who is OTM about artificial general intelligence?

more than likely we won't know it when we see it because it will come from such a foreign environment.

i think something like colonialism studies could help on this problem

I'm inclined to think that it is an emergent property of complex systems (in the modern scientific sense of both words), but that does not necessarily imply that it will therefore emerge as we make computers more and more complex. Or that what intelligence does emerge, if any, will be comprehensible or useful or usable by us.

The history of ideas of what might constitute artificial intelligence is quite a long one. I ended up seeing the last 20 minutes or so of a discussion at the NTK/xcom thingy last year which touched on this. One of the people on the panel had written a book about 18th century automata (robot ducks ahoy). These were at the time considered to be very close to proper self acting agents, in the same way that yr average Turing testee is now. He made quite a convincing case that we were chasing a constantly moving AI target because no-one's really clear what it is.

of course our model is ourselves, and it's ouyrselves at our most bonkers!!

I'm out of the current loop on this stuff because it's been a good five or six years since I've read much of the new stuff about it, but when I was in school I spent a lot of time on AI, particularly from the philosophical/computer science end of it. (i.e. I didn't build actual robots, just read and wrote a lot about AI in theory)

Some of the more well-known objections to the Turing test were penned by John Searle. I found this one with a quick google, but there are several by Searle and arguing that a computer doing a good job of impersonating a human does not qualify as AI.

But the original question was about Artificial General Intelligence, which is a different thing altogether, and a system designed to impersonate, say, a person using AIM could qualify as an expert system under plenty of circumstances.

John McCarthy, you were OTM about artificial intelligence. Thanks for the good times back on the Farm. RIP.

RIP

Yeah, saw this yesterday evening but didn't see a good obit to post then. RIP. Read a few of his papers as an undergrad, they were invariably fun and confusing.

http://i.imgur.com/Dj19a.png

looks like a DUDE, wish I looked half as good at that age. rip

He was so rad, always with his bag he got from flying the Concorde. He didn't look quite that good when I knew him, though.

McCarthy did some amazing things to increase the breadth of usefulness of computers. RIP.

As for General Intelligence in computers, that will come when computers have acquired much better ways to accumulate detailed sensations and can form them into a rich set of experiences from which they may learn over time, and when they survive long enough to learn about the world in depth. If programmers are required to stuff them full of knowledge, inventing an algorithm for every 'thought' they have, then there will be no General Intelligence at the end of the AI rainbow.

I think AI people know that. The real problems are all in your first sentence, with the words "acquired", "accumulate", "sensations", "form", "experiences", and "learn".

Yeah. Humans are compact, mobile and exquisitely equipped with sensory feedback mechanisms, to the point where our brains are so flooded with data that they discard far more than they use. Then we spend years upon years training our brains, just to get to where we can have a good enough grasp of the world to where we can solve "simple" AI problems.

as long as we believe the myth that human intelligence is somehow, like, metaphysically unique from other intelligences (simian, cetacean, porcine, etc), then we will struggle to create an artificial intelligence. or, rather, the turing test really is kind of a terrible benchmark.

but i'm guessing the AI ppl already know this stuff

aaaaaand i actually scrolled just a tiny bit upthread and i see that someone covered this

and maybe i've watched too many terminator movies, but i'm not sure that the first AI will be created from whole cloth, or even recognizable as intelligence at the outset. like, i'd put money on the googlebot (conceived here as the aggregate form of all the various google technologies) starting to exhibit sentience before some lil robot designed by a few guys in a lab.

our intelligence was produced by the accretion of accidents, sorta stands to reason that an intelligence for which we are responsible will be similar; AI will be a product of humanity, not a human. imo.

*reads omni*

what's the best learn LISP in 28 days for dummies book?

Since participating in the kurzweil thread, I've been thinking a lot about artificial intelligence. What are some good (or great) books about A.I.? They can be theoretical or slightly practical or historical, but I don't know calculus or how to program, so I don't need anything especially technical. A little bit technical is ok.

Russell and Norvig's Artificial Intelligence: A Modern Approach is the standard big AI textbook but it's aimed at CS students so it is kind of technical with pseudocode and so on.

Haha. I have an early edition of that and it's really hard to understand.

I don't know the field but this seems interesting, bringing a philosophical perspective to bear, and suggesting that computational and representational approaches are doomed to fail. (n.b. I have not read any Heidegger, an omission I am feeling an increasing need to rectify.)

Why Heideggerian A I Failed and how Fixing it would Require making it more Heideggerian

(My attention was brought to that article by a bookmark I glimpsed in a colleague's browser during a presentation!)

for the source of Heideggerian critiques of AI, see Hubert Dreyfus' classic What Computers Can't Do.

Aye, that paper's by Dreyfus! Might need to get that book - after boning up on Heidegger.

iirc you don't need to know any Heidegger to read Dreyfus' book; in fact it's a pretty good introduction to Heideggerian thought without having to read any Heidegger.

I think I'm approaching that Heidegger time in my life anyway though.

Now I ain't sayin' she a gold digger
But she ain't messin' with no Heidegger

I read the revised version What Computer's Still Can't Do. Don't know if he revisited and rerevised it.

I borrowed and never returned a copy of "Artificial Intelligence" by John Haugeland . It's a good overview.

http://www.amazon.com/Artificial-Intelligence-Very-John-Haugeland/dp/0262580950

(probably a few years behind the curve now though)

https://www.youtube.com/watch?v=AplG6KnOr2Q

what's good to read if i want a simple overview of the math and cs behind neural networks / machine learning stripped of all hopeful / obfuscatory lingo pertaining to neurons and learning?

reading a bit about d e e p d r e a m tonight i was struck by how crummy every discussion of neural nets / machine learning i've encountered as been. i never really cared to learn much about it, popped up here and there in philosophy of mind-related contexts, but it sure would be nice if i could've just been given an account of the machinery for approximating functions / solving optimization problems. instead it's all gotta be goddamn science fiction every third word : /