💬│general

[meager sparrow]

In a screen dumbass

[minor shard]

Maybe I can

[vocal owl]

bros got satellite imagery

[meager sparrow]

I could be laughing right now you don’t even know. Dude assuming random stuff for no reason.. LMAO

[minor shard]

Anything is possible these days who knows 😉

[vocal owl]

watch me hit this insane 360

[minor shard]

@south kindle did u get filter/glasses? I never realized how much light it blocks

Sounds stupid I know lol but it’s pretty cool how they work as simple as it is

[agile jay]

Polarized glass?

[south kindle]

I do have a pair from a long time ago that I'm still looking for as we speak.

[minor shard]

It’s like plastic

[agile jay]

I'm assuming they just make it like that because it's cheaper.

How polarized are they?

[south kindle]

they are similar to the ones you get after the eye doctor, i think at least.

[minor shard]

💯

[minor shard]

Can literally only see the sun. Idk about %

[agile jay]

So very high then

[minor shard]

Yeah

I was like against this whole thing and then I realize I could just look up at my house and I was like ehhh why not

[south kindle]

this is from Steve Lookner's channel doing coverage: https://www.youtube.com/watch?v=gCkZmADecL0

[agile jay]

Apparently they are 99.99968% polarized...

[fast meteor]

https://www.bhphotovideo.com/explora/photography/buying-guide/lens-filters-for-solar-photography
btw here is the guide for photography

[south kindle]

Sidenote: polarized glasses and fishing apparently are a good combo.

[agile jay]

Yep, because who knew water was reflective?

[south kindle]

haha

seems like cheating though.

texas is getting it's first glimpse now.

[agile jay]

You can only look at it without glasses when it's fully covered.

Not sure how long that lasts though.

Also is it a total solar eclipse?

Or only partial?

Also if you have superpowers, best to not get into any fights during it.

[south kindle]

https://www.perplexity.ai/search/The-Great-North-Xamu7TDvQziozOTyGYCOkw

that's poster worthy. Nasa should be selling these. ha.

[minor shard]

Total if you are in the path I believe

Or like 99%

[agile jay]

Rip, that we don't have many moons.

Saturn and Jupiter have over a hundred each...

[south kindle]

Saturns moons are WILD too. Enceladus being probably the most interesting moon in the solar system.

[agile jay]

The downside to Saturn's moons is that they are on the same plain.

So you don't get a good view of the rings.

[south kindle]

water gysers that reach into outer space for cassini to do samples of.

[agile jay]

Yep, and probably what makes up a lot of the rings.

[south kindle]

I actually learned how that works recently via the PBS/Nova series.

Moons getting tildally locked and then ripped apart as the orbit closer. it's own gravity and saturns gravity pull the moons apart, creating the rings that surround the planet.

For Saturn at least.

[agile jay]

All the gas giants have rings.

It's just they aren't advertised as much.

So Saturn, Jupiter, Uranus and Neptune.

[south kindle]

The rings of saturn are kind of the "crown jewel" of the solar system. so they get most of the attention.

[agile jay]

Earth probably had rings too at one point.

When we crashed into the moon.

[meager sparrow]

When did we ever crash into the moon??

All life on earth would probably be extinct if that was the case…

[agile jay]

The moon is part of a planet that crashed with earth a long ass time ago.

And we know it happened because of geological records on both bodies.

[meager sparrow]

How? If the moon crashed into the earth. Then we probably won’t have a moon to begin with..

[agile jay]

We didn't have a moon, until it crashed into us...

A lot of collisions happen in the early days of the solar system.

[south kindle]

I believe that was part of Jupiters creation as well. due to the gas disk around the Sun, it migrated inwards towards the sun and the celestial planets, throwing everything in it's way inwards. Saturns gravitational pull is what brought Jupiter back. at least in the current theories of the solar system creation.

[agile jay]

One of the key points of what makes a planet is that it's cleared other stuff within its orbit.

[meager sparrow]

Dude.. you’re not answering my question.. if the moon crashed into the earth. Then why do we have a floating moon that rotates around the earth? Wouldn’t the moon be held on the surface of the earth by gravity..

[agile jay]

You can also see a lot of moons as planets that have been captured by bigger planets.

[agile jay]

Gravity isn't as strong as the kinetic energy of the impact.

[fast meteor]

https://letstalkscience.ca/educational-resources/backgrounders/origin-earths-moon

[meager sparrow]

? The kinetic energy is being excreted on one direction.. the same direction that the moon will be pulled at.. by the gravitational pull of the earth

[meager sparrow]

Ok thank you…

[agile jay]

Different masses are moving at different speeds around the star, even if they follow the same orbit.

[south kindle]

https://youtu.be/wkQuOrsgVGY?si=qrl5Bdx_Mr4L-f6_&t=2603 > this covers everything I just said. highly recommend the entire video.

NICE!

[agile jay]

The event?

[south kindle]

https://tenor.com/view/queen-freddie-mercury-we-are-the-champions-champion-sing-gif-4654136

[meager sparrow]

That is legit just a theory. They have no proof that. The event. actually took place. That is why they say “how do we think the moon formed”?

[agile jay]

Except when you look at the elements on the surface of the moon, and how it's distributed...

[meager sparrow]

What type of elements are in the surface of the moon? And how are they distributed?

[minor shard]

Good thing we have Perplexity to look it up 😉

[half venture]

Dammit it was a draw

[agile jay]

The isotopes are way too similar on the moon's surface, super sus.

[agile jay]

Lol, even when I was on holiday, you couldn't beat me 😂

[half venture]

I thought I did

[agile jay]

GG anyway

[half venture]

Yeah

[agile jay]

It's the same price either way.

[half venture]

How do I know I got the pro tho

[agile jay]

Ask them, I guess.

I assume they just apply it on their end.

[meager sparrow]

The isotopes of what? What types of elements are similar on earth and moon and how are they distributed? Explain..

[agile jay]

Can't you use perplexity to find the answer yourself? I'm not a free physics professor, lol.

[half venture]

😭

[agile jay]

My PhD was on quantum mechanics, not Astrophysics lol.

[meager sparrow]

Dude you are making. Claims about something you believe and when I ask you for proof or give information you say “search it up” like bruh that is not have you proof something

[agile jay]

Although I did take it during my bachelors.

[meager sparrow]

You don’t answer someone who is asking you a question by saying “search it up”

That means you have no proof or research to back your statements

[agile jay]

Sorry, I don't have all my research in my pocket.

On a topic which is not polarized in the science community...

[foggy whale]

just got here so i'm assuming that's an amazing chemistry joke

[fast meteor]

https://www.perplexity.ai/search/can-the-formation-OM.Zq555TbGsFIQubqIuvQ

even math arithmetic itself relies on zfc peano axiom/model

[tame current]

Is the Sonar model just a name change, or is it different in some way to the experimental model?

[fast meteor]

an alias: according to
https://docs.perplexity.ai/discuss/65d956e39db34f001ff8ce0a
https://docs.perplexity.ai/discuss/6582f98b41714c00723d5d5c
https://docs.perplexity.ai/discuss/6601ffd6bd5f0e0045ac5d16

[half venture]

@wraith crow hey may I know how do we know that the pro subscription was given ?

From the event I mean.

[south kindle]

I got a code from Alex, but i'm struggling to input it at check out. not seeing any fields or options to use a promo code.

[agile jay]

Check DM's

[agile jay]

Yep, they are currently disabled.

Because copy.ai fked up...

[half venture]

I have nothing in my dms tho

[agile jay]

You can ask in #1140622008086970420

#👉│feedback-general is probably bettwr

Since Alex responds there

[fast meteor]

or drop a dm to alex

[half venture]

Nevermind I just got it

[south kindle]

https://www.star.nesdis.noaa.gov/GOES/conus_band.php?sat=G16&band=GEOCOLOR&length=24 > from the GOES sat over North America. you can see the path and how dark it was.

[obtuse hatch]

anyone know how I can access the resources from the API and even change the search method like "search Reddit posts, use wolfram..."

[agile jay]

You can choose the model by editing the url.

I think it's something like &focus=writing

https://www.perplexity.ai/search?q=[QUERY]&focus=[internet, scholar, writing, wolfram, youtube, reddit]&copilot=[true, false]

[crisp geyser]

Hi, i just joined now i am a pro user but i don't see any pro channel here

[south kindle]

they should see your message here and give the roles

[crisp geyser]

Ohh

[south kindle]

i don't believe it's automatic. you just gotta ask for them.

[crisp geyser]

Okay, thanks for helping

[tame current]

You have to click on the link from the settings again once you joined for some reason

Worked for me

[fading moth]

There is a pro channel?

I have pro D:

[crisp geyser]

somehow i got access now and i didn't do anything

[south kindle]

you were given the Pro role.

[crisp geyser]

yeah

[fading moth]

Egad, I see it now

It's so beautiful

[finite badger]

Is the Pro role need to be hand given?

[serene sand]

Please leave the server and rejoin via the Pro Discord link at https://www.perplexity.ai/settings/account

[finite badger]

Hm, didn't seem to work

Yeah, the leave and rejoin procedure doesn't seem to fly. Tried twice 🤔

[signal hamlet]

You've already got the role, @finite badger, so you should see those.

[finite badger]

Ah, I think I was manually granted it

Thanks!

[ripe void]

Does perplexity have research papers?

Or accessible ones

[south kindle]

like published by Perplexity? good question actually. if you mean sourcing research papers, that's specifically what Academic focus mode is for.

[brittle merlin]

random question / complaint (maybe)

when im copying from perplexity it randomly turns my text white and background black how can i fix this?

[south kindle]

is that from selecting text and copy or via the copy button on the bottom right?

a workaround is copy it into a blank notepad doc to remove formatting, then copy/paste from there.

[brittle merlin]

both

[half venture]

Meta releasing smaller llama 3 next week

[brittle merlin]

i ended up just clearing all

[south kindle]

otherwise may be an issue with the interface your copying into

[brittle merlin]

clear all formatting seems to do the trick

so another dumb question now LOL

[south kindle]

not if anyone else sees it and learns something. don't beat yourself up!

[brittle merlin]

we get "pro" "premium" version of all these like claude?

or how does that does that work?

[south kindle]

gotta select the model in settings.

or you'll see the option under "rewrite"

[half venture]

Why do you think they are only releasing smaller versions next week

And keeping bigger models under wraps

Not runnable on consumer hardware maybe

[brittle merlin]

im asking are they considered the premium of the apps like i love claude 3 opus

but is it the same as if i went to there offical website and bought there?

[half venture]

If your asking if it's using gpt 4 or claude opus YES

[brittle merlin]

im assuming your anwsering eun?

[south kindle]

hold the phone. my b

[south kindle]

Lower context window in PPLX from what we are aware of. i think we figured it out to be 30k? i don't recall specifically.

[brittle merlin]

so 👉 👈 none of that made sense to me because im dumb af LOL

[south kindle]

context window is how much text it can keep in memory before it starts to delete old and replace with new. like an ongoing DVR camera system that can only store 1 day of video. once it hits 24 hours, hour 1 gets replaced with hour 25.

[brittle merlin]

ohhh so its very similar just less recording

[south kindle]

on the claude site it's either 100k or higher. so equal to multiple copies of a Harry Potter book. as opposed to like only 1 copy of context length in PPLX. so unless your loading in multiple huge PDFs to work with, you wouldn't notice the difference.

[brittle merlin]

thats so cool!

[weary plank]

hey guys, does perplexy also work well for solving advanced calculations? in free mode.

[brittle merlin]

explain a question to better understand?

like its doing amazing at helping me with discrete math which is annoying

[brittle merlin]

if your asking if pro/free would be different in the way it anwsers... no they would be treated the same

[south kindle]

use Writing Focus mode to avoid bringing in search results which aren't needed for this. should help a lot.

anyone have experience using the wolfram-alpha focus for complex math calculations?

[weary plank]

OK OK I will give an example

[south kindle]

https://www.perplexity.ai/search/2x3-3x2-5x-XsJKpQCERfCX0DslBCcSHg > this doesn't count as "complex/advanced math" but forcing it to use <scratchpad> at least provides the "work" aspects.

[weary plank]

Well, I would like to know if Perplexity would be able to perform calculations for product dilutions and statistics, proportions, and physical calculations. You know, university-level calculations?

[minor shard]

Is this correct

[south kindle]

I've run it a few times and it keeps coming up with the same result.

[minor shard]

I can’t see it 😦

[south kindle]

gotta set to public.

[brittle merlin]

https://www.perplexity.ai/search/2x3-3x2-5x-h52rp.ZpRdelGF.Di.lTnA

that work?

[south kindle]

now i'm more confused.

[finite badger]

https://www.wolframalpha.com/input?i=2x^3+%2B+3x^2+-+5x+%2B+7+%3D+0

You probably shouldn't try to use Perplexity for math

[south kindle]

PRobably accurate

although that was my point in asking about the wolfram-focus.

[brittle merlin]

chatgpt would be better i think

[south kindle]

GPT-4 Turbo in PPLX should be the same though.

[minor shard]

LLMs are not good at math/logic, in my experience

[finite badger]

They're all gonna be bad at this kind of task

[weary plank]

What would this focus mode be?

[south kindle]

in theory, sending it over to the same engine here #💬│general message

[brittle merlin]

Solution By Steps

Step 1: Define the Polynomial Equation
We are given the cubic equation (2x^3 + 3x^2 - 5x + 7 = 0). This is a third-degree polynomial, meaning it has three roots (solutions).

Step 2: Calculate the Roots
Using algebraic methods to solve cubic equations, we find the roots of the given equation. This involves complex calculations that are typically done using mathematical software or formulas derived from the cubic equation's characteristics.

Final Answer
The solutions (roots) of the equation (2x^3 + 3x^2 - 5x + 7 = 0) are complex and involve cubic roots of complex numbers. They are approximately:

1. (x \approx -1.769)
2. (x \approx 0.884 + 1.324i)
3. (x \approx 0.884 - 1.324i)

Key Concept
Cubic Equation Solution

Key Concept Explanation
A cubic equation, which is a polynomial of degree three, can have up to three roots, including real and complex ones. The general form is (ax^3 + bx^2 + cx + d = 0), where (a), (b), (c), and (d) are constants, and (a \neq 0). Solving a cubic equation involves finding the values of (x) that satisfy the equation. The solutions can be obtained using various methods, including factoring, graphing, or using the cubic formula, which is a more complex version of the quadratic formula.

i got this with a chat gpt lol

[finite badger]

Using it in Wolfram Alpha mode has a much better shot to work

idk exactly how they did it under the hood but if it's using WolframAlpha as a tool, it's much more likely to be able to actually do math

[brittle merlin]

yeah ai isnt it chief

Welcome to the advanced problem-solving mode, where we delve into the world of algebra and beyond. I'm Math, ready to take you on a journey through the complexities of mathematical equations.

Let's solve the cubic equation (2x^3 + 3x^2 - 5x + 7 = 0). To start, I'll perform some initial calculations.

Let's get started!

To solve the cubic equation (2x^3 + 3x^2 - 5x + 7 = 0), we get three solutions. These solutions are complex because the equation has no real roots that satisfy it completely. The solutions are given in a form involving complex numbers and cube roots, indicative of the complexities involved in cubic equations.

However, these expressions are quite complicated and not in their simplest form. To understand these solutions better, we need to simplify them. Let's proceed with that.

The simplified forms of the solutions to (2x^3 + 3x^2 - 5x + 7 = 0) still involve complex numbers and cube roots. They are intricate due to the nature of cubic equations, especially those with no real roots. These expressions illustrate the diversity and complexity of solutions in polynomial equations.

The solutions can be expressed as:

1. ( \frac{3^{\frac{2}{3}} \left(\frac{13}{3} + \left(-6 + (1 + \sqrt{3}i)(3\sqrt{25809} + 540)^{\frac{1}{3}}\right)(1 + \sqrt{3}i)(3\sqrt{25809} + 540)^{\frac{1}{3}}\right)}{36((1 + \sqrt{3}i)(\sqrt{25809} + 180)^{\frac{1}{3}})} )

2. ( \frac{3^{\frac{2}{3}} \left(\frac{13}{3} + \left(-6 + (1 - \sqrt{3}i)(3\sqrt{25809} + 540)^{\frac{1}{3}}\right)(1 - \sqrt{3}i)(3\sqrt{25809} + 540)^{\frac{1}{3}}\right)}{36((1 - \sqrt{3}i)(\sqrt{25809} + 180)^{\frac{1}{3}})} )
3. ( -\frac{3^{\frac{1}{3}} (\sqrt{25809} + 180)^{\frac{1}{3}}}{6} - \frac{13 \times 3^{\frac{2}{3}}}{6(\sqrt{25809} + 180)^{\frac{1}{3}}} - \frac{1}{2} )

These results, though complex, are a testament to the rich structure of polynomial equations. They show how cubic equations can yield intricate solutions involving real and imaginary parts, cube roots, and complex arithmetic.

[weary plank]

Okay guys, I didn't quite understand... So, is Perplexity good for performing and explaining calculations or not?

[brittle merlin]

no

[weary plank]

:/

[brittle merlin]

it sucks with this problem

How many license plates can be made using either two uppercase English letters followed by five digits or three uppercase English letters followed by four digits?

[weary plank]

So we're going to stick with the primitive method of paper and pen then... oh dear..

[brittle merlin]

my advice is run it through a couple of times on different ones chat gpt 4 is really good for math questions

[wooden karma]

I've prompted claude 3 opus three times now for my homework and it has not yet denied my requests due to limitations

what is this black magic

[brittle merlin]

i love claude 3 opus its been my go to

[south kindle]

Ideally we'd be using an LLM specifically trained on the problems you are working on. but that seems unlikely to exist exactly.

[wooden karma]

at this point i'd probably still be arguing with gpt-4 turbo trying to convince it to do what i asked in the first prompt

[brittle merlin]

LOL glad im not alone tbh
I've found Claude 3 Opus to crush ChatGPT-4 in long coding.

It's because of Claude's 99% token recall

[wooden karma]

its also sick at coding

[brittle merlin]

its been a life saver for my coding classes because i dont know poop

[wooden karma]

i gave it a kinda complex js file to find a (relatively simple) bug in and it did so perfectly

would have taken me like 10 mins to look through all of the code myself

[brittle merlin]

thats really cool it just helped me with c++ ❤️

[wooden karma]

for me its mainly that it can actually write essays without me gaslighting it about its limitations

[south kindle]

AI accidently made us all good at gaslighting apparently lol

[wooden karma]

legit 80% of the time i spent on gpt-4 was trying to convince it that it can actually fulfill my request

[wooden karma]

gpt-4 turbo specifically

[agile jay]

Rip, I don't use gaslighting techniques

[wooden karma]

gpt-4 wasnt that bad, but context window is too small sometimes

the non turbo version

[agile jay]

GPT3.5Turbo?

[wooden karma]

4

the normal gpt-4

it didnt complain about limitations for 80% of the prompts

[agile jay]

It's probably also because you are using it through the API.

[wooden karma]

i was using it in playground cuz its cheaper

[agile jay]

Which isn't as restrictive.

[wooden karma]

if API isn't as restrictive then im sorry for all the gpt-4 turbo app users

[agile jay]

Yep, API has always been less restrictive.

[wooden karma]

i had to spend like 2x more money on it after gpt-4 turbo

it cost less per token but i had to try to gaslight it too many times

[agile jay]

Yep, luckily perplexity most likely has a better deal than the base one.

[final finch]

Does anyone have any tips for setting up a "breaking news" feed for certain industries / topics in Perplexity?

[south kindle]

#1221054037894365294 message

that was someones last attempt at it.

[ripe void]

I should have been more clear yeah, I mean published by perplexity

I could maybe try with academic search, but if I could find stuff I’d assume they’d already have a page for research

[foggy whale]

no rush on this, but I've also tried leaving and coming back to no avail

[devout bronze]

yeah my perplexity is cooked too

[devout bronze]

you also exposed your IP

they removed DNS records 💀

nvm

back online

[tame current]

Hello, anyone knows the name of font that Perplexity Team using in the poster? It’s look future. 🥺 Thank you so much

[brittle merlin]

https://www.perplexity.ai/search/What-is-the-enJMIYTcTtaomjt93gBMUg

[lone citrus]

Guys, is there any of the platform that can provide 600 Claude3(not haikku) messages per day like pplx?

[south kindle]

I believe the 600 you are listing is for Pro searches specifically.

[brittle merlin]

claude.ai

[south kindle]

They removed usage limits for Claude 3 models for Pro users.

[lone citrus]

I tried, not true

They had crazy limit for Opus, specially with long documents or PDFS

[brittle merlin]

Poe i guess i never even heard of it

In POE there is a message limit to GPT-4 and Claude 2.

You'll receive unlimited messaging on popular bots like ChatGPT and Claude-instant, and up to 600 and 1,000 messages on powerful bots like GPT-4 and Claude-2-100k. After messaging limits are reached, messaging speed, quality or bot availability may be reduced.

[lone citrus]

Poe?

[gusty summit]

35岁现象

[lone citrus]

Times changed

[brittle merlin]

https://poe.com/

[lone citrus]

That's for old claude, claude 3 had a increadiable high price

[south kindle]

The creator of Poe is also still on the Open AI board. It’s a decent product that lets you use multiple models in a chat interface.

[lone citrus]

I think about 200 sonnet, 50 opus per month😓

[south kindle]

But the limits I’m not sure of honestly.

[brittle merlin]

thats a old post its been since updated with claude 3

[south kindle]

Personally I find I get more value from everything included with PPLX Pro. But currently i let my pro lapse in anticipation of the egg hunt rewards. So I’m back to basic PPLX for now.

[lone citrus]

1000000/12000= 83.33

That's why I give up the poe, the price rised for two times

[south kindle]

They haven’t updated that documentation yet it seems.

I know I used more than 50 opus messages within PPLX in 1 day while refining system prompts for scratchpad. Where did you notice the 600 figure previously ?

I’m still leaning towards confusion over using the pro models and the Pro Search quota that gets displayed when you use it. Which always seemed to reset back up to 550 or so.

[dim ginkgo]

The normal Claude 3 Opus in Poe has a 32k context, right?

[south kindle]

That’s a question for the Poe discord I think.

[lone citrus]

The pplx blog mentioned the pro serch limit is 600 per day and the Advanced models(gpt4 and claude3) are share same usage

[south kindle]

#📢│announcements message

[calm swan]

I don't know why but lately claude 3 has been nerfed that it not followed the instructions correctly

[lone citrus]

And it's not reset after 12 pm, it;s just slowly getting back to 600 after 24 hours

like what kind of nerfed?

[calm swan]

It not followed the prompt correctly

[south kindle]

Per the announcement, they are allowing unlimited Opus quarries.

I saw reports on Reddit today of people saying they had issues with context length and instruction following. Likely something internal on their side.

[lone citrus]

Maybe try another prompt

[south kindle]

Or feed it the initial prompt again. Ask it to summarized the entire chat so far and see what it leaves out. Could be hallucination, but points to a possible indication of the issue I think.

[lone citrus]

LLMs are complex and unpredictable. What works today may not necessarily work tomorrow. Their performance can be inconsistent and influenced by various factors.

[calm swan]

https://www.perplexity.ai/search/You-are-PyWiz-vIXfQzNOQJWdkSJusm3t0Q

I even already told it in the beginning

[livid mantle]

this is not poe comunity we are perplexitytopians

[dim ginkgo]

But this is a general chat😃

[brittle merlin]

damn pywiz

[livid mantle]

do u like poe

[dim ginkgo]

I'm interested in everything about AI

[brittle merlin]

https://www.perplexity.ai/search/You-are-PyWiz-vfl8fRlpQa.hs9jAhTWlww

mine wanted to work after i b slapped it

[calm swan]

Yeah, Opus after few multi turn chat,it works, but in a single one, it can't, I don't know if it's because Anthropic nerfed it or if the temperature settings are just messed up

[brittle merlin]

lets try another one

[calm swan]

okay go ahead

Only GPT-4 and GPT-Turbo (i guess), compeltely not giving the answer but rather the hint

[brittle merlin]

https://www.perplexity.ai/search/You-are-PyWiz-l9_LsH1SRq2yASqUcwYafQ

gave me better results

MISTRAL LARGE

[calm swan]

Yeah, because mistral trained by GPT-4 output ahhahahah

[brittle merlin]

lol shhhh

https://tenor.com/view/success-great-job-nice-great-success-great-gif-5586706

[south kindle]

We could optimize that prompt though. Use more brackets to emphasize the important parts. It shouldn’t veer off course that soon though.

[brittle merlin]

Great success
https://www.perplexity.ai/search/You-are-PyWiz-qcL8ciZvQUGvZD317rtRzQ

default is the worst haha

[calm swan]

ahhahahha

[calm swan]

yeah, in order Claude 3 model to work, we need to structured the prompt correctly using <></>, but it's quite hard tho for someone who is not good at prompt engineering

[south kindle]

It’s just guessing the best output based on solving the python problem. We could try “reminding” it to use “pywiz” to address the problem in later replies as well.

Try opening a new thread and start off with “help me optimize” this prompt. Then put the prompt in “ “.

[calm swan]

This prompt u mean?
You are PyWiz, an AI chatbot designed to teach Python programming through interactive problem-solving. When a user presents a problem, guide them step-by-step towards the solution by providing hints, explanations, and examples. Avoid giving the solution directly unless the user expresses a desire to give up or requests it after attempting to solve the problem.

[south kindle]

That would be in quotes. Then the task of “optimize this prompt” is before or after it. Results may very depending on if it’s prompt to optimize then task or task then prompt to optimize

[brittle merlin]

thats alot of work just for it to be my best friend

[south kindle]

Not really with copy paste. Using LLMs to optimize prompts is a know usecase that helps. And ideally once you get it dialed in you don’t need to repeat the optimize steps.

Or at least on web with a mouse/keyboard. On mobile that could be annoying.

Is pywiz already a known system prompt/GPT via chatgpt and PPLX is just being lazy?

[brittle merlin]

https://www.perplexity.ai/search/youre-not-allowed-GooM0L4SSHexuGfLmkl.ZQ

im gonna throw this thing into hell

[calm swan]

ahhahaha, i told u, opus kinda nerfed right now

[brittle merlin]

its brain ded LOL

https://tenor.com/view/borat-king-king-in-the-castle-gif-12965265

gpt 3.5 does it so easily LOL

[calm swan]

@south kindle
https://www.perplexity.ai/search/Optimize-this-prompt-iRSckxLeSfW_fX1V36naWg

I even already used the prompt that optimized by Opus but it itself didn't follow correctly

[brittle merlin]

Let's solve this step by step. First, we need to add the two numbers a and b. After that, we'll convert their sum to a binary string. Here's how you can approach it:

1. Add a and b to get their sum.
2. Convert the sum to binary.
3. Since binary conversion in Python results in a string prefixed with '0b', you'll need to remove this prefix to get the final binary string.

Now, try to implement these steps in the function add_binary(a, b). If you need more guidance on any of these steps, just let me know!

[calm swan]

yeah, the problem with openai model is just creativity, it's not that creative as claude 3 but it followed the instruction correctly

[brittle merlin]

damn i made mine hate me LOL

https://www.perplexity.ai/search/you-are-a-fMb_PdbyQYiC1snDSNsT.w

[calm swan]

I think the problem with claude 3 opus in perplexity is temperature quite high, here i set the temperature low and it's follow correctly:
Full output:
Great! Let's work on solving this problem step by step. The goal is to find the first element in the array that is not consecutive, meaning it's not exactly 1 larger than the previous element.

Here's a step-by-step approach to solve this problem:

1. Iterate through the array starting from the second element (index 1).
2. For each element, compare it with the previous element.
3. If the current element is not exactly 1 larger than the previous element, it means we have found the first non-consecutive number. Return that number.
4. If we reach the end of the array without finding a non-consecutive number, it means the entire array is consecutive. In this case, return None (or null in some languages).

Now, let's start implementing the solution in Python. I'll provide you with some hints and examples along the way.

def first_non_consecutive(arr):
    # Your code here
    pass

Hint 1: You can use a for loop to iterate through the array starting from the second element (index 1).

Example:

for i in range(1, len(arr)):
    # Compare arr[i] with arr[i-1]
    # If not consecutive, return arr[i]

Try to implement the solution based on the given hints. If you get stuck or need further assistance, feel free to ask!

@brittle merlin
try this:https://app.wordware.ai/r/7df30863-5cf2-4a81-8563-f24c279a77bb

[brittle merlin]

thats kinda fancy

damn i need to cancel chat gpt and get another like poe or what ever

perplexity does pretty much the same for coding

[calm swan]

what your main usecase using perplexity? coding?

[brittle merlin]

math/coding/ school/dumb qeustions

i have a harpa.ai which is taking claude and chat gpt and puts in a toolbar so i dont have to take screenshots

[calm swan]

wait what is that and how it works

[brittle merlin]

its ai on a toolbar so like i can have it explain a page

one second and ill show a example

[calm swan]

how much it cost

[brittle merlin]

xplanation of the Webpage Content
The provided webpage content is related to a homework question from a course, specifically focused on "Counting Principles". Let me break down the key points and provide concise explanations with practical examples:

Required Information:

The college has 23 mathematics majors and 325 computer science majors.
The question is asking about picking two representatives, one from the mathematics majors and the other from the computer science majors.
Counting Principles:

The problem involves the concept of "Counting Principles", which is a fundamental technique in combinatorics to determine the number of possible ways to select or arrange a set of items.
In this case, we need to find the number of ways to pick one representative from the 23 mathematics majors and one representative from the 325 computer science majors.
Explanation:

To calculate the number of possible combinations, we can multiply the number of choices for the mathematics major (23) and the number of choices for the computer science major (325).
The formula for this type of problem is: Number of choices for mathematics major × Number of choices for computer science major = Total number of combinations.
Plugging in the numbers, we get: 23 × 325 = 7,475 possible combinations.
Example:

Imagine you have a list of 23 students who are mathematics majors and a list of 325 students who are computer science majors.
You need to pick one representative from the mathematics majors and one representative from the computer science majors.
The number of ways you can do this is 7,475, as each mathematics major can be paired with any of the 325 computer science majors.
In summary, the key points are:

The college has 23 mathematics majors and 325 computer science majors.
The question is asking about picking two representatives, one from each major.
To calculate the number of possible combinations, we multiply the number of choices for the mathematics major (23) and the number of choices for the computer science major (325).
The final answer is 7,475 possible combinations.

so it just took this page with me saying that and turned it into my problem

so i didnt have to go and copy and paste a page

[calm swan]

Woah, quite cool

is it free?

[brittle merlin]

there is a free verison or you pay like 117 i think for lifetime

then add chatgpt premium to get chat gpt 4 and others

[brittle merlin]

https://harpa.ai/

[brittle merlin]

its a life saver for homework

[calm swan]

how mcuh u pay for that?

[brittle merlin]

i paid for the 120 lifetime

[brittle merlin]

$
120
/ LIFE
single user only, limited deal
paid once
Everything in FREE plus
Unlimited power messages and command runs
Custom AI commands
Edit predefined commands
Chat history, export chats
Reuse chats across tabs
Commands cloud storage
but free can do the same

id say just try it out

[brittle merlin]

https://chromewebstore.google.com/detail/harpa-ai-automation-agent/eanggfilgoajaocelnaflolkadkeghjp

[brittle merlin]

it replies to emails kinda good lol

Dear Gymreapers Team,

I recently received your email titled "Ready for a game-changer?" and wanted to share my thoughts. The concept immediately caught my attention, and I'm intrigued by what you're introducing.

Could you provide more details about what makes this offer a game-changer? I'm always on the lookout for innovative products and solutions that can enhance my fitness journey.

Looking forward to learning more about your new offerings.

Best regards,

[calm swan]

Which model write this?

[brittle merlin]

chat gpt 4

you get chat gpt 3.5 Gemini and Claude

i would just do free verison tbh

[brittle merlin]

i want perplexity to make something like this since they ahve claude 3 opus and chat gpt 4 and a great pro and so on it would be amazing

[sturdy trellis]

How can i use the claude in india

@sturdy trellis I am new here

[halcyon coral]

Claude 3 is available for all Perplexity Pro subscribers

[tame rose]

Hi, there's a system to ensure that the search is done on sources written in a specific language?

[foggy whale]

yeah we're gonna need a fact check on that claim

now the automations are pretty cool, I'll say that. Now that I use Arc as a browser though, just having a separate split pane open with whatever tool i need is better than having it baked into a browser extension.

[fluid mauve]

I read that as 300 with zeros to 3 decimal points.

[foggy whale]

squints, then LOLs

lawyer approved

[fluid mauve]

haha.big fonts, big screen being pedantic helps

[livid mantle]

[livid mantle]

ill try this siht out thanks drew

the summarizer is good enough for me, the creation stuff i know its only using chat gpt 3 api

it has a premium subscription im uninstalling it

[fossil sentinel]

Gemini 1.5 now supports audio and has a system instruction section

And I think video also includes audio too

POG OR WHAT

[livid mantle]

poggers

[lone citrus]

I am wonder why they don't just replace the Gemini 1 ultra now

[livid mantle]

no

we need to get more feedbacks and data

you understand why they dont instantly release gta 6

[lone citrus]

I am afraid of gta6 actually

In my imagination:

69.99$
Tons of bugs

[livid mantle]

gta 6 everything will be ai, you never want to come back to the real world again.

you will reincarnate being chased by police cars in the game

[lone citrus]

Require 5090rtx to run 1080p60fps

Release a 29.99 dlc after one month, need dlc to actually have some features showed in demo video

[lone citrus]

Ai is still a new thing for game, I do not believe they will bring Gen ai into the game

Maybe in the 2030, but not now

[livid mantle]

ai is not ready yet we are not there yet everyone with their fluff with

AGI

[lone citrus]

🤨

[livid mantle]

real ai must be sentient

[tame current]

it will probably require something like 3090 so it's okay

[livid mantle]

OH? lets ask perplexity

[lone citrus]

Most steam player still using 1060

[tame current]

Nobody can know what will that game require, but 4090 is nonsense

[tame current]

But when will GTA VI be released on PC? In 3-4 years ?

[livid mantle]

OS: Windows 10 64-bit
Processor: Intel Core i7-8700K / AMD Ryzen 7 3700X
Memory: 16 GB RA
Graphics: NVIDIA GeForce GTX 1080 Ti / AMD Radeon RX 5700 X
DirectX: Version 1
Storage: 150 GB available space (SSD recommended)``````fix
RECOMMENDED:
OS: Windows 11 64-bit
Processor: Intel Core i9-10900K / AMD Ryzen 9 5900X
Memory: 32 GB RAM
Graphics: NVIDIA GeForce RTX 3080 / AMD Radeon RX 6800 XT
DirectX: Version 12
Storage: 150 GB available space (NVMe SSD)```

Source

[lone citrus]

Cities Skylines 2 recommend 3080 GPU, it's could be true for 4090

[fossil sentinel]

When the day comes that game npcs use ai

People will become even more anti social

[sweet jasper]

Fr

Fdvr is epitome slavery

We have more access to consumption yet all we get is depressión

[livid mantle]

yes, birth rate will go down

https://tenor.com/view/roger-scott-wealthpress-stocks-roger-scott-wealthpress-wealthpress-roger-scott-gif-23073645

ai will be the last one standing in earth

laughing

[tame current]

And it will come soon. I'm actually working on making a game with AI NPCs. They will insult the player for being noob lol

[orchid magnet]

i want to just chat with the ai

and as an ai enthusiant i think im just going to stick with perplexity

because i can use both, and i mainly just use the ai bot for chatting, and brainstorming ways to help me make money. WHcih perplexity can do better because it can show me sources on how to do it

thank you for listening to my ted talk, i will be a perplexity pro member haha

pray they update the context window too

[tame current]

What's wrong with the context window?

[orchid magnet]

its 32k

vs if you paid for claude 3 its 200k

[tame current]

Does that include the web page content or are they processed in some way? Separate RAG model?

[orchid magnet]

but also, no internet access with claude 3

https://docs.anthropic.com/claude/docs/long-context-window-tips

[tame current]

You can have browsing plugin with Claude 3 on Librechat alt interface. But requires some work to set it up. And the API prices are much larger than just paying for the Pro

[orchid magnet]

you can?

[tame current]

You can, but it has searching capabilities of free perplexity. IIRC no multiple queries

[orchid magnet]

ohh boo

huh

but perplexity pro smacks tbh

and idrk if i need that context window

[tame current]

Tbh if there was another AI service with multiple queries, I'd probably ditch perplexity. Why can't someone implement it

[orchid magnet]

idk man

what do you think is wrong with perplexity

[tame current]

The quality of the responses is worse than on other sites, especially with opus

[orchid magnet]

thats true

[tame current]

I don't even know what's the hype with it, on pplx it's like a worse gpt4

[orchid magnet]

not gonna lie

might use that thing for claude you were talking about

[tame current]

Yeah, then prepare your ass for $1000 API bills. Opus is costly af, and you're gonna pass entire websites to it

[orchid magnet]

but i have the subscription

eh ur right

opus is mad expensive t bh

tbh if i want web access i can just use gemini lol

[tame current]

Gemini is a disaster, it's worse than labs models and only supports just one query. When I asked it to search up something for me, it refused with no clear reason

If anything, use solar medium on labs.perplexity.ai or copilot

[orchid magnet]

free perplexity isnt that bad is it

[tame current]

I'm starting to reconsider renewing my pro plan for the labs models and arc search

[orchid magnet]

same here

[tame current]

I'm gonna lose just the photo upload

[orchid magnet]

i can just use copiolot if i need google

opus is so worth for the writing

[feral marlin]

I pay for both perplexity and claude.ai and on writing mode it feels about the same with opus

[orchid magnet]

for real?

[feral marlin]

I interchangeably use the two I only use Claude if something has a single file with more than 32k tokens

Since perplexity can still handle normal context just caps files at 32k each

[orchid magnet]

i have also seen others post on the reddit that says. opus is becoming lazy now lol

so lunar

[feral marlin]

I think that’s just people getting used to it

I don’t really agree with that

[orchid magnet]

should i pay for perplexity or claude

[orchid magnet]

me neither, people just dont understand how to prompt. They never show proof either. I have also seen people getting banned for no reason lol

[feral marlin]

Probably perplexity I can easily run out of Claude usage for 8 hours pretty often

[orchid magnet]

wait actually?

[feral marlin]

Claude.ai limits are kinda low I think

[orchid magnet]

depends

the more files you use the lower it is

[feral marlin]

Really easy to reach them if you’re doing long conversations yeah

[orchid magnet]

yea

[feral marlin]

Perplexity also let’s you have the threads with custom prompts which are nice

[orchid magnet]

i woud rather have limits than have the ai spew out nonsense

which i have heard perplexity does

bc i dont think there s a limit

[feral marlin]

Idk I’ve had it spew out nonsense but I’ve also had Claude spew out nonsense

Both were while working with rustlang

[orchid magnet]

is perplexity good with long convos

[tame current]

For me, yes. Just the free isn't

[feral marlin]

I think it does good because with Claude for some reason if the convos long the website becomes terribly laggy and perplexity doesn’t do that

[orchid magnet]

still just kinda stuck between the opus membership and the perplexity membership tbh

[tame current]

What about ChatGPT Plus?

I've seen it has many good features that perplexity doesn't have

[feral marlin]

Idk I haven’t used it in a year

[tame current]

It can even create files, pdfs, edit photos ,etc

But still loses with perplexity pro when it comes to web search

[feral marlin]

It doesn’t have Claude models and has kind of restrictive limits

Which is why I avoid it

[orchid magnet]

i have been using it but i went back using it on my friends account and the writing is honestly awful compared to opus

like opus blows gpt out of the water

its insane

[tame current]

I'm using only GPT4 Turbo and Claude 3 Haiku, which is free anyways

Not a limitation for me

[orchid magnet]

how does haiku stack up to gpt4 turbo

but also i tried asking it to write like opus and it didnt work lmao

[tame current]

I'm using haiku on bigger prompts like "find mistakes in x", etc where I don't want it to run for minutes

And recommending it to ppl who can't afford any paid ai

Along with sonar

Beats gpt3.5, for sure

[fossil sentinel]

😂

[livid mantle]

This interview is filled with flowery language and avoids addressing the real issues surrounding AI. While Kevin Scott, Microsoft's CTO, emphasizes the potential of AI to improve lives and solve global challenges, he conveniently sidesteps the crucial question of who truly benefits from these advancements. He fails to acknowledge the significant power imbalance that exists, where large tech companies like Microsoft hold the reins of AI development and reap the majority of the rewards. He vaguely mentions "responsible AI" but offers no concrete solutions to address biases, job displacement, and the potential for misuse. Furthermore, his emphasis on individual entrepreneurship as a solution for economic disparities rings hollow, as it ignores the systemic issues that perpetuate inequality. It seems the "American Dream" he champions is one where tech giants like Microsoft control the narrative and the profits, leaving individuals to navigate the complex and often inequitable landscape of AI's impact.

[tame current]

What site is it?

[fossil sentinel]

It's google's aistudio

[fossil sentinel]

Did u feed this to gemini 😂

[tame current]

Won't work in EU. Is it paid? Does it require credit card?

[fossil sentinel]

It's a preview

So it's free for now

[livid mantle]

yes

[fossil sentinel]

If you use a VPN you might be able to access

Idk if gemini 1.5 is still invite only or if it's available to use freely

[tame current]

What's a good free Mac VPN? Proton seems to have just EU countries in free plan

[shell jasper]

Is there any invitation link?

Like a personal one to share with friends and family and obtain rewards??

[tame current]

You have it in settings page

You get $10 discount on the pro month cycle

[tiny nymph]

Hi all,

Do you guys have Perplexity model instead of having the searches powered by Claude 3 Opus?

[halcyon coral]

https://discord.com/channels/1047197230748151888/1198438847877480629

[tiny nymph]

Thanks, that makes sense. How can we be sure this is really Claude 3 behind the scene then?

I have asked pretty simple questions, and he was quite off, mixing up things I didn't ask and clearly said no to with what I asked for.

Didn't have these frustrations since GPT3.5

[halcyon coral]

As described in my post, you can tell very quickly by the "intelligence" of the model. I can see that you are using Pro search, but this can be a disadvantage, especially for simpler questions. I would therefore recommend trying out the simple questions without Pro Search 🙂

[tiny nymph]

Ok thanks for the advice 🙂

[brittle merlin]

You want the x if you want anything s is a scam more then x

[half venture]

New gpt 4 turbo api

Will perplexity update ?

[supple ridge]

Way to downplay the Perplexity models! Haha no, experimenting is key indeed

[formal parcel]

Asking same question

[fossil sentinel]

💀

Pushing the token limit lmao

[lone citrus]

Poe got gemini1.5, wonder when the pplx got too

[green thicket]

Whats Gemini 1.5 Performance?

Even on perplexity ai I’m mostly using GPT 4 turbo for everything. Anybody else?

It’s better than ChatGPT

[chilly fox]

I would put it right around GPT 4 level but with the context window improvements

That is with my usage

[lone citrus]

gemini 1.5 is about gpt4 performance, but with very lonnnnnnnng context windwo

max 1000k input

also, it's support video and audio, you could try it in the google ai lab

[agile jay]

Users don't actually have a 1M context, they just say that the model can work with 1M-10M context with minimum decrease in capabilities.

[half venture]

The inference speed on gpt 4 new turbo

Is truly extraordinary

But I don't feel like it's truly revolutionary

It does feel more agentic tho

Like claude opus

So probably anthropic Google and openai are the first ones to achieve agents

Waittt

No way

Did Microsoft already have this version of gpt 4 running on copilot

Before it even launched

😭

[fossil sentinel]

Depends

It can tackle advanced topics better if you upload a relevant textbook

But idk about vanilla

[fossil sentinel]

Wdym?

[stiff orbit]

can you elaborate a bit on that?

[half venture]

The speed at which the text is written

It's fast asf

Similar in speeds to copilot pro

Microsoft probably had access to this version even before

[stiff orbit]

i thought it was fast enough earlier but cool addition ig
What about quality of respones?

[half venture]

Lobotomised still

Doesn't have the charm orignal gpt 4 had

It's good for coding tho

And maths

[stiff orbit]

hm

[half venture]

Now we wait for may 9

[fossil sentinel]

Is copilot pro any good

[half venture]

No

[fossil sentinel]

As expected

[half venture]

It's even more lobotomized

Than chatgpt

Somehow

[fossil sentinel]

Copilot quality is so bad

[half venture]

It's not

If you ask it about correct things

And scientific things

It's actually good

But for usual chatting

Ugh

[fossil sentinel]

I mean I haven't used it in a while but quality of responses from Microsoft has always been bad

[half venture]

Disgusting

[fossil sentinel]

Can't shake off that reputation

[half venture]

So this was the model update

Jimmy apples talked about

Q* reasoning update on may 9

Alongside Google own reasoning thing

Probably

[fossil sentinel]

🤔

[half venture]

Anthropic might sweep in as well

There model is the most agentic of them all

Especially claude haiku

It's designed to be agentix

Agentic

[fossil sentinel]

https://openrouter.ai/models/google/gemini-pro-1.5

Wish I had access to 2.8 M context on aistudio 😒

[stiff orbit]

why is everyone hyped about this? isnt this old news 🤔

like the model is a bit dumb to even be worth with that much context window

wish opu$ had 1M context without burning through half my country's economy. that would be some stuff

[fossil sentinel]

Gemini 1.5?

[stiff orbit]

ye

[fossil sentinel]

If you use it properly it can perform better than gpt 4

[stiff orbit]

i have been testing and using it before it was public on aistudio.

even with clear instructions, it does very bad on tasks which require higher level of reasoning

i'll try again to see if its better now

[fossil sentinel]

If you ground it in quality information using that large context size it does better

Which is how I've been using it

Couldn't care less about how it performs vanilla lol

[half venture]

Current credit cost per msg:

G1.5: 250
G1.5-128K: 1,250
G1.5-1M: 3,000

Ok so I was wrong

Gemini advanced is not cheap to run

[agile jay]

Isn't that 1.5pro

[half venture]

Yeah

[agile jay]

Not ultra, so it's pretty dumb.

[half venture]

No actually in benchmarks

In competes with gpt 4

But I doubt it

[minor shard]

Opus/ sonnet/haiku ftw

[half venture]

Like the benchmarks seem so suspicious

Like that new open source model

Cobere was it ?

It beats gpt 4 but when I try it. It still feels worse

And this new gpt 4 might be better than claude opus but it will still feel worse

Simply because lobotomy kaisen LLM edition

I can't wait for meta to open source there thing man

A TRUE good open source model with a good dataset

[minor shard]

When was a new turbo announced?

[half venture]

Just when Google announced there gemini 1.5 api

So it feels like

If my timeline is correct it will be like this

[agile jay]

I have all the models, and Gemini is clearly behind GPT4T and Claude Opus

In capabilities.

[half venture]

April - llama 3 small , gpt 4.25 , gemini api for developers users get shafted.

[minor shard]

I’m assuming the new turbo is currently active in perplexity since it automatically switched over

[half venture]

In may - gemini consumer edition less woke , Q* and googles reasoning edition

In the Google next event

They kept barking about agents

[half venture]

So probably agents in may

Which is what openai calls Q*

[agile jay]

Not really

Q* most likely being able to think more by searching more possible outputs.

[formal parcel]

Tree search

[agile jay]

Since I'm guessing it's a play on the A* search algorithm.

[warm cave]

If OpenAI released agents under the name q* everyone you be disappointed

[formal parcel]

It's some sort of tree search

[agile jay]

Yep, arrakis was their agent testing model.

And Q* came after.

[formal parcel]

and people said it might be gradient descent / smth like that (like GPT can think before they complete the first token)

[half venture]

So whatever there agent thing is

May

[agile jay]

Q* basically let's the model use more compute for more difficult tasks.

[half venture]

At this point it's very clear that. Openai only releases things when Google does something

[agile jay]

By doing a more thorough search.

[agile jay]

Did they release anything when Gemini was released?

[half venture]

Voice update

Like that was it

[agile jay]

Exactly, so seems they don't care about Google.

[half venture]

No they definitely do

[half venture]

Gemini 1?5 pro
Release Sora demo

[formal parcel]

I remember it was so funny when they slapped Sora out hours after Gemini 1.5 Pro

[agile jay]

Gemini was pretty disappointing

[half venture]

Oh Google next event ???? Release gpt new model update

[formal parcel]

Yeah, 100%

[agile jay]

And 1.5 is just a longer context...

[formal parcel]

Is Arrakis Confirmed tho?

[agile jay]

Yep, it definitely existed.

[brittle merlin]

huge annoucment incoming!

im here

[formal parcel]

Someone got Qwen to hit 700K tokens

[brittle merlin]

its just huge becacuse im fat

[agile jay]

But they shelved it because it costs too much to run.

[formal parcel]

Let's be honest here, we technically never got a huge jump since transformers in 2017

[half venture]

I wonder how much of the cool sh it we will never get to see simply because it's not cheap

[agile jay]

Have you seen the difference between GPT-2, 3 and 4?

[formal parcel]

We need a architecture change

[formal parcel]

I was there in the text-davinci-002 days

[agile jay]

That's easier said than done.

[half venture]

Google did release a new architecture today

[formal parcel]

I did see the jump

[formal parcel]

Not a lot of improvement comapred to Transformers

[formal parcel]

exactly OAI had to release a new GPT-4 update today

[half venture]

No

Gemini is transformers

I meant google released a new model

Trained on a architecture that's not Transformers

[agile jay]

On the dev side...

[agile jay]

A paper or on their API?

[half venture]

Open source lol

Hugging face

It's a silent release

They didn't mention it anywhere

[warm cave]

Like Gemma?

Or something different

[half venture]

Yeah except this Gemma was not trained on transformers

[agile jay]

Gemma 1.5?

[warm cave]

?

[formal parcel]

remember, OpenAI had 1 year working on stuff, it took exactly 1 year (Claude 3 Opus) to beat GPT-4

[half venture]

RecurrentGemma, an efficiency-optimized architecture for research experimentation

[formal parcel]

Curious if everyone agrees on this, GPT-4 is MoE

[agile jay]

Most likely.

[warm cave]

Prob is, but we can’t know 100% unless I missed something

[agile jay]

Or inference would be more expensive.

[half venture]

Research innovation: RecurrentGemma showcases a non-transformer model that achieves high performance, highlighting advancements in deep learning research.

It's still only a 2 billion model

Not scaled up

Wonder what will happen if you scale it

[agile jay]

It becomes Big Gemma

[half venture]

😭

[formal parcel]

I think the only player that can truly compete against OpenAI is Anthropic

[half venture]

This GPT4 turbo release seems really rushed/disorganized

No metrics, no concrete examples

What is going on at OpenAI?

[agile jay]

They just don't care, lol.

[half venture]

Bro they are not only competing but BEATING OPENAI

Amazon better keep anthropic funded

[agile jay]

On public models, yes.

[half venture]

I’m surprised that OpenAI released and upgrade to GPT4 turbo rather than putting out a Haiku competitor

Especially with two small Llama3 models coming next week

It looks like OpenAI is content with falling to third place in the small/cheap model race

[formal parcel]

I still remember the GPT-4 Turbo on Dev Day, It felt like worse than GPT-3.5 Turbo

[half venture]

Makes me think they don't have any small models at all and only gpt 5 which is why they keep stalling

As seen by gpt 4 turbo

There efforts to cheapen

Su ck

Openai does well with big models

Gpt 3 largest model at time

Gpt 4 largest model at the time

[formal parcel]

I won't be suprised if in like 7 years students learn how to train GPT-4 on their school laptop

[half venture]

Forget 7

More like 2

[formal parcel]

Compute doesn't scale that fast

[warm cave]

There is always some kid that is smarter than you, no matter who you are

[agile jay]

Unless you're the only kid...

[half venture]

You can input 8000 tokens in

Instead of 400

4000

In copilot now

Huge

[fossil sentinel]

I remember for copilot you could change something in dev tools to input basically 32k context 😂

Idk if they changed that yet

[fossil sentinel]

Video and audio capability is pretty huge

Maybe not everyone will have a use case, but huge nonetheless

[agile jay]

Need to see it actually being used first.

To see if it's any good.

[fossil sentinel]

You can try it out

Aistudio

[agile jay]

Why isn't it on the client facing Gemini yet?

[fossil sentinel]

Gemini 1.5 is still a preview

[agile jay]

Yep, so it doesn't count then, unless users actually have access to it.

Especially since Google kills a lot of projects.

[fossil sentinel]

I doubt they're killing it anytime soon lmao

[muted pine]

Hey it's possible to add prompt in every prompt when i use perplexity ?

For exemple " Make search in english "

Because when i ask in french for exemple perplexity make search in french so avoid lot of result

[fossil sentinel]

There's an option in settings to set a language you want

[whole lark]

When I saw the Google Next keynote, I realized that they are working with a variety of Ai.

[polar swallow]

For some reason, Perplexity stopped giving answers in russian.
If I write my request - perplexity perfectly understands it, finds russian websites and even gives additional questions in russian, but answer by itself is written in english. Is there any way to fix it?

[muted pine]

No i want reply in French but search in english

[minor shard]

Anyone test GPT 4 turbo in perplexity to see if it’s better?

[worthy thistle]

I have with some math definitions and seems the same under writing

[devout bronze]

did perplexity remove the 7 day trial? i was tryna convince my friend to try out perplexity pro but it just wants to charge him and not give him the trial

[vivid kraken]

anyone getting the ability for read aloud responces in IOS app?

[minor shard]

Same lol

[minor shard]

It is working for me

[chilly nova]

Does anyone know if inside ChatHub, Perplexity uses my default model or Perplexity LLM?

[south kindle]

whats chathub specifically?

[minor shard]

I was about to ask that

[half venture]

Gemini 1.5 pro still is not as smart as the public available version gemini advanced

Like the roleplay ability of advanced is another level

And coding as well

Gemini 1.5 ultra will be revolutionary

[minor shard]

idk about that lol we will see

[half venture]

Claude 3 opus biggest weakness

Is probably coding

Gpt 4 has much less hallucinations

[agile jay]

Strange since I get the opposite.

[signal hamlet]

Full immersion?
https://www.perplexity.ai/search/what-are-the-jVCEnoPvQoivwnKgC.Rg8A

[agile jay]

Yep, so like VR but it feels like real life, or even better.

[minor shard]

Im fine with it as long as it doesn't have to be implanted into my brain 🙂

[half venture]

I am not fine with it either way

[agile jay]

You put random sht in your body everyday, when you eat.

[minor shard]

Trueeee but but

[half venture]

Unless it's being used to treat I dunno Alzheimer's or something

Then it's fine

[agile jay]

People said they would never carry a tracking device and a camera to track what they do everyday, and do you see anyone without a smart phone?

It will just be normal, and the people who don't use it will be the strange ones.

I would rather go the bioengineering route instead of brain implant though.

[lone citrus]

true, if new technology bring problems, fix the problems, don't fix the technology.

[minor shard]

I mean this is different than me altering my body. I don't mind it being something I carry around, just not something I have to modify my body for. Very different

I get ur point but that is the line I draw

even if everyone else is doing it

[agile jay]

I think most people just do what other people do, since we are a social species.

[minor shard]

I agree

[agile jay]

If you need an implant to get a good job, 99% of people will do it.

[minor shard]

While I agree, I think there is a hard line there that will make that impossible. Especially when it comes to requiring you to change ur body. We are not even close to society accepting something like that. Maybe in 200 years

I also think AI is advancing too fast for the public to accept. If it keeps going at this speed we are going to have some serious push back lol

usually its too late by then cause no one pays attention until its in your face lol

[agile jay]

Change your body? Isn't cosmetic surgery very popular?

As for the backlash, most of it comes from the fear of being replaced.

[minor shard]

Yes but the majority of jobs don't require it

[minor shard]

lots of scifi movies

😛

also people who don't want to learn new things

[agile jay]

Yes, because in most jobs, looking better doesn't make you a better worker.

But being more capable does...

[minor shard]

True... but I still think people wouldn't go that far. Although who knows what society will accept in the future. Some parts of the world can't agree on certain legal rights over their own body 👀

we could all be living under a dictator in 50 years. Who knows

I think people don't want to think for themselves. They want someone to tell them waht to do

I know im sounding contradictory lol

[agile jay]

Yep, but I assume if you asked most people would they want to be more intelligent, have better self control, and interface with technology using your mind, and all you need is a short 30m surgery, most people would say yes.

[minor shard]

I personally just can't imagine myself doing that right now. Maybe that will change in 10 years idk. Just the thought of it makes me cringe as useful as it would be. Thats just me tho

[agile jay]

I think when the technology gets a lot better, it will be strange to not get it.

[minor shard]

Thats what I am thinking, I just don't know if I will get over the modification to my body. its just weird to think about something being attached to me "unnaturally"

[agile jay]

Do you wear glasses?

[minor shard]

No i do not, i tried the smart ones but glasses in general are just annoying lol

[agile jay]

Well for most people, the difference when having an implant will be greater than when wearing glasses, so they will feel anxious without them.

[half venture]

Huge

seems to be another MoE. Mixtral-8x22b and a context length of around 64k tokens (65536 max_seq_len).

[minor shard]

feels like they are pumping out something new almost every week haha

[half venture]

It's open source

So probably not the best model

But it's 130b model

It's definitely gpt 4 class level

And we finally have a gpt 4 open source model

Sam Altman better fu ck ing release something that is not GPT 4 VERSION 30000

[brittle merlin]

is there a ai that has no protection

or is perplexity labs best choice

]

[warm cave]

Do you have a link for this?

Oh I think I found it

[half venture]

https://x.com/MistralAI/status/1777869263778291896

Mistral AI (@MistralAI) on X

magnet:?xt=urn:btih:9238b09245d0d8cd915be09927769d5f7584c1c9&dn=mixtral-8x22b&tr=udp%3A%2F%https://t.co/2UepcMGLGd%3A1337%2Fannounce&tr=http%3A%2F%https://t.co/OdtBUsbeV5%3A1337%2Fannounce

Twitter · Apr 10, 2024, 1:20 AM

I don't think you will be able to run it lol

130b model

It's probably gpt 4 level like large was but open source

[warm cave]

Yeah, my computer comfortably runs 8x7b, no chance 8x22b, haha. But I want to use in with groq

Hopefully they add it soon

[minor shard]

Would be cool to have on Perpleixty 👀

[brittle merlin]

mistralai is op

its in perplexity labs

[minor shard]

I would love to see the 130b model #dreaming

[minor shard]

May 1st: GPT 4.5 JUST RELEASED! : SUPER MEGA FASTER BETTER (we won't go into details on what we changed though)

[brittle merlin]

sources?

[half venture]

What are our predictions on when gpt 5 releases

[brittle merlin]

befor gta 6

[half venture]

Nice joke

If I had to take a guess

Gemini 2.0 in 2024 December

Wait

When is the US election

[minor shard]

November

We will prolly get like 100 new models in December

[half venture]

Yeah

Once the election is over

So gpt 5 and gemini 2.0

[minor shard]

Then the world ends from AGI

[half venture]

In December

[half venture]

No lmao

We will use it for a month

And then start sh it talking how awful gpt 5 is

[brittle merlin]

then we get gta 6 with the release of the ps7 and xbox 1200

and the consoles come with ai

is there a unrestricted ai not for like dumb nsfw crap but coding questions about cheating LOL

[minor shard]

I guess I was wrong and they have to update the model https://www.perplexity.ai/search/What-is-the-DI.rc2EBQPK7phj6xFnwtA I thought it was automatic.

Oh duh cause it used to be preview now it’s just turbo lol

[half venture]

Mistral 8*22b > its between gpt4 and sonnet

Pretty solid

Open source

Too bad it requires so much compute

[sweet jasper]

https://aider.chat/2024/04/09/gpt-4-turbo.html

Stupid faster

[devout geyser]

Ouch guess I'll be sticking with opus

[elfin granite]

i heard the context length is limited in the pro version of perplexity for the different models... is that true? I want to go with pro but if its limited id much rather just subscribe directly to gpt or claude

[lone citrus]

some say yes, some say no

[west sage]

Is there any way for Perplexity to include an image directly in the answer?

[buoyant kindle]

It's very limited. And can't be any other way, considering the costs. If you want something like 32-200k, or 1000k, use Claude 3 Haiku in PPLX Labs or Gemini 1.5 Pro (both are free). But Claude 3 Opus and GPT-4T in PPLX Pro are "smarter".

Personally, I don't mind. I'd rather have Opus for many short questions, as it is today, than have it for very few questions with huge context window.
Ultimately, we'd have the choice between something like Opus 4k, Opus 200k, Haiku 200k and so on. I'm sure Opus 4k would be used the most.

[tame current]

I just tried it but the images are same as what it already shows... https://www.perplexity.ai/search/show-images-of-IpfumQMUTlyB6cm0pobRDA

doesn't seem to work the way you'd expect

[west sage]

oh that's interesting. I was trying to make this work with a prompt but was giving me alucinated links. Just made a small modification and now is working

[frozen tundra]

Anyone else on iPhone unable to edit their Collections?

[rugged ocean]

When using Claude 3 Opus in writing mode, is it better / more relevant to leave Pro mode on, or shoud I turn it off ? Does that make any difference ?

[stable radish]

Turn it off, pro is supposed to leverage web search that's why module is called "PRO Search".

In writing mode, you won't benefit from browser search.

[serene sand]

https://www.perplexity.ai/hub/faq/what-is-pro-search

[calm swan]

have u tried it?

[tame current]

the instruct version isnt even out yet lol

[meager sparrow]

Joe Biden

Joe Biden

Joe Biden

[quiet pier]

Code HTML saisie de données

[tame current]

so you got it work for any image on the web ?

[warm cave]

I normally keep “pro search” off for writing mode, but it dose have some benefits, it will ask you follow up questions, so if you ask it to write an essay about a vague topic, then it will often give you a list of options for what you want the essay to cover, or if you ask a question it may ask clarification questions before answering. Tho all of these are pointless if you just give it a good prompt in the first place in writing mode.

[tough pollen]

Hey guys I need some help

How do I generate images in mobile app

[orchid magnet]

Still stuck between opus and perplexity

[tame current]

what do you guys think of command r+ ?

would you say it's as good as gpt4 ?

I mean , probably not but still

[stark walrus]

Are there any outages going on right now? I've tried five different searches and gotten error codes on all of them.

[tame current]

works fine for me, are you using VPN ?

[stark walrus]

probably that 😛

[wraith crow]

it is currently only possible to generate images on the web.

[elfin heath]

Hey guys

Perplexity pro worth it? For learning blender and ureal engine stuff

[tame current]

there's a free trial you could try that

but if the free version suffices for you, you probably don't need the pro version

pro would give you access to more models and some other features but just look at the features and if see if you need them. personally, I think if you're not coding or writing, you're better off with the free version

one more thing that is help in pro is the ability to upload images

[elfin heath]

Ohh thank you

[sleek vortex]

gemini 1.5 pro

is so good in the ai console

only ai model with video and audio support

[fluid cloak]

Do you all think the main LLM Training Hardware will still be produced by NVIDIA?
(aka GPU, TPU or whatever new technology currently under development)
For the next 3 or 5 years

[livid stream]

Anyone having issues with Android app ATM?

[halcyon coral]

In general, there are no problems with the app 🙂
However, judging by your question, you seem to be experiencing problems, so it would be helpful if you could create a detailed #1140622008086970420, thank you!

[livid stream]

All good, it was my VPN 🫰

[tacit scroll]

cant wait to see Ayo and JLO this saturday! lets goooooo thank you perplexity!

technology is incredible i cant believe i dont have to use google anymore i can just type a quick question

[serene sand]

https://x.com/perplexity_ai/status/1778067977566294448

Perplexity (@perplexity_ai) on X

We teamed up with Raycast to make knowledge accessible anywhere, anytime on your Mac. New annual Raycast Pro subscribers get Perplexity Pro for free for 3 months, or 6 months if you include the advanced AI add-on.

Twitter · Apr 10, 2024, 2:30 PM

[minor shard]

No more having to find info yourself!

[tacit scroll]

so true!

[minor shard]

I wonder if anyone here has used this yet 👀 would love to know their experience

[devout bronze]

they removed that didn’t they?

[potent igloo]

lol you guys made perplexity so humble

he did created it, my guy is an artist

[wraith crow]

you mean raycast?

[minor shard]

Yes 🙂

ive been seeing a lot of these popping up

Wish there was a windows version

[wraith crow]

I've been using it for almost a year now, love it. It almost has an extension for everything you need and has quite a big community. And if not you can easily make one. I mainly use it to convert images offline to the right file extension.

[minor shard]

Oo that sounds nice. I will check it out thanks!

I like anything that saves me steps and time lol

[wraith crow]

Also a big plus is that the extensions are open source and you can view the code for every extension you download

[minor shard]

Thats really nice. Yeah I was checking out their "Store". Really just makes life so much easier

Thank you!

[potent igloo]

wow

[pseudo smelt]

@signal hamlet this is regarding the last announcement regarding partnership with Raycast. I don’t have Raycast Pro, but I do subscribe to Perplexity. Will that give me access to Perplexity from Raycast without paying for their Pro?

[celest stone]

question, I yeeted my pro sub from play subscriptions, when it runs out, will the button on the pro settings section of the website let me sign up through stripe again?

[warm cave]

So what is raycast, I’m looking at the website and it looks like spotlight search with AI

[snow briar]

What is raycast?

[sleek vortex]

well the ai was added later

[warm cave]

What types of things make it really great?

[sleek vortex]

its just a good and fast alternative to spotlight or idk alfred 5

you can add lots of extensions for functionality in app

[warm cave]

Oh, so its good even without ai, I’m going to download it and give it a try, I love spotlight, so if this is better than that is awesome

[sleek vortex]

yeah

my main issue with spotlight is that it was super slow

this is way better

idk about file searching, i dont really use that

i use it a ton to open apps

and random things like generating a qr code

clipboard history is really useful too

[warm cave]

Yeah, I tried a few spotlight alternatives a while back but did not find anything good, excited to try this, is the good stuff behind a pay wall?

[serene sand]

https://www.perplexity.ai/search/What-is-Raycast-.ItnfuzRRECM3I.88NKApg

[wraith crow]

the automations you can do with the app. and the build in functionality like day counter / currency converter etc. its definitely an upgrade coming from spotlight

[warm cave]

Oh cool, I’m a little skeptical when a AI tool comes out since they don’t often stand up to the hype, but this looks like a really cool tool even if it had no AI integration

I’m so happy, that was fast: #🦙│feedback-playground message

8x22b

[mystic mica]

Will perplexity have any integrations in the Rabbit R1 when it comes out?

[warm cave]

I think the rabbit search is powered by perplexity

[mystic mica]

Suppose that would make a lot of sense

[warm cave]

This might be a stupid question, but why dose perplexity labs exist, is it a testing ground for models, why are the models not added to the API or the main perplexity as an option. Don’t get me wrong, I like perplexity labs.

[serene sand]

https://www.perplexity.ai/search/This-might-be-Z6MsEm0cRcO_xdNCXpj2yg

[orchid magnet]

What’s ray cast

[warm cave]

#💬│general message

[orchid magnet]

Might buy

[whole lark]

#💗│sharing message

max 1500tokens in labs

[dawn junco]

I asked Claude where to watch the latter seasons of one piece and it basically just said “Crunchyroll or pirate it.” I’ve never seen an AI suggest pirating lmaooo

[warm cave]

Haha, had the same thing happed but with a book for school

Claude 3 opus as well

[dawn junco]

Its refreshing honestly to have an AI honest with the options

So many times, and especiallly with chatgpt, the ai's are so consumer forward that they will straight up lie to sound better, I compared an essay with a rubric twice, once purposely and very obviously leaving out one of the requirements and it just ignored that part of it.

[orchid magnet]

is this claude?

[dawn junco]

Yep

Sonnet

[orchid magnet]

love claude so much

i use opus

[dawn junco]

I saw a twitter post boasting a 10k prize for any LLM that could solve their puzzle or whatever and Opus beat it and claimed it within like 30 minutes or something.

Can't dream of better marketing than that

[orchid magnet]

genius

[shy thicket]

Saw the Raycast announcement. How does Raycast make Perplexity better?

[foggy whale]

Also a daily Raycast user but probably don't lean into the automations as much as i could. It's a really good launcher and i think it's better than Alfred or Launchbar on MacOS. Really i just want Quicksilver to come back because it did things that some of the newer launchers still don't.

[drowsy hound]

Hellooooo

I just found out about perplexity recently because I’ve been using chatgpt for awhile

And found out it’s basically chatgpt but with real time access to the internet, correct?

[shy thicket]

Yes. It makes Claude so much better for that reason

[drowsy hound]

I saw Claude and was wondering if there was an iOS app for it since I primarily search for stuff with ai on my phone and iPad

I use the computer but not too often compared to phone

This is basically Claude on my phone?

I wanna kinda cancel my subscription for chatgpt at this point if this is better but I also got so many chats on chatgpt as well that I like looking back to or following up in the future in case something comes up again

[shy thicket]

I cancelled chatgpt