solving equation of neural network effectiveness parameter! if we can predict topic for the token and then extracts size of data for that topic, we can tell if token is reliable or not. if model got lots of data for that topic then token reliable if not then unreliable.
Now anything more than a year old is dated in AI. Weird. You literally have to be a young teenager to think this is 'normal' progress, because November 2022 was the wake-up call for 99% of us, even those of us who pay attention to tech advances. I read up on AlphaGo, AlphaFold etc., was impressed by the eight months it took to develop the Covid vaccines (unlike the people who complained that it took _too long_) and was still stunned, because a chatbot hits differently than a drug discovery platform, which is too abstract for the present.
Neural networks, including those used in LLMs, are more like a computational approach or framework. They can run on classical computers and don't require quantum computers. Think of neural networks as a versatile tool that can be used to build various AI applications, including LLMs. Quantum computing is a separate field that might eventually be used to accelerate certain AI tasks, but it's not a prerequisite for neural networks or LLMs. In simple terms: Neural networks: A computational approach LLMs: An application of neural networks Quantum computing: A different computing paradigm altogether None of these are operating systems in the traditional sense. They're all distinct technologies that may intersect in future developments but are currently separate fields. I hope that helps clarify, to be fair it is a very technically complicated topic and I only understand the big picture ideas; the underlying math and model architecture has been slow for me to digest but I ultimately want to be able to hand-build my own simple models.
I love Ilya and Elon
Keep uploading, very interesting
Is it just me? I can’t help but notice he kinda has that bit of Justin Timberlake vibe
He does kind of look like him, I just don't know why he thinks that hair is a good idea. Like dude, even Demis Hassabis got rid of it.
🇧🇷🇧🇷🇧🇷🇧🇷👏🏻, Robots will do most of our jobs!
Not will but already has.
AI is taking over my job and now I am using AI to try to get a living out of it.
The world is becoming a lonely and scary.
When was this?
Nov 5, 2023 - afaik
The man is genius
very interesting and enlightening
solving equation of neural network effectiveness parameter!
if we can predict topic for the token and then extracts size of data for that topic, we can tell if token is reliable or not.
if model got lots of data for that topic then token reliable if not then unreliable.
good content - keep it up
Ilya left OpenAI in May 2024 so this interview is OLD OLD OLD
Ilya's last contact with the public was on Twitter, where he announced the initiative of SSI, his new company to achieve superintelligence.
Now anything more than a year old is dated in AI. Weird. You literally have to be a young teenager to think this is 'normal' progress, because November 2022 was the wake-up call for 99% of us, even those of us who pay attention to tech advances.
I read up on AlphaGo, AlphaFold etc., was impressed by the eight months it took to develop the Covid vaccines (unlike the people who complained that it took _too long_) and was still stunned, because a chatbot hits differently than a drug discovery platform, which is too abstract for the present.
date > Nov 5, 2023
you need to put this song while he is speaking -----------> music.th-cam.com/video/iSQcqtKY1cA/w-d-xo.html
He said bajan not asian.
Bayesian. Jesus...
0:40 I thought neural networks was the operating system for quantum computing? Did I get it wrong Again? LLM’s are like apps in this scenario.
Neural networks, including those used in LLMs, are more like a computational approach or framework. They can run on classical computers and don't require quantum computers. Think of neural networks as a versatile tool that can be used to build various AI applications, including LLMs.
Quantum computing is a separate field that might eventually be used to accelerate certain AI tasks, but it's not a prerequisite for neural networks or LLMs.
In simple terms:
Neural networks: A computational approach
LLMs: An application of neural networks
Quantum computing: A different computing paradigm altogether
None of these are operating systems in the traditional sense. They're all distinct technologies that may intersect in future developments but are currently separate fields.
I hope that helps clarify, to be fair it is a very technically complicated topic and I only understand the big picture ideas; the underlying math and model architecture has been slow for me to digest but I ultimately want to be able to hand-build my own simple models.
@@mattolsen3345 thanks for the thoughtful reply.