugins other than Internet access and (maybe academic filters) is a little much. I might be open to go with pro for $10 a month

[devout geyser]

Though on the other hand you're getting 100 messages with GPT-4 every 4 hours, which is much higher. Copilot is helpful for doing additional research on the internet which GPT-4 might not have up to date knowledge about

If you reduce that window to every 3 hours, to match with the other service, in comparison you're still getting 75 messages every 3 hours with Perplexity compared to just 25

[astral crow]

cno problem

[tame current]

sometimes perplexity gives a lot of information for certain questions that have to do with certain things and like Zero info for others questions

[signal hamlet]

Hi @tame current! For some specific questions very few sources can be found, I assume. But if you feel that the lack of info found for certain questions is more of a bug, then please share those questions with permalinks in our feedback channels, so we can check those, too, improve or fix.

[final prawn]

does perplex have any input/outpu plugins

[fossil sentinel]

Ayo? 👀

I like it

[south widget]

Does anyone know if queries/threads/data sent to perplexity also feed into the training data of openai? I can't find an answer in the terms of service or privacy policy.

[neat elk]

no it is not, our data is not used by open ai

[grizzled canyon]

read the openai policy

[south widget]

which one?

[south widget]

thank you

[fallow locust]

you can opt out of training if you email them

[south widget]

or sign an enterprise license with them

[grizzled canyon]

yep

[indigo quarry]

Will probably email

[mental edge]

Hi,
i just tried this via the website without signing up, etc.
seems to work really well. I'm paying for GPT-4 at the moment, what's the catch/limitation with perplexity?
couldn't find any payments / plans page

[indigo quarry]

Didn’t know there was a paid version

[mental edge]

Maybe there isn't, I've just found this (perplexity)
I've been paying for the openai chatgpt-4 which lets me use the newer model 25 times per hour

[wraith crow]

Hi the pro version is currently only available to buy on the beta website: https://beta.perplexity.ai/

[vital pike]

#Motivation

[rough haven]

What's the history of perplexity? How did you guys start? Find such massive fundings etc?

[lyric magnet]

[tidal holly]

Nooooooooo

[fossil sentinel]

How come it shows option to get pro on the normal site then?

[wraith crow]

You're right they pushed it to the main website too now. I don't see that popup anymore since I already have the pro subscription. So didn't notice that.

[tame current]

That does not answer the question.

I think Cryptex was asking whether it is Vanilla Tailwind or using an additional UI kit on top of that.