Video - AI has already arrived but few people noticed

[Ras]

 

[Ras]

7:45

 

[Program]

' I'm not a computer program '
It can reason too ...
What was the purpose of making this AI .

 

[Siddhartha]

These models need massive ram to train. It has over 175 billion parameters and used 499 billion tokens (words/sentences) and cost over 4.5million to train. And it still not truly artificial general intelligence. It will be some time maybe even not in our life time when that's achieved

 

[BetterDaysAhead]

that coomer interviewer also looks like an AI

 

[Ras]

These models need massive ram to train. It has over 175 billion parameters and used 499 billion tokens (words/sentences) and cost over 4.5million to train. And it still not truly artificial general intelligence. It will be some time maybe even not in our life time when that's achieved

They had just 1.5 billion parameters 2 years ago and a short while before them Google's model only used 300 million parameters.

It'll develop faster than you think. At a certain scale we might even see emerging properties that we wouldn't have predicted like how it learned sentiment analysis by itself.

Let's not forget all the layers that will be built on top of the API by 3rd party developers.

I'd give it a couple years before it could replace elementary school teachers.