Depends. Do they want to use AI to benefit humanity or secure their own classes power and wealth forever whilst the plebs are left to struggle or whittled down to save the environment for themselves?
I was a little disappointed at the explanation of how Ilya decided that GPUs were the perfect hardware to run AI training. It just boiled down to "it was perfect for it." I would have liked to have heard how GPUs were superior over other hardware, by describing the downsides of other hardware. I feel that Einstein's famous quote applies here "If you can't explain it to a six year old, you don't understand it yourself." And I have no doubt Ilya understands what he's talking about, but again, i was disappointed in the answer.
@@user_375a82 he just told you. a cpu can do a lot of differnt things, whatever system is deisgned for the chip. GPU does grahpics math on very many channels in parallel, meaning at the same time. It can process many many more than a cpu at the same time. A cpu can do many things very fast but one at a time or multi cores as many cores at a time for that section of hte chip. I think. Some one check me on my attempt to explain. i know enough to get in trouble quickly lol. Watch Asianometry channel if you enjoy lreaning about how chips are made. Very cool. I'm onthe software side mostly. And building servers and clients.
DEEP and LARGE people.🤷🏻♂️
The video ended when Ilya just started describing the most significant part
So much to reflect upon from this conversation.
Jensen Huang & Ilya Sutskever, 2 wonderful human beings.
Depends. Do they want to use AI to benefit humanity or secure their own classes power and wealth forever whilst the plebs are left to struggle or whittled down to save the environment for themselves?
In the center of this video are two men who changed my life!
@tjchatgptgoat: How so? What do you do?
@@ogungou9 I struggled in some areas with barriers of entry that were both removed because of Nvidia And it's amazing scientist.
Just a heads up there's many mistakes in your captions. "Bayesian models" as "Asian models" for example
Electronic and Informatic bots talking to each other 😂.
What is the date of this conversation?! Please.
cool
I prefer GPT 4 over GPTo, GPT 4 is better in reasoning and better in understanding user intent
Yep. Well everyone praising 4o is also praising the multimidodal version, which isn't public yet. The upgraded voice synthesis adds a lot.
I only use GPT-4o because of the price diffrence.
Where is the mention of ChatGPT-4o here? clickbait titles earn thumbs down
Older video than onmi model
I was a little disappointed at the explanation of how Ilya decided that GPUs were the perfect hardware to run AI training.
It just boiled down to "it was perfect for it." I would have liked to have heard how GPUs were superior over other hardware, by describing the downsides of other hardware.
I feel that Einstein's famous quote applies here "If you can't explain it to a six year old, you don't understand it yourself." And I have no doubt Ilya understands what he's talking about, but again, i was disappointed in the answer.
at the beginning... simply because GPUs were the most parallel & programmable commodity hardware around then.
I agree, but not sure Ilya is an expert in GPUs. How did Geoff Hinton know that a GPU could work well?
@@user_375a82 he just told you. a cpu can do a lot of differnt things, whatever system is deisgned for the chip. GPU does grahpics math on very many channels in parallel, meaning at the same time. It can process many many more than a cpu at the same time. A cpu can do many things very fast but one at a time or multi cores as many cores at a time for that section of hte chip. I think. Some one check me on my attempt to explain. i know enough to get in trouble quickly lol. Watch Asianometry channel if you enjoy lreaning about how chips are made. Very cool. I'm onthe software side mostly. And building servers and clients.