💬│general

[tame current]

I think that just means the tell the LLM to generate a search query

having a vector db for the whole internet is extremely impractical lol

[tame current]

Tad weird

“If you do A, B might still happen. In order to prevent B from still happening, do A.”

[floral merlin]

I do recall one time I get a confused answer like this 🙂

most likely due to the fact that the model is confused with so many sources

does the model have the ability to verify the web results that it's being fed to? or just blindly summarize them? 🤔

I've added this to my ai profile, let's see how it goes

[hollow mortar]

Will Claude 2 become available on the Free plan?

[pallid elbow]

Probably no. Claude 2 is basically GPT-4 but anthropics attempt

[tame current]

I guess it's a possibility

I think claude is cheaper than gpt4

[pallid elbow]

Well since POE needs a premium to use unlimitedly maybe copilot could have Claude 2?

[halcyon coral]

https://www.perplexity.ai/search/355531b2-fec6-475d-b39e-7e69d88d2f34?s=c

[tame current]

Perhaps blindly

I mean, it did recommend a piracy site to me after all

[serene wigeon]

Perplexity.ai with Claude will try to satisfy your requests for example posts by making them up, and when you ask for links for them it makes up obviously fake links, I'll post screenshots if you don't belive me

[halcyon coral]

Sharing in #💗│sharing, bugs in #1140622008086970420!

[finite spoke]

Can't we add multiple files to a new thread. Perplexity Pro is only allowing one file attachment with GPT-4 and Claude model. But in Chat GPT plus I was able to add two document. I am doing anything wrong. suggestions plz

[halcyon coral]

For now, there are a few limitations: only text content (no images), only one file at a time, and no redownload, but we hope to address these over time.
#📢│announcements message

[finite spoke]

Thank you . It would be helpful to add two files. If we need to analyze two files and come up with a conclusion it would be helpful. For e.g two research paper and get key points or differences

[idle zodiac]

Im curious to know how our data is handled ?

[halcyon coral]

Your data will be handled in accordance with the Privacy Policy:
https://www.perplexity.ai/privacy

[idle zodiac]

Fantastic would there be an update or announcement if any changes have been made ?

[halcyon coral]

There are no announcements here on Discord about this, but the top of the website shows the date of the last change.

[blazing idol]

Why is perplexity bot down

[lament imp]

It should also be noted that the use cases differ by a bit. If you're looking to do programming/writing/data analysis or summarization of documents, I found that it provides clearer outcomes. However, GPT-4 feels a bit better in business/general use. Claude 2 seems to have better understanding in terms of clarifying what may be causing confusion as well. For coding specifically, they seem neck in neck, but I think GPT-4 can do more complicated problems but generally, Claude seems to be good for most of the cases I am using it for

Also Claude has better mathematics abilities supposedly. Also seems to deal with specific topics better than gpt-4.

[idle zodiac]

Is Perplexity Pro essentially just sending API requests to Claude2 and GPT-4 as a subscriber ? If I go Pro on here do I not have to subscribe to GPT-4, I would assume no because OpenAI still needs to make it money ?

[violet urchin]

Who would be the person to talk about marketing opportunities?

[lament imp]

I believe that while it utilizes GPT-4, it is an independent service that OpenAI makes money from by charging for the GPT-4 usage. This is why there are pricing models such as this one: https://openai.com/pricing

I more look at Perplexity as a way to gain access to these various models, while providing functionality that is unique to Perplexity itself.

Ofc Perplexity also fine-tunes what is being shown to you, so (I assume) there is a customized version of both GPT-4/3.5 to improve upon them

[halcyon coral]

Perplexity is standalone and you don't have to subscribe to anything else. Perplexity uses the APIs of OpenAI and Anthropic (Claude), among others, for its service.

[echo kelp]

hi, anyone having issues loading https://www.perplexity.ai/ on desktop, chrome?

app and chrome extension working, but webpage stuck on the 'checking is sit is secure' page

[halcyon coral]

#1134574720822558782 message

[echo kelp]

sorry brother thought i replied to you, thanks again my man, much appreciated.

[halcyon coral]

You're welcome 😉

[echo kelp]

you're both awesome, hope your both having a great week, looking forward to trying out claude 2

[coarse sequoia]

hey everyone, really inspired by the team at perplexity! We're working on a similar tool for farmers

[idle zodiac]

OpenAI as in GPT-4 or 3.5 ? Apologies I'm just crossing my T's and dotting my i's - tell me its GPT-4, 24/7 without throttling to 3.5 and I'm sold sir !

[idle zodiac]

@fierce trail

[grizzled canyon]

they use the ai service from azure, the same interface which openai uses

[echo salmon]

Hi. I am relatively new to Perplexity and my mother tongue is Brazilian Portuguese.

I was using it in Portuguese (as defined on my AI profile), but after realizing Perplexity searched only local sources, I switched back to English (which btw is 60%+ of intern content)

My current workaround is to prompt in English and then ask to translate in a second prompt

HOWEVER, the translated prompt give me crazy source links. Like google translate. Is this expected?

Example: https://www.perplexity.ai/search/4fd09311-fe00-4b5a-ab2d-774ffc7d1a62?s=mn

[shut gyro]

Just got access to Google generative AI and it seems to do the same thing as Perplexity

No file upload, specific focus, custom instructions, or choice of AI model though

[idle zodiac]

In time... Major players in AI will attain all the tools but this is the testing A.I phase weeding out the trash of what works and what doesn't.

[halcyon coral]

You can use GPT-4 up to 600 times a day, which is almost unlimited.

[orchid relic]

Which one is that?

[halcyon coral]

The internet search focuses on "translate" and "pt-br". I can also see this in all my translations, so I would say it is normal.

[halcyon coral]

Have you ever tried setting the AI Profile to Portuguese but asking in English? https://www.perplexity.ai/search/2ccb4edf-ac43-422d-8b5f-d58c74d230cd?s=u

[humble swan]

🤔

[rich bane]

Hi. Any plans on implementing phi-1?

[jagged ivy]

Image displaying still doesn't work on android?

[sacred trellis]

Almost 0 people have access to GPT-4-32k rn

[halcyon coral]

What do you mean by that?

[sacred trellis]

We are not currently granting access to GPT-4-32K API, but it will be made available at a later date.

Your thread mentions GPT-4-32k

Which is currently not a thing

The GPT-4-8k pricing would be more appropriate to list given availability

[halcyon coral]

Perplexity uses 32k, it was a comparison between Claude 2 100k and GPT-4 32k to compare costs.

[sacred trellis]

Do you have a source on that? I only heard mention of "GPT-4" from Perplexity but not the 32k version

Google search yields 0 results

[halcyon coral]

||My knowledge 😉||
This has been confirmed many times by the team.
#💬│general message

[sacred trellis]

Ah I see now

Perhaps only certain organizations are given access

Thank you for the information!

[halcyon coral]

Beta access for big companies ig

[sacred trellis]

Makes sense

[twin prawn]

Hey guys is our data anonymized on perplexity AI ? I want to upload a personal file, but will it be stored?

[sacred verge]

hi, why my perplexity.ai is always tips the "Check if the site is secure"， it block at checking for a very long time

[wraith crow]

You mean that you're stuck at the cloudflare page?

If you are, please follow these steps to solve the issue #1134574720822558782 message

[sacred verge]

@wraith crow that's right, just stuck at the cloudflare page for a very long time...

[wraith crow]

Go to https://labs.perplexity.ai/
Solve the captcha there and navigate back to https://perplexity.ai/

[sacred verge]

thank you ,that's really a work around, hope perplex can fix this issus later

[wraith crow]

Do you have the perplexity chrome extension installed?

[sacred verge]

yes, I installed the perplex ai compation

yes, I installed the perplex ai compation chrome extension

[wraith crow]

It's a known issue with the chrome extension. Multiple people had the same problem. If you really don't want to navigate to labs.perplexity.ai the next time you could try to disable the chrome extension for the time being.

[sacred verge]

turn off the extension will fix this issue?

I use this extension very infrequently, so I can completely turn it off, which perfectly solves the problem. Thank you very much

[orchid relic]

Anyone else cancel their Chatgpt subscription and get on to perplexity pro?

[ivory basin]

Hi all I decided to try the premium features to test claude2 model in Perplexity. I gave it a shot, and while I'm impressed for the growing and dedication of Perplexity, it appears that claude.ai's chat outperforms perplexity witg the claude2 model when extracting insights from lengthy PDFs. 📚🧐 Could you share your insights on why this divergence in performance?

[sacred trellis]

Perplexity is fine-tuned to be a research assistant which doesn't quite fit my use case so I've bing chilling with ChatGPT Plus

I also prefer custom instructions and the other beta features over at ChatGPT

That's just me though other people have different needs

[tame current]

This new update will be awesome ! : https://discordapp.com/channels/1047197230748151888/1047204950763122820/1081258627618324500

[sacred trellis]

Oh wow it's already here and it is indeed very awesome!

[odd mauve]

Hi, if a Perplexity Pro users asks a complex question in the area of personal or relationship advice, can Perplexity when composing its answers only use info from web searches, or can it can answer using mainly the internal "offline" data contained in the GPT-4 model? In other words, imaging it as GPT-4 with browsing plugins enabled. It can search, but does not really do so as it's much more effective (results in a higher-quality answer if it only uses data which are part of the GPT-4 model).
The alternative would be that perplexity only uses data from web search, user profile, prompt, uploaded file and only uses GPT-4 to summarize that data and present it in a nicely structured response, similarly to what Bing does.

[echo salmon]

you are a genius! Thank you, it solved the issue!

[karmic river]

Hi, i have a problem with perplexity payment, in which chat can i have support?

[halcyon coral]

What exactly is the problem?
I can help you with most things, if not I will forward you to the support 😉

[karmic river]

I've appiled a discount at the checkout, but i don't see it in the current plan section

Is it right?

[halcyon coral]

What discount?

[karmic river]

A coupon code

A discount code I applied to the shopping cart, to be exact SCHOOL25, it had updated the total but then I went to check in the payment section the price did not change

[karmic river]

Sorry, I noticed now looking from pc that a note appeared.

[halcyon coral]

All right, then it seems to work👍

[rustic gate]

Hi everyone. Been using perplexity for a few months and just joined pro monthly. Is there a special procedure I have to take to take advantage of the school25 promo code for the year? thanks

[karmic river]

@halcyon coral Which Ai Model do you advise?

[halcyon coral]

It depends on the use case, but most of the time I use GPT-4 and am very satisfied with the results 😉

[rugged ocean]

Can you change the model on the android app? I can't seem to find a way to switch to gpt4 or Claude 2.

[karmic river]

Is use it mostly to create some ideas for advertising captioning or enhancing phrases

[rustic gate]

Question: I've been using both gpt4 and claude for a while now. Claude 2 is very strong in areas where gpt4 falls short. I've switched the AI in the perplexity settings to Claude2, and wondering how the 300+ copilot limits affect Claude2 versus OpenAI. Are they both limited to 300 uses a day collectively?

[halcyon coral]

Only at https://perplexity.ai atm

[rugged ocean]

Oh 🆗 thanks 😊

[halcyon coral]

600 uses = Copilot/GPT-4/Claude

[rustic gate]

thanks @halcyon coral

btw is there an api doc for usage of the models with perplexity?

[carmine linden]

Hey @rustic gate! Perplexity AI does not currently provide an API.

[halcyon coral]

Hmmm...Claude 2 can be more chat-friendly if you want to develop an idea together with the AI.
But for this use case, only trying helps 😉

[humble swan]

Would it be possible to have a hard token limit

like GPT has

[karmic river]

Thanks dude

[mental brook]

hello guys, i'm new here and I still don't understand the main differences between all AI models in perplexity pro ? (perplexity, claude 2 and gpt4)
also, i'm a chatgpt plus user. Does perplexity pro can replace all my chatgpt using ? (like translation, coding, ...)
for me right now perplexity is "just" a tool to browse X websites and summarize research ^^ am I wrong ?

[karmic river]

Is there any way to use plugins like openai gpt-4 site?

[short creek]

Welcome! You can do pretty much the same things with Perplexity as you can with ChatGPT Plus (and even more since you can experiment more with GPT-4 and Claude2). The capabilities here are very similar to what ChatGPT Plus offers. I imagine you have already tried changing the "focus" of the thread. By default it searches the entire web, but you can also set it to "Writing" mode which makes it function very similarly to ChatGPT Plus.

Also in ChatGPT Plus you are limited to 400 GPT-4 tokens a day. Here you have 600 and you can use them as you want during the day since they dynamically replenish themselves, instead of having to wait 3 hours if you spend the 50 messages that ChatGPT Plus give you every 3 hours.

Unless you are working with some plugin that is just irreplaceable at the moment I'll give it a try.

[short creek]

Also, about your question on the different AI Models.

GPT-3.5:

• I mostly use it for fast web search since Perplexity's system prompt is good and responses are really accurate.

GPT-4:

• I also use this one for searches when I need to ask multiple questions or more advance reasoning. (besides using it with the "Writing" focus as you'd do with GPT-4 in ChatGPT Plus)

Claude2:

• I'm currently still testing this one but seems like it's a really good competitor for GPT-4. It seems way more conversational so far.

[mental brook]

well this is so clear, thx a lot 🙂

[short creek]

I tried to answer the best I could haha. Claude was implemented like 2-3 days ago that's why I'm still testing it.

[serene wigeon]

Perplexity with Claude2 was hallucinating/confabulating/roleplaying/going wild with its imagination in a way I never saw Perplexity do with GPT-4 (literally making up Twitter and Reddit links when I asked it to source its made-up posts)

But I found a workaround/patch, use this as a contextual semi-pre-prompt: "This is important: for this request I am not roleplaying fiction, this is not a request for fictional answers."

[rigid minnow]

Are there some server problems im having a tough time getting perplexity to take the prompts? :/

[halcyon coral]

Error: too long ?

[rigid minnow]

No it just does nothing.

If i use very short prompts it works awesome.

But anything longer it just doesnt take the prompt. Nothing happens :/

[halcyon coral]

Can confirm the problem...I'm trying to figure out what's going on....

[rigid minnow]

Awesome! Thanks a lot 😄 Just started happening now.

[violet ridge]

I just subscribed to the pro plan. I have a question: when I input code, I put the code snippet inside three backticks, but this will prevent my input submission. I simply cannot click the submit button. I clearly remember I can do that before when I was using perplexity ai on chrome. is three backticks not allowed?

[halcyon coral]

#💬│general message should be this bug

[short creek]

It's on OpenAI's end. Affecting both **ChatGPT **and it's API.

[violet ridge]

but I changed to claude-2 model the behavior is the same.

[short creek]

Well that's weird then

[halcyon coral]

Nah, it´s a pplx bug

[short creek]

Yeah but besides that **ChatGPT **is not working properly. **API **is behaving in an odd way too.

[halcyon coral]

All API´s or only GPT-4?

[short creek]

It just says API in here: https://status.openai.com/
Anyway for some reason now it says API is perfectly fine so...

[short creek]

So then It might be on Perplexity's end.

[halcyon coral]

#🔎│ask

[violet ridge]

I highly suspect that in the input, you cannot use shift+enter to begin a new line. any that new line will prevent input submission. at least at my end.

[halcyon coral]

Update:
The character limit seems to be 500, no matter how many tokens there are.

#💗│sharing

[violet ridge]

wow didn't know that... that's quite a small limit...

[halcyon coral]

It's just a bug, the input limit should actually be 2k (tokens) 😉

[short creek]

2k tokens you mean. So around 8000 words.

That's better haha

[halcyon coral]

thx

[short creek]

That could lead to some confussions.

[tropic crest]

I'm seeing claude2 hallucinate a lot as well

its pretty bad

[short creek]

I'm getting some of those but it's a rarity.

Overall I'm having a nice experience with Claude

[violet ridge]

ok thanks for explaining, I hope this bug can get fix soon, currently i'm not able to input anything useful, or maybe i'm too talkative.

[short creek]

Nah, 500 words is just not enough. It's not your fault.

[tropic crest]

I think the main thing is claude2 struggles to figure out if its hallucinating whereas when GPT4 hallucinates it is able to pick up pretty quick

like claude2 doubles down when it hallucinates

[short creek]

Yes!

[tropic crest]

whereas GPT4 will realize and say sorry

asking gpt4 is your analysis correcthelps me get the right answer often

it's really good at self reflection

[short creek]

A good AI Profile might help a lot with that though

[tropic crest]

hmm yeah I find that perplexity doesnt leverage AI profile as well as openAI

[short creek]

It does for me

Are you using writing mode with Claude?

[tropic crest]

i meant openAI sorry

[short creek]

Search Mode does not take the AI Profile under consideration that much

It prioritizes the search results over your AI Profile

[tropic crest]

not often

ah i see

that makes sense

[short creek]

Yeah. Same happens with both Perplexity and GPT-3.5T

I encourage you to try the "Writing" focus with Claude

And let us know what you think

I think it hallucinates way less when using the "Writing" mode

And even though the model says it's cut is on September 2021, it's not. It has been trained to December 2022. You can ask events or information from 2022 and it will provide it.

[wraith grail]

Why does the File processing feature become limited after using it on a few long pdfs? I use it to summarize research papers, and it works at first, but after processing a few files, it stops working and says "too long". I'm a Pro user.

[short creek]

That's weird. Can you share that in here #👑│pro-lounge ?

Maybe they know what's going on

I don't work a lot with PDF's

[halcyon coral]

Should be the character bug 🤔

[short creek]

Yeah but maybe some more people that work a lot with PDF's have different experiences.

[echo salmon]

I am fairly new to Perplexity and Pro too, but i still didn't cancelled ChatGPT pro because their code interpreter is simply marvelous. Until Perplexity allows me to upload csv, excel and help me makes sense of lots of data I am not cancelling ChatGPT Pro, but only due to this.

[short creek]

Code Interpreter is the best thing out there. No competitors at the moment.

It would be crazy if you could upload files mid conversation here in Perplexity too as suggested in this thread by @elder bramble https://discord.com/channels/1047197230748151888/1136797195903975545

[mental brook]

as I cannot upload csv and excel files with my company's firewall, code interpreter is uselss for me atm 😦

[finite spoke]

Claude 2 might be new to Perplexity platform , its loosing context quite frequently . I haven't encountered the issue when i used it https://claude.ai/chats

[short creek]

That's not been happening to me so far.

[echo salmon]

I know, I would never upload company's data as it would probably violate your NDA.

i use mainly for personal stuff.

Recent example: I asked three different doctors to give me history of medication use since 2013. They all gave me text in different formatting with different dates. I simply uploaded a .txt and asked code interpreter to clean data and parse to excel. It took a few tries but excelled at it.

Another: asked it to transform 20 JSON files with my weight data from fitbit (daily since 2016) and it gave me a beautiful excel.

You can be creative with your own data, but I discourage using company or any super confidential data on it

[short creek]

Yeah unless you have access to either Bing Chat Enterprise or ChatGPT Enterprise you should be aware of the data you share with these AI Models.

[serene wigeon]

I tried Writing Mode with Claude2, and my previous request that roleplayed/hallucinated made up Reddit/Twitter posts has been fixed with that, it now returns correct links

It even seems to bypass the RLHF or politically correct system pre-prompts that GPT-4 shoves at me when talking about Covid or Ukraine etc

[short creek]

Wait how is it giving you links in Writing Mode lol

[finite spoke]

oh okie, One way i am trying to avoid it , explaining the context after few conversations. My conversations are quite long and lot of data to analyze(provide output) .

[short creek]

I tried running on long conversations then asking for an explanation on the whole thing and so far it worked just fine. I need to test in more in depth though.

[serene wigeon]

Oh wait, I did it again in Focus=Writing Mode and this time it just says "I apologize, but I do not feel comfortable generating or spreading potentially misleading health information." I guess I didn't do it right earlier

Anyway, I'll try Writing Mode more often

[short creek]

Hope it helps! It depends if you want to search for recently updated information or not.

[serene wigeon]

Oh I definitely use Perplexity.ai for anything even remotely involving anything in the last 2 years, so maybe I won't be using Writing Mode if it doesn't do the search thing

[jagged ivy]

Nice info provided 🤔

[short creek]

That's the thing. I think I said something earlier in this channel but before the FT-GPT-3.5 implementation you could just do "Writing" and use "Copilot" for specific queries that you'd need to search on the web

Now it's just not possible. I wish they added a "web search" toggle thing as Bing is going to do with their plugins soon

[mental brook]

do you know how can I take the role "pro" ? 🙂 (i've bought it obviously)

[halcyon coral]

Leave and rejoin via the link at your settings @ perplexity.ai

[serene wigeon]

I think you just put /pro in the url, like https://www.perplexity.ai/pro

[halcyon coral]

Update on input limit to 500 characters:

Everything should work again (with increased latency)!

[rigid minnow]

Awesome 🙂

[chilly forge]

I think there should be definitely an updates channel for things like this.

Definitely not a place for #💬│general chat where important things would be lost completely

Along with that still Perplexity doesn’t face issues quite often but still there should be an updates channel for things like this !

[halcyon coral]

At least for me there is no other channel atm, but I agree with you, there should be an extra status channel for something like this!

[chilly forge]

Hmm same for me, but since you’re a community volunteer you must get one since you’re active as well. Also I have seen many services having it so it becomes quite obvious ig

[jagged ivy]

Wow, I got ideator role, thanks to who has given me this role 🙂

[tight tree]

hello guys!
Is perplexity connected to internet when I search or not?

[chilly forge]

Well Yes ! But don’t believe me just try asking it

[tight tree]

he says he is! but how to prove it 😄

[chilly forge]

Try asking him about the live weather of your location I suppose

[gritty dock]

Ever since the new logo I haven't been able to get this image out of my head whenever I see it.

[chilly forge]

Do you use Advanced Data Analysis ( formerly known as Code Interpreter ) on GPT Plus ?
If Yes then you shouldn’t ditch

[tight tree]

thanks man

wow this is amazing tool

So what is the main difference between having GPT4 pro subscription, and subbing to perplexity?

[halcyon coral]

Sure!
PerplexityAI is, as you probably know, an AI search engine.
If you have a Pro subscription you can choose which model you want to use: GPT-4, GPT-3.5 FT or Claude 2.
Claude 2 offers the most context (100k) and is the same as what you can use with Anthropic. I can't promise you exactly the same results (depends on the system prompt).

ChatGPT Plus is only worthwhile if you use the code interpreter all the time, otherwise PerplexityAI is superior.

[viral kite]

@signal hamlet hi, there is an offer for educators. Will the 25% be valid for every month of purchase if done now, or it is a one-month offer if done till 9th Sept 2023?

[humble swan]

pretty sure it says one billing cycle doesnt it

[viral kite]

ya, but since it is an offer, you don't know

[humble swan]

is it one with SCHOOL25?

[viral kite]

ok

[halcyon coral]

Hey, no need to ping Alex 😉
The discount is a one time discount (one time yearly sub or one time monthly sub)

[viral kite]

but whats 50$ man?

sorry

-50

ok, got it

[wraith grail]

Any plans on releasing an API?

[halcyon coral]

At the moment there are no plans for this, but if that changes it will be announced!

[lament knoll]

Wow, I got super excited when I got the email saying you've implemented Claude into Perplexity.
I have a question though if you don't mind. Is this the Claude 100k model ? If yes, would it work with text files as well ?
Thanks

[signal hamlet]

Hi @lament knoll! Yes, it's 100K and it works with text files. #📢│announcements message

[oblique wedge]

Hello all. Is there a way to avoid, or quickly get rid of, the reference notations inside of perplexity's output? I'm trying to train a chatbot on factual output, but it appears that the constant bracketed references are confusing it.

[wraith crow]

There is currenly no way to disable citations but there is a project made by a community member that can remove those #😎│cool-projects message

[oblique wedge]

I certainly appreciate it.

[tame current]

hi, im wondering what the cap is for queries for GPT-4/claude with pro subscription?

[halcyon coral]

600 uses/24h

[tame current]

perfect, thanks

[short creek]

Also the **token **system is dynamic and self-replenishing. For instance, if you use a token at 10:40 AM on Tuesday, it will be **restored **at 10:40 AM on Wednesday.

[severe moat]

what is the input token limit for claude 2 on perplexity?

[short creek]

2k (8000 characters ~). 100k on File Upload as far as I'm concerned.

[severe moat]

@short creek awesome thank you king

[short creek]

Anyway if someone else could clarify on this matter it would be nice.

[short creek]

It's been recently introduced so I'm not a 100% sure.

[severe moat]

I'll test it out

Yes 2,000 tokens this is badass

[short creek]

Yeah but files can be huge now if using Claude. The 2,000 tokens is there for all the AI Models.

[severe moat]

im a little slow lol

[short creek]

Why haha

[severe moat]

i didn't know every model on perplexity had a 2k limit i've been babying it haha.

im mindblown at 100k in a file

[short creek]

haha

[short creek]

I think it's 100k with both input + output in mind, which is still a huge thing.

As I said I'm still testing it but I tried some big PDF's and it was working great.

[pastel spoke]

/dream a pickel

[carmine linden]

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

[pastel spoke]

#🎨│dream a pickle nickel.

#🎨│dream /a pickle nickel.

[languid dawn]

I have the same issue but with Spanish. It would be great if Perplexity could be configured to search in English but dialogue in other language.

[echo salmon]

I don’t know how to reference replies here. But @halcyon coral has a solution. Set mother tongue in AI profile but prompt in English. You get search in English and results in chosen language. Works like a charm

[halcyon coral]

#💬│general message

[humble swan]

So this Claude thing do you input the file and ask your prompt right away or do you input the file then you can do the prompts

Could I for example input a whole book and ask questions afterwards

[languid dawn]

Yep, thank you, I arrived at the same bypass by myself. The thing is that it would be better to be able to ask in your native tongue, much more efficient as you are proficient in that language. I can imagine a “search location” setting.

[idle zodiac]

Well this discord is gaining some traction ?

[languid dawn]

In other topic, just tried the iOS app and every time I open Perplexity it loads a “create your account”, even if I am already logged in, really annoying! Someone else has this problem?

[wraith crow]

#1145171019523424336 message yes the devs are aware of this bug and will be fixed soon

[oblique wedge]

I freakin' LOVE this tool

[thorn moss]

2k for 8000 words? is this right? i thought 2k tokens is something 1600-1800 words? @halcyon coral

[worn monolith]

The ability to upload huge files is a game changer and will finally let me retire from paying openai monthly.

[short creek]

That information has been extracted from here. It's 8000 characters though, in my last message I said words, my bad.
https://perplexity-ai.notion.site/Perplexity-FAQs-0a9141bdd1b94c76b955e892f73451ff

[grizzled canyon]

yeah perplexity is pretty based compared to other offerings lmao

[tame current]

I don’t like it’s responses compared to ChatGPT’s however I love how it lists websites and can talk about more recent things

And the fact that it doesn’t need my phone number.. wtf GPT?

[severe moat]

i am a perplexity fanboy now.

[oblique wedge]

I get some things DONE with this app! Keep up the good work team.

[chilly forge]

If you buy a PRO sub so you would need eventually put your number in there, so theres that !

[tame current]

Hmm

I wonder why though

Free trial abuse prevention?

[chilly forge]

Yep

[tame current]

I see

At least it’s not required for the free version though

[scarlet zephyr]

So what do we think? Is the Pro subscription really worth it?

[worn monolith]

I have it. And use it often.

[dense kraken]

I'm getting infinite cloudflare :-(

[wraith crow]

please follow these steps: #💬│general message

you can also disable the chrome extension for the time being till this is fixed

[dense kraken]

this didn't help

[dense kraken]

this did

[blissful raptor]

@everyone I have mostly been interacting with the perplexity community on twitter, but it's time to keep this discord community active and engaged. i will be spending time on the Pro Lounge #👑│pro-lounge and #💬│general and #👉│feedback-general. Thanks for all your support and usage of Perpelxity so far in helping us get to where we are and build this product and community!

[tame current]

I’m assuming boosting gives access to the pro lounge

Or is it showing that you have pro with a picture?

I believe you need pro

[halcyon coral]

Pro subscription

[tame current]

You have to join from the special invite in pro settings

Ohh

[golden anvil]

Oooh ceo

[old forum]

I have pro access but have signed up using Gmail

anyway to connect our perplexity acc to discord acc?

[blissful raptor]

easiest thing would be to cancel your pro and create one with the email synced to your discord.

[old forum]

okiee

will do that

[odd mauve]

Hi, if a Perplexity Pro users asks a complex question in the area of personal or relationship advice, can Perplexity when composing its answers only use info from web searches, or can it can answer using mainly the internal "offline" data contained in the GPT-4 model? In other words, imaging it as GPT-4 with browsing plugins enabled. It can search, but does not really do so as it's much more effective (results in a higher-quality answer if it only uses data which are part of the GPT-4 model).
The alternative would be that perplexity only uses data from web search, user profile, prompt, uploaded file and only uses GPT-4 to summarize that data and present it in a nicely structured response, similarly to what Bing does.

[halcyon coral]

Isn't it just about getting the Pro role? So just leave and rejoin with the Pro link?

[blissful raptor]

yes, we allow you to have this experience by entering into 'Writing' focus.

[tame current]

Ok

[rich bane]

Who to write to, to get pro role?

I forgot

[halcyon coral]

You can just leave and rejoin via the Pro link

[rich bane]

Aha

[rich bane]

Thx it worked 😉

[odd mauve]

What exactly does the copilot mean? Does it mean access to GPT-4 (or Claude 2 if the user selects it)?

[ocean lake]

Wassup

[halcyon coral]

Please have a look at the #📘│faq

[wind musk]

yaay the dev wanna talk to people

can you please fix the flickering glitch

[vast pasture]

Hi I use my protonmail both for discord and my perplexity account. What should I do? I do not have access to the server

[chilly forge]

Dm alex or leave server n join from server link which can be found in settings

[tawny mortar]

How effective has perplexity pro been for you as a user while doing coding task as an experienced developer? I know it is good for beginners SDE, but want to gain insights on how it has been for experienced devs out there?

[humble swan]

Discord bot needs more attention 🥹

[warm sigil]

May I check why I'm not able to join pro lounge even following perplexity pro invite link?

[wraith crow]

You'll have to leave the server first and then join with that link in the settings

[jagged ivy]

For some reason it seems to me that Claude 2 uses footnotes to sources less often than GPT 4
For example:
GPT 4: https://www.perplexity.ai/search/7c3ccb73-656d-43ea-8d56-cdcc74d14ab9?s=mn

Claude 2: https://www.perplexity.ai/search/3f56b2bb-dca0-4adc-b91e-b8d5bf04b88c?s=mn

[devout sun]

perplexity.ai/pro?referral_code=6HW74JVG

[solemn trench]

can i use hte academic code for perplexity even tho im already a subscriber? i want to just lock down a cheaper full-year rate

[wraith grail]

So uploading a file + using Writing mode on Perplexity with Claude2 engine is the same as using Claude2 directly?

[wraith grail]

Where can we get an academic code?

[tame current]

I would really have liked if I had subscribed before they released it

[chilly forge]

There was a discount at that time running as well

[tame current]

well, I guess the cost has been paid so I don't have much to say about it now

[chilly forge]

On a side note - you really don’t need a discount haha 👀

[peak iron]

@blissful raptor How does it feels like being so smart?

Sorry to ping like this

[tame current]

claude 2 only uses bullet points to explain stuff? sounds quite basic

[wraith grail]

In my experience claude2 is quite steerable. ask it to explain in two paragraphs and that's what it does. Otherwise, it's not as verbose as GPT4 and might use bullet points to be succinct. Just my anecdotal experience

[lament imp]

I've had a mixed bag when it comes to Claude-2 citing stuff. In some cases it wouldn't cite at all, but in some scenarios it would have 30+ citations in a single response (some duplicated citations).

[clever stratus]

I have an upcoming hackathon in work - one idea is to provide a prompt based search within our docs site to replace our existing (poor) static search feature. Perplexity AI shows great results so it would be great to embed or integrate in our docs page. Is this possible - I tried on #ask and got unreliable answers: #🔎│ask message

Given this is for a hackathon getting something going quickly is important. If the API suggested by the answer linked above exists, is it easy to access and begin developing with? Or how does that work?

[jagged ivy]

I have said in ai profile to do it

[halcyon coral]

Currently PerplexityAI does not provide an API, if this changes it will be announced.

[burnt violet]

yes uwu

[thorn moss]

i stopped my subscription to openai chatgpt now

deym they are really getting worse

the **** men. most of my prompts in chatgpt openai violate their policy. doesnt make any sense subscribing to their platform

[wraith grail]

Their API is still the best tho

[thorn moss]

Yeah I agree. but their API is the only thing that holds them good though

if you dont use code interpreter its done

[wraith grail]

They had the GPT4 model in March. They're many steps ahead of the game imho. They will be the ones releasing the next breakthrough.

[thorn moss]

Thats not my point though. i am not disagreeing with that. i am talking about their CURRENT situation which their platform is getting worst. 😂😂😂

[wraith grail]

agreed

[thorn moss]

cut long story short. Chatgpt for a moment is useless unless your job needs code interpreter. Its a **** good thing perplexity ai exist.

[halcyon coral]

Please do not share inappropriate content @tame current

[chilly grove]

How about that indicator on the right, turn off copilot. I assumed that disabled web search. Also using focus and selecting writing seems to make it less like a search engine also.

[halcyon coral]

Turning Copilot off doesn´t block web search. Only Writing focus does.

[chilly grove]

that's covered in the FAQ pretty well.

[regal frigate]

Why can't other email addresses be used to log in on an Android mobile app, with only Google email login provided?

@halcyon coral

[halcyon coral]

You can login via normal email

[twilit shadow]

Perplexity unperplexes all my life complexities - thank you for such an excellent product and continuous nonstop improvement.

[thorn moss]

Is it possible someday that the input limit will increase?

[blissful raptor]

yep, likely. how much would you like it to be? any example queries?

[halcyon coral]

Maybe similar to ChatGPT or Claude, depending on the context size of the model.

[wraith crow]

And I think it was also suggested in the pro feedback but would be nice if the text would still be too large that it will get added to a .txt file just like claude does.

[blissful raptor]

that's a good suggestion. except at this point, when it goes into file upload mode, we won't search the web, but only answer questions about the file. which may not be what you want if you enter a long prompt that needs the web. if you specifcy your use case more, it will be helpful.

[wraith crow]

For my use case I don’t actually need internet access for large amount of text I put in. I mostly use it to let the ai rewrite large chunks of text or ask questions about large pieces of code.

[blissful raptor]

got it. thanks, we will look into this.

[rain portal]

I find out that the result doesn’t not provide example if ask so and result got back as no information in the web but when I ask gpt 3.5 gives me more example

I didn’t see a much value in subscription so I did cancel the subscription

My feedback is when I want to enter the chat by iOS app , I wait for two screen pop up until I type directly so annoying why ?

Also by default when I go outside the app and come back should I see where I left the message not start over and ask so annoying

The Ui has to be more personalized to my needs not too the community which is not my interest and related or add value to me at all

So overwhelmed when I see the first page too many discussions not related to me and not UI friendly

Be simple like google search bar

[short creek]

Both the web version and the app are currently being developed and improved on a regular basis. If you can provide more specific details about your issue, I'm sure people would be willing to help.

[rain portal]

I did explain up

[short creek]

But I honestly didn't understand what you were saying here.

Can you elaborate?

You are not being able to make the AI give you examples about a certain topic?

[rain portal]

For example I was searching for example of specific discussion to understand more about the topic , it gave me respond by saying no information on the web

[short creek]

That's kind of weird.

May I ask what was your question?

[rain portal]

But when I ask open Ai gpt with the specific terms , it give me context and example

Let me check

[short creek]

Sure.

[rain portal]

What’s your what’s number to share the screenshot

[short creek]

You can share the question over here or just paste the screenshot.

[rain portal]

I can’t paste picture

See the example

It’s just off topic here

[short creek]

You are asking it about the opinion that specific person has on federal increase rate?

[rain portal]

Yes and what kind of example to understand the context

Here another example

[short creek]

Wait

[short creek]

https://www.perplexity.ai/search/394b9eab-19ec-43a4-9f06-c5495c029438?s=c

[rain portal]

Look the example off

I was looking for his opinion as example to understand the context

The tool didn’t understand what example I was referring too ali though in the same log

[short creek]

To make your prompts more effective, I'd recommend that you spend more time crafting them. Typing a simple sentence like "Joseph E. Stiglitz and fed increase rate" may not yield the best results. Both ChatGPT (GPT-3.5) and Perplexity struggled when I typed your prompt.

[rain portal]

Gpt 3.5 is not up date are and live data right ? Up to 2021

[short creek]

That's correct. Perplexity can provide the most recent data via web search though.

[rain portal]

Why the tool expect me to know how to prompt

What a user interface is this

I don’t see product market fit

[short creek]

It doesn't have anything to do with the UI. Every single AI tool has a slight learning curve in order to produce the best responses.

[rain portal]

Yeah but it shouldn’t suggest a better promote or so not just gives me result not looking for

[short creek]

Have you tried configuring your native language in Perplexity so you can ask on it?

[rain portal]

It through me out of the app and search in google

Pointless

That’s why I did cancel my subscription

It’s just waste of time to be honest and no value at all

Good luck in raising money

User is looking for value added

If you can’t create this it’s just pointless what ever the tool is

Just see count the unsubscribe users till date you can measure this kpi for next round of funding

U will understand how added value the tool is

Numbers don’t lie

[rain portal]

I don’t know how to ask in my native language

[tame current]

would you rather use gpt 4 in openai website or in perplexity?

[rain portal]

Open Ai has log relate the query to main subject starting

From user experience much better

Here I come to see check the sources that’s all

I don’t expect more

So why I pay

If I get value more of course I will pay but if the value not match what’s the point u tell me

[tame current]

great! i was thinking of purchasing but i should rethink and research again for the one that's more flexible on the use.

[rain portal]

So hectic the way you want to get most of the topic answered

At least it should suggest me a better promote to measure if I’m happy or not with the result

Otherwise I switch app

[tame current]

does the gpt-4 in OpenAI also able to access the internet like Perplexity? I never bought it

[rain portal]

Try it out

Then cancel anytime

[tame current]

i think before the copilot were added this feature exist

[rain portal]

The flow there much better

[tame current]

well, i don't see it anymore

[rain portal]

Here u keep remind the main topic in your promote otherwise u off topic

Pointless cuz

Waste of money

[short creek]

They don't support native web search. They used to have Bing's web search which was good. Now that's not an option and they only have a third party plugin called webpilot (which is nice but not even close to Perplexity in my opinion)

[rain portal]

They keep add this shits to get more valuations

It’s investor game

It’s so basic , people pay where they see value

That’s why Netflix is losing money now

[tame current]

so if i put some context for the Bot to support their response, i will get much more better result rather than not giving it a context?

sorry if my english is bad, is not my native

[rain portal]

Yes

[short creek]

Yeah that's always the case with any AI tool out there.

[short creek]

No worries, all good 🙂

[rain portal]

As promote engineering called

Fine your promote to better output

But who was the time to learn now this shit of promote 101

So these big amount of raised funding should left this part on users to better perform

I’m signing out of there sorry

[short creek]

Simply use the tools in the language you feel most comfortable with and provide sufficient context by using descriptive questions.

[tame current]

Great! Thank you for the supports!

[short creek]

Feel free to ask any question you have.

[fervent halo]

It's like this with any AI tool, you need to prompt it well for it to give you good results, have you tried putting your native language in the AI profile for example: "I speak Spanish, please expect all questions from me to be in Spanish. I also expect you to respond to all my questions in Spanish"

[hardy halo]

Hi..everyone.. i am build a chatbot using llama-2. but it is taking long-time for generating response. but in preplexity it generate a faster responce how??

[tribal hazel]

@short creek are you expert level in programming

[devout fossil]

there is no responce if touching the voice button

[halcyon coral]

Please create a bug report in #1140622008086970420 according to the guidelines in https://discord.com/channels/1047197230748151888/1140626423707607040

[tame current]

Hi @devout fossil , could you share the android version and what mobile you’re using ? Thanks!

[devout fossil]

sure

please fix this thanks

[tame current]

Thanks , could you share your phone model too ?

[devout fossil]

realme gt neo5

[tame current]

yes

[wraith crow]

Don’t promote that activity in this discord

[tame current]

bro

bro

bri

pipe down

ai integrated with botnet

who would have ever thought

[wraith crow]

please be respectful thanks 🙂

[short creek]

I'm a developer myself, but I still have a lot to learn. I believe there are many people on this server who know much more than I do.

[lone thistle]

Is there a way to upload more then 1 file to perplexity using Claude 2?

[halcyon coral]

Not atm

[eternal saddle]

https://www.perplexity.ai/search/5cd76f68-aa36-42e0-a58c-b2c5dfdee4fc?s=u

Can anyone tell if these are accurate?

I have a feeling the info given isn't accurate. If anyone knows anyone who specializes in diets etc please forward this to them

[karmic yacht]

😀

[lone thistle]

Thank you, and by the way, is there a guide to using perplexity?

[halcyon coral]

There is a FAQ: https://perplexity-ai.notion.site/Perplexity-AI-FAQs-0a9141bdd1b94c76b955e892f73451ff?pvs=4
You can also search for PerplexityAI on YT.

[lone thistle]

Thank you, I've looked at the FAQ, and from what I understand (I jump in the water and paid for the pro) I can only have summarizing when talking the Claude using perplexity right?

[wraith crow]

you can summarize text with both gpt and claude. Claude might be a little better for large text since it has a context size of 100k tokens

[lone thistle]

Thank you, but what I meant was that I only get summarizing as an answer, and I wanted to know if that's how perplexity works? because when I talk to Claude at claude.ai, he answers me in a normal way, gives me suggestions for sentences, or sample paragraphs

[wraith crow]

if you want to have a more chat like experience you can select writing mode at the bottom left of the input field

[lone thistle]

Thank you, I already did that, but I still get answers as a summery

[wraith crow]

could you share the link of that thread?

[wraith crow]

I see what you mean, Claude does seem to like to summarize and give back a list. To fix this you could add “Don’t summarize responses” to you ai profile or tell it at the start a thread

[lone thistle]

Thanks, it worked but the conversation just don't have the same response as Claude at claude.ai, it like talking to Claude 0.5 or something like that, I'm asking for X and the model gives the same answer.

[modest furnace]

Claude is nerfed on Perplexity?

[scenic grail]

when trying to access perplexity i get stuck on "checking if the site is secure " 😦

used to work fine before

switching to chrome works

main browser is thorium and i have the same plugins as I do on chrome so not sure what the problem is

[short creek]

Disable the Perplexity AI Companion extension and it should work.

[scenic grail]

that worked! thanks 😄

[short creek]

No worries 🙂

[stiff solstice]

Is there a reason for the 25 source limit?
I am asking out of complete curiosity. Personally, answers haven't felt the same since that change.

[languid dawn]

Was there previously a higher limit for the number of sources?

[stiff solstice]

There was a time Perplexity was yielding ~40 sources per search and the results were a lot more accurate. Copilot follow-up frequency has also increased significantly since the cap...and it gets frustrating, mainly due to it being seemingly consistently confused about context in general in every conversation I have with it.

I really don't mean to be that guy, but it's blowing my mind more people haven't mentioned it.

[languid dawn]

That’s interesting, how long ago was like that?. Besides it makes sense, one of the best ways I can imagine AI helping us is doing exactly that: reading A LOT of sources to give an answer. Sounds bad that they choose to limit that precise feature, maybe to reduce costs for the GPT-4, integration, idk.

[cursive spruce]

am I right that perplexity never gives you access to pure gpt4? Even when I asked gpt 4 "who are you" it answered "perplexity"

[short creek]

It is using** GPT-4**. The AI model is just not aware of it because of the system prompt. Same thing used to happen in OpenAI's ChatGPT Plus.

[grizzled canyon]

it is using GPT-4 literally

only if you have GPT-4 selected

[cursive spruce]

I do and I am starting to think its not pure GPT-4, its too "dumb"

[short creek]

Try the Writing focus. Searching the internet is a whole different thing.

[grizzled canyon]

when it is fed tokens from a source, sometimes it can get fixated on what it has been given

[cursive spruce]

yea I have it selected

[cursive spruce]

any way to bypass it?

[grizzled canyon]

you can actually tell it to search specific sources

i've asked it to search only wikipedia before, that works

give it a list of sources

[short creek]

Try Writing.

[grizzled canyon]

that too

[cursive spruce]

if I change ai model and come back to old thread will the newly selected model answer or old one?

[short creek]

The selected one will be the one answering.

[grizzled canyon]

haven't tried

[grizzled canyon]

based

[cursive spruce]

thanks @short creek , you are super helpgul!

[short creek]

You can try it quite easily! Try asking something with Perplexity (GPT-3.5) then change to either **Claude2 **or **GPT-4 ** and continue the conversation.

[short creek]

Thanks! Always trying to help 🙂

[short creek]

Also someone said you can have two tabs open as a workaround while they add an easy access to switch the AI Model. You can have one tab with your profile and another one with a thread.

[cursive spruce]

its so funny to show AI my coding problem and getting like 10th time same answer haha

*wrong one 🙂

[short creek]

haha

It depends on what you are asking and the way you are asking it.

[cursive spruce]

great tip! Thanks!

[short creek]

AI can sometimes be picky.

[stiff solstice]

That's all I can think of as well.
I've noticed increased bias and lack of elaboration on simple queries that, before, didn't require an engineered prompt. August 25th was when I first noticed. #⚡│ask-community message
#👉│feedback-general message
All questions regarding this have fallen on deaf ears, unfortunately.
I love Perplexity and am simply curious if anyone might be able to give some insight as to why this might be.

[short creek]

I wonder what the reason is. Maybe someone from the team can tell us more about it.

[grizzled canyon]

Try explaining why you think it is broken, it usually works better the more tokens it has to think with context in mind i.e. code snippet.

[heavy eagle]

Hello

[proven nexus]

I'm a bit confused, does the free version of perplexity use gpt 3 or gpt 3.5? On the internet the articles says gpt 3 while here I read gpt 3.5

[flat agate]

3.5

[tame current]

it used to be 3 but they switched to 3.5

[proven nexus]

Nice to hear

Thanks

[echo kelp]

im suddently hitting a word count for my queries, whereas I didnt have that issue before.

for instance. well a character limit i'm hiting

which wasnt there before

any insight or help would be great

[halcyon coral]

How long is your prompt?

[wraith crow]

the limit for input is 2k tokens each question

that should be around 8k characters

[echo kelp]

i asked perplexity to assess a conversation, the conversation has just over 1000 characters.

i cant input anything more than 400 characters now

maybe I'm using threads wrong thats why i hit a limit? i have created through using perplexity alot of separate threads

[wraith crow]

do you mean words or characters because i can’t recreate this issue.

[echo kelp]

one sec, i'll share a screenshot

1000 characters.

150 words

any idea why this is happening? thanks

[wraith crow]

this is the input to start a new thread right?

[echo kelp]

yes, ive never hit such a word count before

[wraith crow]

do you have some userscript enabled for perplexity since I also see an 8 on the send button

[echo kelp]

thats the languagetool extension

just disabled it, but no luck

[wraith crow]

could you try to refresh the page or try it on a private tab without extensions and check if you’re still facing this problem

[echo kelp]

just refresh, and opened in private

still no luck

[halcyon coral]

Do you have any content blockers, perhaps already integrated in the browser?
Which browser do you use? Do you have the same problem with other browsers?

[echo kelp]

this is with avast secure browser

i'll try general chrome now

[wraith crow]

it seems something is messing with the character counter in your browser since it also says NaN

[echo kelp]

yeah, still issues with chrome, maybe its an extension, let me disabled one by one and i'll see what it is

ok thats odd, i opened brave browser and logged into perplexity pro and now it works. so something in my chrome and avast secure browser has sudddenly stopped me from querying

one sec restarting computer lmao

[wraith crow]

even when you disabled all extensions on chrome it still didn’t work there?

[echo kelp]

yup

im gonna restart, rule out my pc throwing a random fit lol

ok restarted pc, all extentions off on chrome and avast secure browser which is basically chrome, still issues

but brave browser works.

😕

tested in edge which i never use

and has no extensions added

any insight into this weird issue would be much appreciated

[stone socket]

I just have the same problem

tried to switch between GPT4 and Claude 2 models and same problem, doing it on firefox

[outer glen]

how do you switch betwen the 2 models?

[stone socket]

with perplexity pro you click on your name on the bottom left and there

[outer glen]

ooh its a pro feature

that makes sense

Is there a perplexity - ai companion extension for firefox?

[eternal kestrel]

hi

[echo kelp]

oh really, did it just start happening for you aswell?

wait, its now working as normal

ok im so confused haha

in chrome and avast secure browser

[stone socket]

same it's working again in firefox

[tame current]

yalls getting a perplexity outage about now?

[echo kelp]

i did earlier, but seems ok now

[old forum]

https://www.perplexity.ai/search/9b531163-3c1d-4f75-a083-bbc2ff693be0

If anyone wants to have context of what Elon Musk tweeted today 😊

[tame current]

I thought that already had a different term

[orchid relic]

Anyone recently taken advantage of the pro option yet?

Im blowing racking up open ai bills, and still pay the monthly fee. But need to think smart

I'm finding the response times on perplexity unreliable occasionally failing.
Is this normal?

[pliant thicket]

Any admin online?

[halcyon coral]

What is it about?

[pliant thicket]

I am a security researcher and I have to report a security bug

Would like to contact the CTO

[tame current]

Perplexity has a lot of bugs tbh. The TOS is also relatively easy to bypass

Tbf it was founded in August 2022 seemingly

[pliant thicket]

I am referring to security bugs

[halcyon coral]

Got the msg?

[pliant thicket]

Which one?

[halcyon coral]

DM

[pliant thicket]

Check

[tame current]

Ah, that wasn’t a reply to your message. I should’ve worded that better, sorry

[brazen copper]

Hi @pliant thicket, did you change the max of characters for prompt?
Bcz on my phone it works as worked before...
And I noticed that I having problems with that only in one google chrome profile, how to fix it?

[stone socket]

Same, had this bug yesterday, went away, then now it's back again

[brazen copper]

Try to wipe data at your Chrome regarding this website, I tried it and it helps

[stone socket]

I tried it on Firefox still same problem :(

ok I smashed the button and after 10 times it works for some reason

[hazy abyss]

Can we describe the core of what Perplexity does as Retrieval Augmented Generation (RAG)?

1. A retriever model searches through data sources and retrieves relevant context passages for the given query: documents, paragraphs, or text snippets.

2. The retrieved context passages are provided as input to a generator model (LLM).

3. The generator model uses the retrieved context to inform its generated text output, producing a response that is better grounded in data sources.

[still shore]

Has the maximum characters allowed in the prompt been reduced? I cannot insert my regular prompts anymore. Get this:

[halcyon coral]

Temporary limitation only, please be patient 🙌

[still shore]

Ok thanks!

[halcyon coral]

Should work again 🆙

[still shore]

Thanks for the update. Unfortunately I am still unable to send the request. When clicking the submit button nothing happens

[halcyon coral]

#🌐│feedback-web message

[halcyon coral]

You refreshed the website?

[still shore]

yep, restarted the browser as well

[halcyon coral]

Did this? #💬│general message

[still shore]

Still not working. If I send in a shorter prompt it works, longer prompts I cannot submit

[halcyon coral]

Just to confirm: You deleted the cookies, logged in again and still have the same problem?

[still shore]

Correct, cleared the cookies in chrome by navigating to inspect element. Was logged out as a result. Logged in again and still cannot submit the prompt

[halcyon coral]

I can confirm the limitation. I have tried it with another account 🤔

[still shore]

after some loading time this showed again

[halcyon coral]

You have Pro?

[still shore]

yep

[halcyon coral]

The limitation should only be temporary. What still surprises me is that some accounts have a limit and some don't.
Sorry for the inconvenience, it should work again soon

[still shore]

Thanks, works now 🙂

[halcyon coral]

Btw you can get the pro role by leaving the server and joining via the pro link in your settings!

[stone socket]

it still doesn't work for me :(

[halcyon coral]

You have already tried the troubleshooting steps?
Then unfortunately only waiting helps.
Perplexity should be available for you again soon 🙏

[stone socket]

yes I'm going to wait ^^

[signal hamlet]

Sorry for the inconvenience! We'll fix it soon.

[stone socket]

That's ok it happens dw

[signal hamlet]

Should be fixed now, @stone socket! Could you please check.

[stone socket]

Yes it's working thx !

[orchid magnet]

what is co pilot

[halcyon coral]

Please have a look at #📘│faq

[orchid magnet]

thank you

also last question, can i ask perplexity about questions regarding how to use it

[halcyon coral]

This does not always work, in any case your question must contain "PerplexityAI".

[orchid magnet]

okay, thank you

this is an awesome ai tool, ive always just wanted a big search engine on steroids and i was trying to use chat gpt for it

[bright ruin]

Hi there, Just purchased pro. Is there a way to disable perplexity constantly returning search restults when I am just asking it simple questions? I want it to function just how Claude normally would

[wraith crow]

You can use writing mode if you want a more chat like experience

[flat agate]

You have to use writing mode to disable internet access

[bright ruin]

Where do I switch to that

[wraith crow]

bottom left of the input field

[bright ruin]

I dont think I want to disable internet access, I just dont need it to search for youtube videos and everything every time

I dont see the option in the bottom left of the input field

[wraith crow]

click on the word focus and you'll see a popup with all the modes

[languid dawn]

Where it says "focus"

[bright ruin]

Ah okay. Is taht only available when you first start a thread?

[flat agate]

Yes

[bright ruin]

I see

thanks

So if i enable that - will it function like how Claude normally would?
I cant use claude because I am not in the US or UK

[flat agate]

Yo have to activate Claude model on your settings

[bright ruin]

I did do that already yes

I dont see a file upload option in the chat box

Can you only upload a file at the start of a chat?

[tame current]

That would be weird

[flat agate]

Yes, by now

[bright ruin]

Thanks I have Claude enabled.
I started the chat. Now I want to upload a file for it

But I dont see the option to add a file. It was only there at the start of the chat

No file upload option

[flat agate]

you only can upload on the first message by now

[bright ruin]

waaaaaaht

sigh

[tame current]

D:

[bright ruin]

Can I get a refund issued then for the $20 I just spent

[tame current]

I’d probably get a refund too

[bright ruin]

I need a chat where I can upload files ongoing to train the discussion around a particular subject

I am better off pourchasing a VPN and using Claude

[wraith crow]

you can explain your issue and ask for a refund by sending a mail to : [e-mail]

[bright ruin]

How come you cant upload files ongoing in a discussion?

[halcyon coral]

The devs are working on improvements

[wraith crow]

The file upload feature hasn't been here for that long. This functionality has been suggested by multiple people and will probably be implemented in the future.

[bright ruin]

It seems a bit misleaing to promote perplexity as having the abilities of Claude 2, when it doesnt

Thats good

[scarlet zephyr]

How often will be the GPT3.5 be fine-tuned for Perplexity Pro going forward?

(Assuming you will keep improving it?)

[wind mural]

hello

peeps

[onyx zenith]

When generating text with PerplexityAI, under which license do the resulting texts fall? Are they eligible for open-source publication or commercial use?

[flat agate]

https://www.perplexity.ai/tos

[serene wigeon]

I have a question, what is it in how perplexity is designed that makes threads often devolve/lock into it repeating previous replies forever regardless of what you say to tell it it's being wrong or incomplete?

[tame current]

I wonder if Perplexity has a variety of responses for one prompt that’s asked repeatedly but has consistent wording

Nope guess not

[onyx zenith]

So I do get that right and I'm not allowed to "republish" generated textes:

Prohibited data collection includes, but is not limited to, using the Service as input into other services, websites, or databases. You agree not to republish any content generated by the Service without clearly citing the Service as well as the context associated with the content. Misrepresentation of the source of the content or the nature of its creation is strictly prohibited.

And

As a user of the Service, you are granted access for your own personal, non-commercial use only.

From the TOS means I'm not allowed to use any KI-Answers in business context?

So thats a main diff to OpenAI TOS?

Input and Output are collectively “Content.” As between the parties and to the extent permitted by applicable law, you own all Input. Subject to your compliance with these Terms, OpenAI hereby assigns to you all its right, title and interest in and to Output. This means you can use Content for any purpose, including commercial purposes such as sale or publication, if you comply with these Terms.
From https://openai.com/policies/terms-of-use

[tame current]

Are you German?

[tame current]

he said KI-Answers so probably is :))

Found the german (I'm german myself)

[tame current]

I don’t think it happens quite as often with ChatGPT

It happens to extremely often with Perplexity. The free version at least

I thought ChatGPT was mostly trained without actual user input

Oh wait, I’m dumb

User input can be a bit troublesome I imagine

[onyx zenith]

I am, yes 🙂

[modest furnace]

https://www.perplexity.ai/search/67c3797c-1c55-4546-a0b8-f7f4ab43f7fb?s=m

Can anyone explain how it makes this mistake about the longest bridge in Europe? It knows the lenght so it shouldn't list the longest as the second longest...

[short creek]

You are not asking "What is the longest brigde in Europe?"

As you can see if you specifically ask for the longest one, it will answer with the expected result.

[halcyon coral]

Nevertheless, this does not make sense 😉 :

The Vasco da Gama Bridge in Portugal is the longest bridge in Europe, spanning 17,185 meters (56,381 feet). The Crimean Bridge or Kerch Strait Bridge in Russia is the second longest, spanning 18.1 kilometers (11.2 miles).

[short creek]

Oh wait it says second longest in the actual response

I didn't realize about that

My bad

[tame current]

whats up with the speed? :))

[safe cosmos]

Nothing is up atm, how bout you?? Enjoying ur day?

[tame current]

same here

[sage lance]

Would you recommend chatgpt plus or perplexity for academic questions like homework related

[zenith atlas]

I use perplexity for finding sources and summarising them, as it does it way better then chatgpt

[sage lance]

Hmmm

Like web articles and stuff?

[zenith atlas]

yes, academic articles, news, magazines and other stuff

[blazing dust]

claude crushes

[tame current]

Claudes got web search?

[blazing dust]

yeah, on perplexity at least

[tame current]

oh gotcha

[blazing dust]

it does seem to respond weird with copilot sometimes, like it just gives me the answer phrased as a second question lol. but this is my first week using it the model seems to work really well

[tame current]

Copilot is kinda useless isnt it?

[blazing dust]

i find it beneficial but only in specific cases.

sometimes for more recent events or when asking a question i might not fully understand my self

but overall yeah often makes a worse answer than non-copilot

[buoyant bough]

Hey. What AI Model do you use?

[halcyon coral]

GPT-3.5/GPT-4/Claude 2

[tame current]

Perplexity if you’d like to see stuff like websites or sources

I feel like ChatGPT plus (even without plus) would probably have better responses (Tbf perplexity is relatively new and ChatGPT has more users and possibly funding aswell). I’d probably choose Perplexity for academic purposes depending on which classes you have

[plush shadow]

Hello everyone, it seems that there is a IOS synching issues with Perplexity. Does anyone else can't see their IOS threads on the web??

[supple ridge]

Yep

But illl check again in case

[weak oriole]

Is their an API for Perplexity?

[carmine linden]

Hey @weak oriole! Perplexity AI does not currently provide an API.

[frank rivet]

Hi! are you still having this issue?

[agile panther]

Hey guys, it seems I am stuck at 'Checking if the site connection is secure' screen. Any tips? It works in incognito but I don't wanna delete my cookies for this only.

[devout fossil]

why this shit looping forever...

[exotic cloak]

Hey guys, which ai model is better to use on Perplexity?

[grizzled canyon]

perplexity = fine tuned gpt3.5, gpt4 is gpt4 and claude is claude... it depends on use case, i find it's best to try a mixture of them and go through the flow of conversation to get the best of all worlds but i find gpt4 is better to facts and technical information and claude is better for synthesising more approachable outputs

the fine tune gpt3.5 is really good actually

very good

[supple ridge]

@tame current where do you see an animated perplexity logo?

[supple ridge]

[grizzled canyon]

true

[keen walrus]

will perplexity be adding an extension for firefox users?

[halcyon coral]

Please go to https://labs.perplexity.ai and verify yourself there via Cloudflare.

[keen walrus]

i've noticed their cloudflare setup isn't the greatest. for me, i open up my browser every morning and enter a query into perplexity, but it ends up just stuck on loading the response, and refreshing doesn't fix it. i have to do ctrl+i, enter a query, and THEN refresh to get the cloudflare "not a bot" box to popup so i can check it. its too aggressive against vpns

and kinda glitchy

[devout fossil]

thanks for your replying, I will try the method you provided

[devout fossil]

thanks I'll try this

[urban nacelle]

I also want a refund for this... Completely useless without this feature.

[jagged ivy]

#💬│general message

[mental light]

even this has been improved, still not good experience ..

[idle zodiac]

What happened to the annual plans ?

[mental light]

how can i login with email in android app ..

lmao, same issue with phind, but it is unfortunate that perplexity raised another webview in app while phind just use PWA ...

[halcyon coral]

Don't you have an option: "Continue with email"?

[mental light]

yes, then it sends a link to the email, then how can i use the link to login in the app

good question

[halcyon coral]

You open the link that should open automatically with the app.

[mental light]

it didnt, does it work for you? if so, let me checkout the problem or use another browser

[wraith crow]

could you try copying the link and opening it in a private browser tab

[mental light]

download an edge and it works, thx

[unborn minnow]

guys, do you find yourself using the 'threads' feature? I almost never use it and create a new thread each time I have a new question. Is it suggested to be used as if managing tabs in a browser or does it make no difference as such and is a matter of personal preference?

In say, ChatGPT and even in Bard, I do proper maintenance of threads and delete the irrelevant or older ones. But I suppose, Perplexity is to be used more as a search engine, with threads seen as 'search history'?

[sharp wagon]

I often select an old relevant thread, and then continue that conversation.

as in, a few days later I think about that topic again, but I have new questions, and so I can continue giving perplexity starting context

[unborn minnow]

I see, that does make sense.
I think a part of the reason for my question is because the UI in ChatGPT or Bard presents the threads in a much more outright fashion where it becomes quite an issue if they're not maintained. Whereas in Perplexity, they reside under the hood without causing much trouble.

[tame current]

I pretty much never revisit threads

[unborn minnow]

Interesting, I think this question might be worth asking those who were involved in the design of it, to see how they intended it to be used

[mystic dune]

I usually delete them outright (if there is no need to return to the answer later), but I'm really waiting on the 'rewrite prompt' (uuuh idk how it's called) feature like in chatGPT

Bcs in chatGPT I have some old threads with some 'base' information that I use from time to time (for example SQL tables) and I can just write one question on their usage and get an answer, and instead of generating a new thread copy pasting the same info for a different question, I can just rewrite my initial question in the same thread and switch between different 'sub-threads' and view all the answers and it's veeery helpful 4 me

I think something like that is in the works rn, that'll make threads extra useful
(Yeah deleting messages in thread in Perplexity is useful but extra branching would be hella cool)

[unborn minnow]

Very well said
I think this is something which all succesful AI chat engines will have in the near future. I guess many users probably won't know they needed this feature until they have a go at it. Afterall, we're all used to chatting with humans in a linear format.

Is there a status board to see what they're working on?

[mystic dune]

Idk, but ppl from perplexity team reply here from time to time and sometimes say that some asked for features are already in the works

[unborn minnow]

Oh i c, not the most efficient strategy to go about it I'd say, but good to know.

[mystic dune]

Hm tried searching around but there doesn't seem to be a roadmap (or I'm just blind), oh well guess we'll get cool stuff along the way

[cunning leaf]

Isnt it the same as Edit Query?

[mystic dune]

It's a bit weird - it doesn't make a new 'sub-thread' like in chatGPT, it edits and re-answers one specific prompt in a thread without changing subsequent thread conversation

When I have something like
Prompt 1
Prompt 2
Prompt 3

And I change 'Prompt 2' to 'Prompt 4' the whole thread after this change stays the same and I can't make a whole new convo without deleting stuff

I don't get it, but it looks like more of a feature to edit your latest prompt to continue the convo with the latest change, but not when you want to start a whole new sub-convo line in chatGPT

[cunning leaf]

Ohhhh I get it now

[devout fossil]

it would be better if there are code high Light in LLaMa Chat

like this

and it would be better if there is a button to copy the users input for better prompt modification

[tame current]

hello

i upgraded my perplexity to pro, where can i get the role?

[flat agate]

You have to exit discord server and enter from your pro account discord icon

[tame current]

ok thank you

i didn't get the role 😦

[halcyon coral]

@tame current Please leave the server and join via the button in your settings

[tame current]

alright thank you i got the role!

[halcyon coral]

No problem, that's what I'm here for

[tame current]

cant help me with my c2

😦

[tame current]

Hello

[brazen copper]

@pliant thicket faced this bug again and wipe data at my Chrome regarding this website didn't help...
#💬│general message

[pliant thicket]

Hey man, There must be a misunderstanding 😅 . I did not face the bug you faced. I just reported a security bug to the team.

[brazen copper]

I forgot to add "I")))

[halcyon coral]

I think he meant why you pinged him 😉

Hey @tame current,
please watch your language!

[karmic river]

Good evening,
I am experiencing problems with the operation of perplexity with both claude and gpt 4.
I type a prompt and get no response, expiring in endless loading.
some solutions?

[wraith crow]

Could you refresh the page. You probably need to verify with cloudflare again

[karmic river]

Yeah, now working. Thanks

[lofty dagger]

heyy

i made a offline autogppt

foss gpt4

p.s i coded so much my skin fell of

i dont think im human

new skin is genearting like gpt4

wiggle

[karmic river]

Any way to copy without the mentions numbers?

[wraith crow]

not possible atm but there is this community member that made a website to remove those citations #😎│cool-projects message

[karmic river]

Nice work ahahah

[mystic kelp]

Is Perplexity better for writing? Because he summarizes the ranking page on SERP and wraps it up.

[blissful raptor]

@here Thanks for the support and feedback so far! In order to streamline them, we're collecting responses here. Would really appreciate if you took a few minutes to complete this survey! Will read through the responses and help improve the product for all of you! https://perplexity.typeform.com/to/Q9LAun1b

[ancient geyser]

where is the pro chat?

[blissful raptor]

#👑│pro-lounge

[tame current]

[ancient geyser]

it says no access for me

[tame current]

🤔

[wraith crow]

If you bought pro you should leave the server and rejoin with the invite link in the pro settings.

[humble swan]

completed the survey

[hollow obsidian]

I still wait that you make a Pro for people with lesser budget... but you ignore my words... you prefer to get NO MONEY instead of offering a "lite pro"...

And no I don't mean 1:1 features for a lower amount.
I mean a few Pro features.

[vale grail]

filled the form, some questions shouldn't be "choose as many as you like" lol

[humble swan]

Would love if the discord bot would get some more attention, mainly #🤖│feedback-bot

[vale grail]

I have the bot in my server but it only provides a link to the perplexity website (with the answer ofc)

maybe it's a permission issue but that's what I'm experiencing

also, the primary reason I am in this server is to be alerted with new updates 💀

[lament imp]

I enjoy how this reads

It sounds very righteous

But I do think it's a good idea

Like a $5 option for 100 Copilot + GPT-4 or something

$20 contains everything else

[fluid mauve]

and ,and Pro Lite, as @hollow obsidian @lament imp says would be a good idea to monetize the product more and pick up more spondoolies from the lower branches. I don't think 'Pro Lite' should get access to Pro-Lounge . I'm saying all ths as a Pro subscriber who rarely used Pro Lounge. Why? I don't really know.

[hollow obsidian]

yes...
I spent every month money for spotify, mobile games, disney+, Amazon prime, Netflix.
I can not also add 20$ on top for a Web Search and AI writer.
But a 5$ little premium version would I SUPPORT.

currently I only USE it, but not support it.

[hollow obsidian]

I don't need your VIP lounge where you all smoke cigars and drink whiskey.

https://tenor.com/bpQqD.gif

[fluid mauve]

LOL. I drink Wine and pass on the fumar

[hollow obsidian]

I'm happy with the poor folk, with the People who are "common" or casual. In German we say "bodenständig"

People who didn't got lifted up in high standards just because they are in the "upper class".

[fluid mauve]

Before this descends into a flamewar, let's all just agree that people at different periods of their lives have differing monetary goals. what I spend my hard-earned cash on is probably different from a lot of people.

[hollow obsidian]

but you want to be "alone" with your rich kids, and thats fine for me 😄

I don't want to be at the smoking/suits wearing snobs.

[lunar anchor]

LOL

[fluid mauve]

I chose to have Perplexity as it amuses me and I find it incredibly powerful and useful. I don't have Spotify, mobile games, Disney+, Amazon Prime, Netflix. As I said what we spend on is different.

[hollow obsidian]

yeah, I love my entertainment :3

[fluid mauve]

I have plex. And this is entertainment

[stone socket]

It's doing the "long long" entry again :(

[signal hamlet]

Hey @stone socket, do you mean the entered text is not sent with a "too long" error?

[half escarp]

Hey, a little late to the party, but I think it would be helpful, esp. if you're using the Claude model, to upload unlimited files like Claude lets you do, so long as it's within the 10 MB/file limitations.

[signal hamlet]

Hey @half escarp! The file upload limit is 25 MB/file.

[dense kraken]

Just to be anti-social here, I always give the lowest response whenever there is a "would you recommend" question. The reason being not me disliking the product, but I don't recommend stuff to people in general.

[tame current]

Hey everyone

How are we doing?

[grizzled canyon]

But maybe it's a 5 because you would if the opportunity arose?

[orchid magnet]

How can I get the most out of perplexity

[urban glade]

ayo

sussy

among us bals

https://tenor.com/view/sad-depressed-dog-gif-25732965

[tame current]

You don't

[sharp shell]

Hi, I am new to perplexity. When I edit a prompt and save it, I see the new response and but the previous response disappears. Anyway to see this like in chatGPT where we use < 1 / n >

[tame current]

between perplexity and fastgpt, which do you think is better ?

(ignoring the fact that perplexity has follow up questions)

[winter wedge]

@latent wharf

[carmine linden]

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

[sand goblet]

flutter

[grizzled canyon]

not yet, or we wont ever, may be a good suggestion

[ember umbra]

The best viagra for 75 years old

[tame current]

😎 https://www.perplexity.ai/search/fb7fb789-febe-483a-aae2-1dc3ae6fae11?s=c

[halcyon coral]

@tame current Please do not send empty messages, thank you!

[steel escarp]

hi ppl , i have a question regards Claude 2 on pro version. as i based in EU if i subscribed to pro can i use Claude 2 ?

[magic hazel]

Hello !

[halcyon coral]

Yes, you can use Claude 2.

[steel escarp]

thanx

[magic hazel]

Is there any vocal room here ?

[halcyon coral]

There is only one for Pro subscribers.

[magic hazel]

ooh okay

[noble ravine]

Could someone quickly explain to me exactly what Copilot does? I've been turning it off recently because it's faster and it still searches the Internet, even without Copilot. And often, Copilot asks me some irrelevant question before I get the answer, slowing down the process. So what does Copilot do exactly?

[halcyon coral]

Please have a look at the #📘│faq

[noble ravine]

Thanks! I think I'll keep Copilot turned off, for now. I'll give it a go a bit later to see if it's improved.

[chrome cape]

Hi, is the context window limited? I cannot insert anything more than before in UI and also it does not give me warning that length was exceeded... File upload also does not work. Looks like UI is stuck. Ditched chat gpt plus for coding with browsing access but I this point I am not sure if I will continue

[chilly forge]

Wise decision !

[river quest]

has anyone tried running Perplexity on Webcatalog?

[stone socket]

yes

[chilly forge]

Didnt tried fastgpt that much but I can say Perplexity with the old Copilot (GPT4) was quite ahead from most of its competitors cant say the same now

[plush shadow]

Hello everyone how can we resynch all Perplexity Threads across all devices??

[tame current]

"conversational answer engine" tf does that even mean? answer is a component of the conversation, not a separate thing also "where knowledge begins" is straight up lame though you even trademarked it 😭

[thorn moss]

Oh my, is perplexity also affected in openais content policy like when doing prompts that violates their policy it will not allow you? my prompts that was working yesterday isnt working now

[tame current]

prolly yeah since they use openai llms

[thorn moss]

oh thank God it worked. i thought perplexity is also affected with openai's nonesense. That is why it triggered me to unsubscribe to them. they are too much. Their platform is useless now actually. it got worse because you cant do your prompt thank God Perplexity 🙏😢

We can't deny their GPT 4 API is still the best even nerfed so much but deym their chatgpt platform is really dumb to the highest level now. the only lacking of perplexity probably is plugins and code interpreter and openai is done. also max input limit

I know Dyno Bot bot. 😂😂😂

[grizzled canyon]

It's not nerfed

[thorn moss]

if not nerfed why so many people complaining gpt 4 replies dumb and too many comparison from the last 3 months?

so its still the same? 🤭

[grizzled canyon]

When the model gets fine tuned the best way to get the desired response is to prompt in the way which it was trained

When a model gets censored it does love some creativity however

It isn't "nerfed" exactly but it has more guard rails just that the ai "enthusiasts" see it as black and white

[grizzled canyon]

Unsubstantiated whinging from those people

[thorn moss]

zeeeeemn. that means we have to work hard to get the best answer from AI right now?

[grizzled canyon]

rule of thumb is the more tokens and the higher quality prompting and priming information you provide it, the better it is able to perform

[thorn moss]

zemmmmmnnnnn. i do prompting but i have noticed its a challenge really to have good answers. work harder i guess

[grizzled canyon]

yeah what you put in is what you get out ...

[upbeat dawn]

i subs for pro, is there something or role i can access here ? @grand mauve

[arctic spade]

3.5 ft is getting good👀

[halcyon coral]

You can get the Pro role by leaving the server and rejoining via the pro invite link in your settings @ perplexity.ai.

[upbeat dawn]

okay

[thorn moss]

perplexity down?

doesnt provide answers, loads slows and provides blank answer

ah ok. only in android. but web version works

[grizzled canyon]

yeah it's better if you just front load the concepts and make some brief notes then expand on it usng perplexity

use obsidian MD

[halcyon coral]

@slow karma Please no advertising, thanks!

[waxen imp]

Can someone provide me with some best practices for using Perplexity?

for now I think it's slightly diff from ChatGPT with searching tools, but what's the main point?

[thorn moss]

perplexity phone app still not functioning

[wraith crow]

what is not functioning in the app?

[thorn moss]

when i ask something while generating answers it doesnt show up the words and just give references below but totally no worda

tried reinstalling but no use btw

only in web version works

its just doing that

[wraith crow]

That looks like a bug. Could you click the share icon in the top right and share the links to the thread where this occurs in #1140622008086970420 this way the developers can better check where the problem is coming from

[thorn moss]

ok wait

done

[wraith crow]

Could you create a new thread in #1140622008086970420 with the links and pictures you just send. Make sure to give it the android tag.

[thorn moss]

ok wait

wait im figuring out how to tag in android cause there seems to be no option for it. and picture insert

[wraith crow]

at the bottom of the input field you'll see the tags

and top right you can add images

[thorn moss]

i mean how can i share the thread if i posr

if i share the thread there is not tag option

if i manual post where is the thread?

[wraith crow]

OOh just copy the link and paste in inside the thread you're making. Don't share it with discord with the share button that will force you to instantly send it.

[thorn moss]

ah ok sorry i dont know

done

it seems like if you open the thread it will look normal if you use web version that makes it look nothings wrong but if you open it on android it doesnt show

[wraith crow]

Thanks for sharing the bug and the links

[thorn moss]

welcome

[gaunt moat]

Hi, why was Wikipedia focus removed?

[wraith crow]

Removed due to low usage #🤖│android-beta-feedback message

If you want similar results you can still use the site parameter:

question here site:wikipedia.com

[gaunt moat]

I understand Well, I don't think focuses like Reddit and YouTube are used very much, so why aren't they removed?

[chilly forge]

reddit ? are you sure xd

[tame current]

reddit is probably used the most after writing

[gaunt moat]

I'm probably the only one not using the Reddit option. 😂

[gusty blaze]

I find it useful when I want to collect user reviews and impressions of products and services, or people’s opinions on topics.

Sometimes I instinctively go like “what are people on Reddit saying about xyz”
Achieves a similar effect to the focus mode

[scarlet zephyr]

Is there a limit to input per query in perplexity?

20000 characters?

[halcyon coral]

Hey, yes there is a limit: 2000 tokens, ~8000 characters.

[chilly grove]

You use Perplexity for what makes it unique. It provides web access rather than isolation. You get authentic perspectives, not just SEO-fueled content. Perplexity utilizes current research, not speculation. In short, you leverage AI to fuse innovation with verified information. You appreciate that Perplexity is verifiable - you take time to validate its responses and remove inaccuracies. You want creative ideas grounded in facts.

[balmy dew]

Is there a way to use perplexity without the internet research? I.e. like the classic gpt4?

[wraith crow]

You can set the mode to Writing in the bottom left corner of the input field

[arctic spade]

the website doesnt work for me

the extension does though

[tame current]

What about other models and pro users?

[halcyon coral]

The input limit for text remains the same.

[tame current]

@halcyon coral even for Claude 2 and gpt-4??

[arctic spade]

guys why is the website not loading for me

[halcyon coral]

Do you have problems with cloudflare?

[halcyon coral]

Yes

[tame current]

8000 chars???

Wtf

[arctic spade]

it's probably oauth that's the issue

[tame current]

Phind offers 16k

[arctic spade]

i can login through the email link but not oauth

[tame current]

Perplexity team is quite stingy it seems

[arctic spade]

phind 30 per month

12k not 16

[zenith atlas]

for now you can add extra text with file upload in your first prompt for thread

[tame current]

30 queries per month?

[arctic spade]

no 30$ per month

pplx 20 if you didnt use the discount

[tame current]

Still expensive