Technology is advancing at an exponential rate often called the "Law of Accelerating Returns." If futurist predictions prove correct, we'll have advanced molecular manufacturing by around 2025, and possibly the replacement of humanity by vastly advanced machines a decade or two later.
This is a chronicle of our journey to that future, one advancing technology article at a time. I post the more significant and interesting articles as I come across them.
Wednesday, April 28, 2021
New Artificial Neuron Device Runs Neural Network Computations Using 100 to 1000 Times Less Energy
There's a number of memrister methods. But, for perspective on just how ridiculously advanced computers could get, and even A.I.(and they still won't become conscious), Cerebras released their second generation chip a week or so ago.
The Cerebras CS-2 has 2.6 trillion transistors. The previous Cerebras CS-1 was said to be so fast, that it could calculate physical processes before they occurred. And, the Cerebras uses regular transistors - not memristers.
Even if Cerebras builds their future chips out of memristers, the future of A.I. still won't be conscious.
Every living thing is conscous. Consciousness involved the living process. The living process is non-equilibrium thermodynamics and chaotic dynamics. Things you can't simulate; well, you can only let evolve naturally.
Eric Drexler notes that a nanotech A.I. would basically be a million times faster in everything; but, then says thermal considerations maybe makes us slow the A.I. down or add some redundancy. But, with memristers, we won't have to slow them down so much.
So, his prediction of doing tens of thousands of years of science and engineering in a day is even more possible than he had thought!
3 comments:
There's a number of memrister methods. But, for perspective on just how ridiculously advanced computers could get, and even A.I.(and they still won't become conscious), Cerebras released their second generation chip a week or so ago.
The Cerebras CS-2 has 2.6 trillion transistors. The previous Cerebras CS-1 was said to be so fast, that it could calculate physical processes before they occurred. And, the Cerebras uses regular transistors - not memristers.
Even if Cerebras builds their future chips out of memristers, the future of A.I. still won't be conscious.
Every living thing is conscous. Consciousness involved the living process. The living process is non-equilibrium thermodynamics and chaotic dynamics. Things you can't simulate; well, you can only let evolve naturally.
Eric Drexler notes that a nanotech A.I. would basically be a million times faster in everything; but, then says thermal considerations maybe makes us slow the A.I. down or add some redundancy. But, with memristers, we won't have to slow them down so much.
So, his prediction of doing tens of thousands of years of science and engineering in a day is even more possible than he had thought!
Then, you add in quantum computers . . .
and what are these going to be useful for? To automate cars and trucks . . .
Post a Comment