AI Won't Be AGI, Until It Can At Least Do This (plus 6 key ways LLMs are being upgraded)

preview_player
Показать описание
The clearest demonstration yet of why current LLMs are not just ‘scale’ away from general intelligence. First, I’ll go over the dozen ways AI is getting murky, from dodgy marketing to tragedy-of-the-commons slop, Recall privacy violations to bad or delayed demos. But then I’ll touch on over a dozen papers - and real-life deployments, of LLMs, CNNs and more - making the case that we shouldn't throw out the baby with the bathwater. Crucially, I'll cover 6 key approaches that are being developed to drag LLMs toward AGI. this video will hopefully, at the very least, leave you much better informed on the current landscape of AI.

Рекомендации по теме
Комментарии
Автор

I've seen people who haven't reached AGI level either.

rumenemilov
Автор

This video shouldn't be considered just a piece of content. It is a timeless reference that greatly explain the real state of AI. Your work is a huge contribution to the free community, very appreciated!

Abdul-qoeb
Автор

Harnessing hallucinations as something constructive instead of viewing them as wholly detrimental is a fascinating and very exciting prospect.

TheTabascodragon
Автор

Claude Opus solves 17:45 correctly with this prompt: "
saa → Green
guu → Red
ree → Blue
fii → Purple
hoo saa → Green, Green
hoo ree → Blue, Blue
ree muo fii → Purple, Blue, Purple guu muo ree → Blue, Red, Blue hoo ree muo saa → Green, Blue, Blue, Green
fii muo hoo saa → Green, Green, Purple, Green, Green

hoo guu muo hoo fii →"

GPT-4o, GPT-4T, Gemini 1.5 Pro couldn't solve it despite many retries. Thanks for another great video!

drhxa
Автор

I've said it once I'll say it 1000 times: this is the best AI YouTube channel. Nothing else comes close in depth of analysis, balanced takes and engaging presentation.

tar-yyub
Автор

Best start to the week when AI Explained drops. The best AI news around!!

MegaSuperCritic
Автор

Perfect video, you stand out of the AI news crowd with it.
Great, measured perspective.

daPawlak
Автор

It’s incredibly refreshing to see ai coverage that’s factual and doesn’t dismiss the issue without promising utopia or Armageddon within the next 12 months. An internet for you!

lucacarey
Автор

Thank God you uploaded! This week has been so confusing and i need your voice to whisper sweet papers into my ear…

Citrusautomaton
Автор

You're legitimately the only source I trust when it comes to AI. Continue the good work.

lost-one
Автор

Great video as always. Let's not forget that GPT-4's vision can't capture all the necessary details; the visual representation isn't precise enough to be useful for the ARC test. I don't agree with the ARC test authors' argument, as the task's presentation to LLMs differs significantly from how people experience it. We'll have to wait for future models with improved vision capabilities to truly assess their performance on such tasks. Excellent points on AI marketing.

AICodingAdventures
Автор

This actually made me feel much more optimistic for the future of the landscape, great work as ever!

NowayJose
Автор

I really appreciate your honesty in this video. Exactly why youre the only AI channel ive stayed subscribed to. Great work.

Hiroprotagonist
Автор

This is such a well made video, Phil. I'm going to have to watch it twice to fully digest it. After not seeing a video from the channel in two weeks, I can tell you've been pretty busy!

keeganpenney
Автор

26:10 please make these videos as long as you want. We are here for this! Thanks!

TheRaizerx
Автор

This is one of the best videos on here. You are definitely a cut above most of the other AI fluff channels on YT.

AkhilBehl
Автор

"The world is more complex than it seems". Well said, and this is a typical phrase someone utters when in the midst of a journey. Once the destination is reached, the discovery complexity goes away and those not involved in the journey will call the result "obvious"*. Fascinating, I am glad to be alive while this technological revolution is unfolding, meandering, hitting and overcoming dead ends.

*To give an example, Aristotle was convinced that in order to maintain the speed of an object, you need to apply force. Which is very natural because friction is everywhere on Earth. Once Newton et al. posited that force is actually causing acceleration, the concept of friction was naturally integrated into the framework. To the high school student that I was once, this solution seemed obvious because it was presented to me as self-evident.

SaintCergue
Автор

I just don't understand why developing agi and therefore consolidating all power in one hands is such a worthwhile goal. I can't imagine what social ladders will exist to climb out of poverty or have any sort of freedom when agi arrives

makstheimba
Автор

Keeping it straight, your grounded perspective is appreciated. AI can be AGI, just not inside this prevalent LLM token-predictor.

Dannnneh
Автор

Great stuff! Really appreciate how each video distils so much information from loads of sources, and that they're also all individually cited in the description :)

someguy_namingly