On Wed, Feb 21, 2007 at 12:28:38PM -0500, Richard Loosemore wrote: > Aki Iskandar wrote: > >I'd be interested in getting some feedback on the book "On Intelligence" > >(author: Jeff Hawkins). > > > >It is very well written - geared for the general masses of course - so > >it's not written like a research paper, although it has the feel of a > >thesis. > > > >The basic premise of the book, if I can even attempt to summarize it in > >two statements (I wouldn't be doing it justice though) is: > > > >1 - Intelligence is the ability to make predictions on memory. > >2 - Artificial Intelligence will not be achieved by todays computer > >chips and smart software. What is needed is a new type of computer - > >one that is physically wired differently. > > > > > >I like the first statement. It's very concise, while capturing a great > >deal of meaning, and I can relate to it ... it "jives". > > > >However, (and although Hawkins backs up the statements fairly > >convincingly) I don't like the second set of statements. As a software > >architect (previously at Microsoft, and currently at Charles Schwab > >where I am writing a custom business engine, and workflow system) it > >scares me. It scares me because, although I have no formal training in > >AI / Cognitive Science, I love the AI field, and am hoping that the AI > >puzzle is "solvable" by software. > > > >So - really, I'm looking for some of your gut feelings as to whether > >there is validity in what Hawkins is saying (I'm sure there is because > >there are probably many ways to solve these type of challenges), but > >also as to whether the solution(s) its going to be more hardware - or > >software. > > > >Thanks, > >~Aki > > > >P.S. I remember a video I saw, where Dr. Sam Adams from IBM stated > >"Hardware is not the issue. We have all the hardware we need". This > >makes sense. Processing power is incredible. But after reading > >Hawkins' book, is it the right kind of hardware to begin with? > > For the time being, it is the software (the conceptual framework, the > high level architecture) that matters most. > > If someone has naive views about the AGI problem, about the various > issues that must be relevant to the design of a thinking system (like, > if they have no comprehensive knowledge of both cognitive science and > AI, among other things), and yet that person has really strong views > about the hardware that we MUST use to build an intelligent system, what > I hear is "Hey, I don't know exactly what you guys are doing, but I know > you need THIS!". Hmmmm. Just keep banging the rocks together.
I wouldn't put it that strongly. Old skool AI used to think that speed
was largely irrelevant ("5 MIPS is enough for the human equivalent").
Other only look at speed of data manipulation (a 300 mm waferful of 65 nm
adders or ring oscillators is surely superhuman, then). If you want to
match the number of bits and the rate of their twiddling between our ears, and
track reasonably
accurately what's going on in there, we're somewhere in 10^17 words and
10^23 ops/s country, on that words. (You might think that's a high upper limit,
but
we know that there are cases where each single spike counts. You might
disagree, but let's work with these numbers. They're as good as any).
Any reasonable man looking at these numbers would immediately see a problem.
You can't do this in a sequential machine. Not in this universe. Others
would hand-wave, and mutter Moore. Moore is sure neat, but integration
density has only very little to do with benchmarks, especially memory
bandwidth minus CPU "speed" gap, which also happens to be an exponential
function. Nevermind that if we look at spiking memory access is unpredictable,
and you're back at the mice, frog and pumpkin stage, aka 5-10% of peak.
This asks for dedicated hardware. Carver Mead & disciples (Rodney Douglas,
Kwabena Boahen &c) would like to trace out circuitry in analog neurons.
This is a great idea in principle, but it falls short on connectivity over
large distances (there's a reason CNS went for spikes over cm distances),
and rapid reconfigurability.
A different approach would do everything in discrete logic, and mesh up
things by a packet-switched signalling mesh (notice that there's a very
close analogy between spikes and packets, and it's not a coincidence).
You can simulate very high connectivity spaces on modestly physical connections,
if the hardware is fast enough. The hardware is fast enough.
This doesn't have to be ASICs, since FPGAs with embedded RAM and a router
are pretty close and pretty dense.
Of course you could use hybrid approaches, too, which however require
lots of A/D and D/A, though on few-bit values, and only when you leave
the circuitry block.
But yes, we need very different hardware. I hope to see this on our
desktops in 5-10 years, at least the beginning of it.
> Having said that, there is an element of truth in what Hawkins says. My
> personal opinion is that he has only a fragment of the truth, however,
> and is mistaking it for the whole deal.
I was going to mention the three blind men, and the elephant,
but you beat me to it. The human intelligence is rough, hairy,
and wrinkled.
--
Eugen* Leitl <a href="http://leitl.org";>leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com
8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?list_id=303
signature.asc
Description: Digital signature
