The submission heading should really be changed to the title of the paper:
"Simulation and assimilation of the digital human brain"
Beyond that, they used 14,012 GPUs each with 16GB memory to simulate a human brain, which at retail is what? roughly on the order of $7 Million or so plus a bit for fancy interconnects.
Assume you invest two orders of magnitude more, e.g. $700 million, could you get it to think like a human baby? That's couch cushion money for the FAANGs.
I feel that the only obstacles to full AGI at this point are like the obstacles for getting to the Moon in 1961 - money and a dedicated engineering effort.
At best, they made a digital (aka binary logic) simulation of how they *think* the human brain works.
But every time someone *thinks* they have it all figured out, a new discovery comes along to show they really don't.
You can't simulate what you don't understand. AI is like a modern version of alchemy.
In the same way it is not possible to turn lead into gold using a chemical reaction, it may not be possible or practical to simulate human level intelligence using a digital logic playback device (aka a computer) of a finite, workable size.
2 comments
[ 2.9 ms ] story [ 16.4 ms ] thread"Simulation and assimilation of the digital human brain"
Beyond that, they used 14,012 GPUs each with 16GB memory to simulate a human brain, which at retail is what? roughly on the order of $7 Million or so plus a bit for fancy interconnects.
Assume you invest two orders of magnitude more, e.g. $700 million, could you get it to think like a human baby? That's couch cushion money for the FAANGs.
I feel that the only obstacles to full AGI at this point are like the obstacles for getting to the Moon in 1961 - money and a dedicated engineering effort.
But every time someone *thinks* they have it all figured out, a new discovery comes along to show they really don't.
You can't simulate what you don't understand. AI is like a modern version of alchemy.
In the same way it is not possible to turn lead into gold using a chemical reaction, it may not be possible or practical to simulate human level intelligence using a digital logic playback device (aka a computer) of a finite, workable size.