I honestly think they're right insofar as the activity of thinking being "dumb" at its core, but that also describes humans - it's all binary decisions when you get down to it IMO. Of course, they're utterly wrong about what that means you can do with these things.
The jump from serial processing via traditional CPUs, was enabled via the growth of gaming and graphics where you had to paint a million pixels all at once. GPUs were made for parallel processing. Machine learning researchers took it from there.
We (wintel, intel/msft) kept trying to add parallelization to the CPU architecture and that plus faster clock speeds and more transistors was Moores law but it was incremental change not true parallel processing.
For a step change we needed a new way. This also why Nvidia is now the most valuable company in the world and Intel peaked 5yrs ago.
"The inside of a computer is as dumb as hell but it goes like mad!"
Replace "computer" with "LLM" and it probably reflects how many people see AI these days (especially AI critics and sceptics).
I honestly think they're right insofar as the activity of thinking being "dumb" at its core, but that also describes humans - it's all binary decisions when you get down to it IMO. Of course, they're utterly wrong about what that means you can do with these things.
01010100 01110010 01110101 01100101 00101100 00100000 01110100 01110010 01110101 01100101 00101110
01010111 01100101 00100000 01100001 01110010 01100101 00100000 01101110 01100101 01110010 01100100 01110011 00101110
01010011 01110000 01100101 01100001 01101011 00100000 01100110 01101111 01110010 00100000 01111001 01101111 01110101 01110010 01110011 01100101 01101100 01100110 00101100 00100000 01101110 01100101 01110010 01100100 00101110
My favorite Feynman quote: “If you’re the smartest person in the room, you’re in the wrong room”
The jump from serial processing via traditional CPUs, was enabled via the growth of gaming and graphics where you had to paint a million pixels all at once. GPUs were made for parallel processing. Machine learning researchers took it from there.
We (wintel, intel/msft) kept trying to add parallelization to the CPU architecture and that plus faster clock speeds and more transistors was Moores law but it was incremental change not true parallel processing.
For a step change we needed a new way. This also why Nvidia is now the most valuable company in the world and Intel peaked 5yrs ago.
I get into this towards the end of this post
https://open.substack.com/pub/wirepine/p/cats-made-ai