https://www.youtube.com/watch?v=53K1dMyslJg

I just watched this video and was very impressed. It provides a short but comprehensive overview of the current state of artificial intelligence research. It includes many many video clips from real life and science fiction too, so it’s quite entertaining as well as being informative.

Spoiler: most of us won’t see “Strong AI” in our lifetimes.

I enjoyed the quick overview in the first half. I just get put off by the 80’s story that rule-based AI doesn’t work just because those who tried found it “too complex” and “too much work”. The rules they wrote back then must have been very inflexible, to lead to such an assessment.

I liked that thanks

Thanks for the video link, but I wasn’t particularly impressed by that video. There wasn’t even any mention or artificial neural networks, which some people still think is the future of AI (that doesn’t include me, by the way!). The related PBS video that really fired my imagination, though, was the one about 3D printing, a topic about which I’ve been hearing a lot lately but until now I never read details about it:

Will 3D Printing Change the World? | Off Book | PBS Digital Studios
https://www.youtube.com/watch?v=X5AZzOw7FwA

Don Patrick - Jul 13, 2013:

The rules they wrote back then must have been very inflexible, to lead to such an assessment.

Not at all. Do you really think that in all these decades since the ‘80s people haven’t thought of better ways to write rules? I’ll repeat something I’ve stated a few times before: I believe the secret of AI is in representation, not in the language or the paradigm. Production rules are simply the wrong representation for strong AI.

Sorry, but I wasn’t talking about “strong AI”, otherwise I’d agree with you.

Strong AI is here right now:

http://ai.neocities.org/AiMind.html—in English;

http://ai.neocities.org/Dushka.html—in Russian;

http://ai.neocities.org/DeKi.txt—in German;

http://ai.neocities.org/mindforth.txt—in English.

http://ai.neocities.org/AiSteps.html is Steps to True AI.

http://www.chatbots.org/ai_zone/viewthread/240 has details.

Don Patrick - Jul 14, 2013:

Sorry, but I wasn’t talking about “strong AI”, otherwise I’d agree with you.

Sorry for the misunderstanding. I just started rewatching the video, and it does seem to be talking about strong AI, though it talks about the engineering difficulties in “capturing” (= representing!) the complexity of the real world, so the video nicely mixes the ideal (strong AI) with the practical (GOFAI). In that, I can see why people like the video: it does nicely bridge that gap.

Here’s a blurb about production systems I came across this weekend. (Now you all know how I spend my spare time!) It gives some specific reasons production systems are inherently difficult to work with, reasons I didn’t know about until now…

Counterbalancing the obvious advantages of production mod-
ularity, however, are two limitations. In the first place, not every
task is conveniently divisible into components that can be handled
autonomously; that is, sometimes the subtasks must be explicitly
coordinated, with particular information or instructions being di-
rected to particular units at particular times. Such organization,
while not impossible in production systems, is distinctly awkward
compared with “centralized” subroutine or function calls. Second,
production modularity is single-level; that is, in a production sys-
tem, you don’t build high-level modules out of lower-level mod-
ules, built out of still lower ones, and so on. All the productions
are on a par, except for priority. LISP definitions, by contrast, are
inherently hierarchical, and von Neumann programs often are.

(“Artificial Intelligence: The Very Idea”, John Haugeland, 1985, page 163)