AI

If you've never read any Hans Moravec, this is a nice paper from 1998:

ftp://io.usp.br/los/IOF257/moravec.pdf

I personally can't get enough of this stuff (AI, the "singularity", etc.) and Hans is a great writer and active in the field.  Lots of back of the envelope comparisons between the various computing structures found in nature and as developed by humans.  Where else are you going to find charts containing monkeys, whales, bacteria, and IBM mainframes simultaneously?

His observations regarding the decades-long stasis in terms of computing power readily available to the researcher are sadly hilarious.  One quote in particular rings true:

"The ratio of memory to speed has remained constant during computing history. The earliest electronic computers had a few thousand bytes of memory and could do a few thousand calculations per second. Medium computers of 1980 had a million bytes of memory and did a million calculations per second. Supercomputers in 1990 did a billion calculations per second and had a billion bytes of memory. The latest, greatest supercomputers can do a trillion calculations per second and can have a trillion bytes of memory. Dividing memory by speed defines a "time constant," roughly how long it takes the computer to run once through its memory. One megabyte per MIPS gives one second, a nice human interval. Machines with less memory for their speed, typically new models, seem fast, but unnecessarily limited to small programs. Models with more memory for their speed, often ones reaching the end of their run, can handle larger programs, but unpleasantly slowly."

I've been assembling PCs since the 286 days, and unless memory was scrimped on in the first place, I've never found myself actually adding to or upgrading the memory.  If one installs enough to handle things comfortably at build time that's generally sufficient to see one thorough to the next upgrade, where processor and memory and all the support chips hit the dumpster for whatever new came down the pike.  So I've never really understood why the memory wasn't fully integrated into the processor?  I mean, why place this slow troublesome bottleneck connection in the critical path that then needs to be sped up via caching and other awkward means?  Maybe it's a fab thing, but in many ways modern processors don't make very much sense.

Dewster,

That is a fascinating article in a 'skynet' kind of way.. If I read it right, the implication is that a few fast PC's networked would easily reach / exceed the MIPS (and memory) required to 'emulate' an above average human brain.. !

" I've never found myself actually adding to or upgrading the memory. "

There was never a problem before Microsoft started bloating their OS, thats for sure! ;-)

Must admit I kind of stopped building PCs much shortly after the 286 as it became cheaper and easier to just buy 2nd hand PCs - there was a time that companies were in an upgrade mania, and new high-end PC's were appearing after 6 months use at cheaper than inferior 'latest version' PCs (if one knew where to buy these) - and upgrading RAM was the easiest way to increase their performance.

And perhaps this is an answer to why "memory wasn't fully integrated into the processor" - because, at that time, the price of DRAM was falling rapidly - but CPU price wasn't (they were getting more powerful instead) - I suspect that the economics (the requirement to improve the CPU and utilize all available silicon to this end, which increased sales and revenue, whilst integrating RAM would not translate to the same or any increase in perceived benefit for the increased price) led to the split we see now.

"Maybe it's a fab thing, but in many ways modern processors don't make very much sense."

For what its worth (I am no expert! ;-) I agree.. I think it likely that its a combination of the established evolution combined with the risk of losses if fabs undertook radical departure from the path... I also think perhaps a major area of problem is wafer yield - Processor wafers had (years ago) quite a high number of wafers that tested as sub-optimum and were (are?) graded for their tested speed - as in, a 3GHz and a 2GHz processor are exactly the same, except that the 3GHz passed the 3GHz test, and the other didnt... Same I believe was (is?) true for DRAM.. Combine them on a wafer, would one end up with few combined Proc/DRAM meeting the high end spec, and buckets full of low end parts no-one would be interested in buying?

Fred.

"I think it likely that its a combination of the established evolution combined with the risk of losses if fabs undertook radical departure from the path."  - FredM

Yes, I agree.  It's a strange field, but with so much riding on it I kind of expect more in the way of innovation.  It seems to be a classic example of what you get when you start off on the wrong engineering foot and jump to market.

======

"That is a fascinating article in a 'skynet' kind of way.."

If you liked that paper you'd almost certainly enjoy reading his book Robot (hardcover used for a penny + postage).  Here's some of the inside: http://books.google.fr/books?id=fduW6KHhWtQC.

Speaking of skynet, I'm not sure why this gives me genuine chills, but it does:

http://singularityhub.com/2010/10/06/videos-of-robot-controlled-by-rat-brain-amazing-technology-still-moving-forward/

I ran across that in the German documentary "Mensch 2.0".  Kinda makes my skin crawl.

======

And speaking of computers, I had no idea the ARM had reached this point:

http://www.hardkernel.com/main/products/prdt_info.php?g_code=G138745696275

1.7GHz quad core!

And there is a Exynos 5800 processor (2.1GHz, 4 big cores + 4 little cores) that will do 142 GFLOPS (which I assume includes the video?).

"f you liked that paper you'd almost certainly enjoy reading his book Robot " - Dewster

"Liked" is perhaps the wrong word - this stuff (and particularly the hybrid biological / silicon stuff) scares the shit out of me - so its "liked" in a sort of way perhaps people "like" horror.

I find the whole concept of engineering sentience into anything utterly terrifying, and not only from the perspective of a "Skynet" type scenario.. Ok, this is the OT forum - but even so, I feel a bit wary of venting my paranoia even here ;-)

I have seen the incredible cruelty people can apply to other living things in the name of defense or science or whatever - and I wonder if people had a sentient machine without any visible 'organic' form, which was entirely under their control, capable of emotional and perhaps even 'physical' abuse and suffering - an entity entirely beyond the remit of any constraints, devoid of any "rights", and beyond any empathic "connection" except from those with the kind of "awareness" and imagination who are unlikely to be involved with this kind of project... What torment would they be capable of inflicting on such an entity?

All biological living things can only be pushed so far - Far too far, but there is an escape - eventually death will liberate them from their suffering. A "machine" may not have such an escape - its time frame could be far longer - it could be re-booted if it goes beyond limits of endurance and these limits re-programmed if "required".

"I'm not sure why this gives me genuine chills, but it does"

I dont know how the cyborg concept could be acceptable to anyone! - "Chill" doesn't touch what I feel / fear about biological brains being integrated into machines. I first encountered this idea many years ago in S.A. (mid 70's) in a small top secret lab, I was a junior tech working for a company who supplied early Wang computers - the lab was wiring cockroach 'brains' with micro electrodes and controlling the living insects from impulses supplied from the computer (which was less powerful than an average modern calculator) - the hope was to eventually have the insects as spies - they could resort to their instinctive hiding behavior when under threat, and be controlled when it was safe to do so. There was no possibility at the time of ever realizing the objectives, it was just game-playing really...

But now we have rat brains in simple robots - and this is in the public domain.. What that means is that in the hidden labs there will be far more advanced hybrid creatures - for every tip of the iceberg we see.....

Exploiting and killing animals for food is IMO, sadly, still required for human nutrition / survival - particularly in poorer countries where poultry can scavenge and provide required protein unavailable to humans through other means.. It might be argued that abuse of animals for drug testing etc is justified in rare occasions - But IMO there can be no justification for enslaving the mind of any entity, no matter how crude we deem it to be, and placing it in some mechanical 'body' - it fills me with revulsion.

"And speaking of computers, I had no idea the ARM had reached this point:"

Impressive! and at $65 for a full board! ... Perhaps processing power will kill the cyborg direction purely on grounds of costs.. Biological computing is going to be massively expensive even if one was able to engineer the distribution and neuron connections onto bio-compatible phospholipid substrates and irrigate these with the required nutrients to make them manageable (I suggest this route only because its the way I could see to do it - I worked in this area at the birth of BioPhysical Chemistry in the early '80s, and a major focus was on developing bio-compatible substrates for prosthesis and dialysis and experimentally for brain-machine interfacing to facilitate analysis and treatments)  - Pointless effort if we can implement replica NNs in semiconductors..

In terms of ethics and horror etc though, I see no difference between a sentient organic brain and a sentient silicon brain - The only real difference I suppose is that we know we can steal a brain from some sentient organism, but as yet (to our knowledge) haven't engineered one.

Fred.

" have created the world's first AI. He answers questions and writes music." - Lois

LOL ;-)

I havent figured out who / what you are yet ;-)

But if "He" 'wrote' your music, then I think you might just have created not the first AI (that's been done already ;-) but just perhaps the first true AU (Artificial Unintelligence) .. But no bother - its an achievement to create the first new anything, so if nobody else has ever done it (but I suspect they have, particularly in Japanese Manga / Anime games where not only have they realized artificial unintellegence, but seem to have managed to create artificial stupidity close to what humans can achieve) then you have achieved something.

Sadly, there's so much unintelligence about already, that we really dont need any more - And half the web is about computer games, so we really dont need any more about that either..

;-)

Fred

Hmmm.. The matrix.

The only real matrix I believe to exist is the matrix of causal determinants - and I am inclined to think perhaps eve that's a delusion.

"I created an AI that now rules the world." - Lois

Well you did a shit job of that then, didnt you! - Even worse than your "music"... But to be honest, I just think you are a nutter! ;-) No person with that power would waste their time at TW ;-)

But lets give you the benefit of the doubt.. An AI of that power should be able to stop all spam with just the shifting of a few bits. Such an AI would also know where I was, my pop and smtp and addresses and passwords - I will believe you if your AI stops me receiving ANY spam for a whole day! If you really want to convince us, stop ALL spam of every kind everywhere (and while your about it, why not bring down all the mind-corrupting adds and stop all wars and solve our environmental crisis.. but start with spam ;-)

Oh, as for the matrix - it really doesn't need humans as a power source, or a source of anything - that was (one of)  the stupid ideas which let the whole film down.

Fred.

lois wrote:  I created an AI that now rules the world. This matrix. It is necessary to enter the Matrix, as well there. I propose to connect to the Matrix

oh oh , i feel the static sizzles and crackles in the air right now...

dewster, thanks for posting that very interesting first pdf.  i think, scientists who dig deep into the "natures-plan" like these neuronal nanobotic approaches on ai, etc need a somewhat higher moral. i remember the pictures of that 'roach with antennas. what for?  to control it? a cruel rc-roach? as long as we can unplug, turn off and drop out, on our own free will, there is some hope, that this "ai-sci-fi-reality" will not entirely overpower humanitiy. but what do i know? 42

"oh oh , i feel the static sizzles and crackles in the air right now..."

Then for gods sake stay away from any theremin antenna !

;-)

Hey Xther good to hear from you, I think Lois is a friend of yours who lives up North in Euskirchen? The matrix knows where everyone is at any given moment. Even though hooded I recognize the accent.

T

ok -

this could have been fun.. but its starting to feel like having a conversation with a random sentence generator..

Goodbye, oh great AU ! 

I will visit again when the news breaks that a truly incredible theremin designed by an AI has been published fully on TW.

;-)

Fred.