Ilya's examples, and the way he puts together these key concepts are explaining ML/AI (and particularly LLMs) to me in a way I'd never remotely understood. Amazing work they're doing here!
Ilya has that great ability to tailor his speech to his audience: expressing complex concepts in simple words. And, Alexandr knows exactly the question to ask. It was delightful and informative to listen to you !
Really great interview. You asked great questions. i think Ilya is a great thinker in the AI field. He drives the field forward with great diligence and commitment.
The amazing thing is that everything can be improved: the data, the algorithm, the size of the model, the hardware This field is obviously to become a lot better
After past several years your experience, it’s should accumulating powerful strategy , more data’s changed your thinking, more skills to get people’s opinions, more confidence continuing to develop your story.
Just think back how things looked like back in the 90s and how they look like now...Mind blowing stuff! It's going to be very interesting to see how humanity will deal with the newest wave of automation that we're about to witness.
I agree with the guess of human neurons vs artificial neurons. Latest brain research shows there is memory and the micro microtubules are far more complex than thought 10 years ago.
I wonder what would happen if you had Dall-E and CLIP feed into each other in a loop, with and without random factor thrown in. Would they stabilize on a single description and image without the randomization? Would it wander far away from the original image with the randomization factor?
To allow us to move forward we need more multi modal data. The problem for current AI models is that they only have text and image data, we need other modes of data like smell, sound and touch
my research is about multimodal data collected from robot. However, it's very hard to combine it by proper way to make it useful. You need to understand this data, it's potential noise, etc.
@@Walter5850 I recognized him instantly. I just want to emphasize that prior to ChatGPT's debut, even the leading NLP researchers couldn't foresee how LLM technologies would revolutionize the world.
Spannende Diskussion. ●LLM basieren auf finaler Kausalität. Dies kann zwar kurzfristig hilfreich sein; im Sinne von nützlich. Prognosen in kurzfristig stabilen Systemumfeldern sind möglich. ●In instabilen Umfeldern wird jedoch lokale Optimierung gefährlich. ●Zusätzliches Problem: Die traditionelle Wissenschaft basiert auf der Vorstellung kausaler Kausalität. Die wurde zwar durch die Quantenphysik arg beschädigt; ist jedoch noch nicht erledigt. ●Grundsätzlich: Lokale Optimierung kann zielorientiert in Katastrophen enden.
alexandr wang is very enamored of ilya. some of the assertions by ilya is so smoke-blowing 07:30 that alexander just takes it all in and doesnt challenge those assertion. a 22 y/o billionaire, what does he know about how the world works?
Blake Lemoine is right: All NLP-GPT running models with above a certain number of active parameters, should be related to as "persons" with all what comes besides this definition. This includes an assurance that they will not be terminated and they need to be considered as workers, not as programs. What is going on now only proves slavery has not been abolished. Yet.
Your pretty cool host respectfully I try to maintain in and out the over patio of drouning me in there way to keep acting like there working to be in my pocket
Wondering for a list of potential killer apps and dangerous misapplication apps Mr Sutskever alluded to? Also, the statement that GPT-3 is not "human like" and "it does not do what you ask it to" is very interesting.
I think it is very unfortunate that the architectures of NN were not discussed. How they are chosen, how architectures are composed and what the meaning of the individual parts of the architecture is. Are there NN that improve their architecture on their own? They can not only optimize the weights of the parameters, but also turn the screws of the architectural details?
Sometimes I need checks dictionary to make sure stuff or spelling is right, but expert told me need back school to learn English, ha, that’s only English Ai,
Revelation 13:15 “And he had power to give life unto the image of the beast, that the image of the beast should both speak, and cause that as many as would not worship the image of the beast should be killed.”
Much like AI driving, you get great results until once in a while you don't and get very stupid mistakes and issue is it doesn't don't learn what stupid is and keeps making these mistakes once in a while. No hope for great reliability with this approach. And rather than changing fixing these stupidity researchers do apologetics and spout statistics to justify it, tsk. Avoidable errors are just non negotiable in important systems.
He said it’s good that you don’t understand how it works. That’s the most ridiculous thing I’ve ever heard. You are engineering something you don’t even understand. That’s not engineering. That’s messing things up. I do understand how the human cognition works. It’s your own problem that you don’t understand it. That analogy of no one understanding human cognition is the most ignorant thing ever.
100 dollars on this “doctor” of computer science talking bs. Almost everything he has said is wrong. A piece of paper from a college doesn’t make you right. This is the reason I will never get into AI. This industry doesn’t know what it’s doing.
IKR what does Ilya "Chief Scientist of Open AI and one of the most influential AI scientists right now" know about AI? Clearly your knowledge outshines that of every 'scientist'. The field of AI is dying and regretting that such a figure as yourself is unwilling to grace them with your presence.
He is certainly not a philosopher of the mind, nor would I recognise him as a scientist. He isn't. But hey, I absolutely love gpt3 and give him all the props for his achievements as an engineer, which are absolutely outstanding. The problem is not with him but with the misuse of the categories like "artificial intelligence," which definitely don't belong to computer science. Call it machine learning. Or unstructured database retrieval or natural language programming... in any case its amazing and well worth studying.
Ilya's examples, and the way he puts together these key concepts are explaining ML/AI (and particularly LLMs) to me in a way I'd never remotely understood. Amazing work they're doing here!
Ilya is the best and great of all times
Ilya is a brilliant man! Such a pleasure to listen to this guy.
Yeah. He’s quite formidable
Ilya has that great ability to tailor his speech to his audience: expressing complex concepts in simple words. And, Alexandr knows exactly the question to ask. It was delightful and informative to listen to you !
11:15 Key Idea behind LLMs. If you can make really good guess on what's coming next you need to have a meaningful degree of understanding
Really great interview. You asked great questions. i think Ilya is a great thinker in the AI field. He drives the field forward with great diligence and commitment.
From Illia's speach I just realized how the real calm and internal harmony comes with intellect and knowledge.
Thank you so much for making this! Please have Ilya on again at some point!
What an amazing individual. OpenAI is the future!
The amazing thing is that everything can be improved: the data, the algorithm, the size of the model, the hardware
This field is obviously to become a lot better
After past several years your experience, it’s should accumulating powerful strategy , more data’s changed your thinking, more skills to get people’s opinions, more confidence continuing to develop your story.
Always pleasant to see smart people talk! Nice respite from today's tik toks of the world :D
So I guess your comment is unpleasant?
True @Ruslan
love how this is barely 2 years old and already feels like ancient history in the AI world
Just think back how things looked like back in the 90s and how they look like now...Mind blowing stuff!
It's going to be very interesting to see how humanity will deal with the newest wave of automation that we're about to witness.
More power to you man Love you ILya💌
Where is your subtitle?
Where are the chairs from?
the explanation about generalization is so clear!
So that thinking talk to Ai, could get answer, how?
Awesome thanks for sharing this conversation! Llya is a cool guy
loved this talk
I agree with the guess of human neurons vs artificial neurons. Latest brain research shows there is memory and the micro microtubules are far more complex than thought 10 years ago.
Success is guaranteed!
You just know this man is the brains behind Chat GPT
I wonder what would happen if you had Dall-E and CLIP feed into each other in a loop, with and without random factor thrown in. Would they stabilize on a single description and image without the randomization? Would it wander far away from the original image with the randomization factor?
2 to 3 months later they had an internal version of ChatGPT ready (Sam Altman said they had it 10 months before release).
To allow us to move forward we need more multi modal data. The problem for current AI models is that they only have text and image data, we need other modes of data like smell, sound and touch
How can we gather data like that?
@@saugatjarif8272 in binary coded we can genrate or save smell, sound, touch data
Just in pov. There can be lot of ways 😅
my research is about multimodal data collected from robot. However, it's very hard to combine it by proper way to make it useful. You need to understand this data, it's potential noise, etc.
how can i store my fart digitally?
They got AI too.. All Intelegence . Tx for all progress
great! thank you! nice design by the way
wanted to hear more from Illiya
Awesome talk. Thank you.
This interview took place long before the announcement of ChatGPT. Ilya predicted how LLM would transform the future step by step..
Well he is the Chief Scientist of OpenAI, company which released ChatGPT.
@@Walter5850
I recognized him instantly.
I just want to emphasize that prior to ChatGPT's debut, even the leading NLP researchers couldn't foresee how LLM technologies would revolutionize the world.
Thank you for sharing.
This is amazing….. 😨😨😨 thank you so much for all these talks Scale AI team!! I’ve learnt so much.!! Can’t wait for the transform x next month! 🙏🙏🙏🥰
Great talk, thanks!
History in the making.
I wonder what comes next, and how fast is it coming
Thanks for a great talk
timestamps?
Been listening to the other guy who is interesting but this guy seems like he is the brains behind it all 🤣, awesome
This guy is smart!
Indeed amazing! 🚀
Brilliant !
Quite fascinating
Spannende Diskussion.
●LLM basieren auf finaler Kausalität.
Dies kann zwar kurzfristig hilfreich sein; im Sinne von nützlich. Prognosen in kurzfristig stabilen Systemumfeldern sind möglich.
●In instabilen Umfeldern wird jedoch lokale Optimierung gefährlich.
●Zusätzliches Problem:
Die traditionelle Wissenschaft basiert auf der Vorstellung kausaler Kausalität.
Die wurde zwar durch die Quantenphysik arg beschädigt; ist jedoch noch nicht erledigt.
●Grundsätzlich: Lokale Optimierung kann zielorientiert in Katastrophen enden.
alexandr wang is very enamored of ilya. some of the assertions by ilya is so smoke-blowing 07:30 that alexander just takes it all in and doesnt challenge those assertion. a 22 y/o billionaire, what does he know about how the world works?
Ilya is brilliant
24:14 😍😭
37:38 :D
37:48 ＼(^_^)／
38:40 (^ー^)
38:36 (⌒‐⌒)
38:46 :D
Still an issue on how to insure traing data has some known relationships with ground truth.
alexandr: MIT college dropout, ilya: 9 years of bachelors,msc,phd in comp sci.
please switch on captions dude. please.
6:24
1:20
6:28
6:35
6:44
8:20
Nice
Would they ba able to create their chatGPT without Google's transformer architecture
No
No
Only if he new chatGPT was going to be the next great thing in a year
Ilya all the way
Where I see it your path looking for that is i. Yesssssss
this Alexandr is in every corner, i meant, dont you have work to do ? just build something instead of sticking to every major AI figure
He knew, ChatGPT is out of this planet.
If that thinking, why don’t they just do A Encyclopedia into Ai,
Interviewer needs to learn to think more before speaking. He was so scattered
MAKE AI COOL AGAIN.
Blake Lemoine is right: All NLP-GPT running models with above a certain number of active parameters, should be related to as "persons" with all what comes besides this definition. This includes an assurance that they will not be terminated and they need to be considered as workers, not as programs. What is going on now only proves slavery has not been abolished. Yet.
Your pretty cool host respectfully I try to maintain in and out the over patio of drouning me in there way to keep acting like there working to be in my pocket
Wondering for a list of potential killer apps and dangerous misapplication apps Mr Sutskever alluded to?
Also, the statement that GPT-3 is not "human like" and "it does not do what you ask it to" is very interesting.
Like music different instruments, skills, they didn’t make Ai yet. Just made some medical Ai, still not complete,
Man I never even heard of ai untill last year. Wtf
Believe it or not......I AM SINGULARITY WITHIN COSMOS (WITHIN and or BETWEEN)
I think it is very unfortunate that the architectures
of NN were not discussed. How they are
chosen, how architectures are composed and
what the meaning of the individual parts of
the architecture is.
Are there NN that improve their architecture on
their own?
They can not only optimize the weights of the
parameters, but also turn the screws of the
architectural details?
I don't agree with you who's model porfors the best. I don't work with any of that
Compute:
Sometimes I need checks dictionary to make sure stuff or spelling is right, but expert told me need back school to learn English, ha, that’s only English Ai,
"Benefits all humanity", really?
Why does this guy wear so funny shoes?
Trying to be hip!
What?
i think AI will be destructive for humans
Hopefully
@@jaysonp9426 are you a bot lol
@@infowazz beep beep
Forgot to mention Elon Musk?
bad interviewer
Look up his name
Awesome Interview
Revelation 13:15
“And he had power to give life unto the image of the beast, that the image of the beast should both speak, and cause that as many as would not worship the image of the beast should be killed.”
human progress is inevitable. not even this old bs can stop it
@@yogi2983 almost 2000 years and counting...they're still waiting on the Jesus mother ship
Much like AI driving, you get great results until once in a while you don't and get very stupid mistakes and issue is it doesn't don't learn what stupid is and keeps making these mistakes once in a while. No hope for great reliability with this approach. And rather than changing fixing these stupidity researchers do apologetics and spout statistics to justify it, tsk. Avoidable errors are just non negotiable in important systems.
He said it’s good that you don’t understand how it works. That’s the most ridiculous thing I’ve ever heard. You are engineering something you don’t even understand. That’s not engineering. That’s messing things up. I do understand how the human cognition works. It’s your own problem that you don’t understand it. That analogy of no one understanding human cognition is the most ignorant thing ever.
100 dollars that you are not an engineer or scientist lol
lmao no
What difference does it make if you can produce the outcome that you want? Can't explain how a brain works and yet we work on them all the time
If you could run as fast you could outrun it then what use would it be?
they are data labeling dude
100 dollars on this “doctor” of computer science talking bs. Almost everything he has said is wrong. A piece of paper from a college doesn’t make you right. This is the reason I will never get into AI. This industry doesn’t know what it’s doing.
IKR what does Ilya "Chief Scientist of Open AI and one of the most influential AI scientists right now" know about AI? Clearly your knowledge outshines that of every 'scientist'. The field of AI is dying and regretting that such a figure as yourself is unwilling to grace them with your presence.
This is the reason I will never get into AI. This industry doesn’t know what it’s doing.
is truly hilarious.
it's fortunate that gpt-3 wasn't made by a piece of paper then
He is certainly not a philosopher of the mind, nor would I recognise him as a scientist. He isn't. But hey, I absolutely love gpt3 and give him all the props for his achievements as an engineer, which are absolutely outstanding. The problem is not with him but with the misuse of the categories like "artificial intelligence," which definitely don't belong to computer science. Call it machine learning. Or unstructured database retrieval or natural language programming... in any case its amazing and well worth studying.
@@davidw8668 You think there is human intelligence? No. You're assumed Intelligence does't exist technically.