Opinion on what occurs when we load, not an LLM, but a LLM + a Neural Net on
a low-error, high entanglement, quantum computer. Will this create a mind?
On Saturday, March 30, 2024 at 08:31:25 AM EDT, John Clark
<johnkcl...@gmail.com> wrote:
On Fri, Mar 29, 2024 at 10:28 PM Russell Standish <li...@hpcoders.com.au>
wrote:
>"There is a big difference between the way transistors are wired in a CPU and
>the way neurons are wired up in a brain."
Yes, but modern chips made by companies like NVIDIA, Cerebras and Groq don't
make CPUs or even GPUs, they make Tensor Processing Units, or in Groq's case
Language Processing Units, chips that have been optimized to do best not in
floating point operations but in large neural networks that all current AI
programs are. In the recent press conference where Nvidia introduced their new
208 billion transistor Blackwell B200 tensor chip, they pointed out that when
used for neural nets, AIs chips have increased their performance by a factor of
1 million over the last 10 years. That's far faster than Moore's Law, and that
was possible because Moore's Law is about transistor density, but they were
talking about AI workloads, and doing well at AI is what NVIDIA's chips are
specialized to do. I also found it interesting that their new Blackwell chip,
when used for AI, needed 25 times less energy than the current AI chip
champion, NVIDIA's Hopper chip, which the company introduced just 2 years ago.
And I do not think it's a coincidence that this huge increase in hardware
capability coincided with the current explosion in AI improvement.
> "In the future, I would expect we'd have dedicate neural processing units,
> based on memristors"
If memristor technology ever becomes practical that would speed things up even
more, but it's not necessary to achieve superhuman performance in an AI in the
very near future.
> "The comparing synapses with ANN parameters is only relevant for the
> statement "we can simulate a human brain sized ANN by X date"."
I don't see how comparing the two things can produce anything useful because
one is concerned with software and the other is concerned with hardware.
Comparing transistors to synapses may not be perfect but it's a much better
analogy than comparing program parameters with brain synapses, at least
transistors and synapses are both hardware. Comparing hardware with software
will only produce a muddle.
> "he [Kurzweil] said human intelligence parity (which I supose could be taken
>to be avergae intelligence, or an IQ of 100) [...]
AI passes 100 IQ for first time, with release of Claude-3
John K Clark See what's on my new list at Extropolis
lnm
--
You received this message because you are subscribed to the Google Groups
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to everything-list+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/everything-list/CAJPayv0TftwZ7N09tmzP5DDsyfRcmcL40Au1-4mzTG8anvTDpA%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to everything-list+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/everything-list/1070662887.2783537.1712099933698%40mail.yahoo.com.
