Talk:History of artificial intelligence/Archive

From Wikipedia, the free encyclopedia

Contents

[edit] "both" or "either"

Under section Experimental AI research, the last sentance of the second paragraph reads: "However, it has become clear that contemporary methods using both broad approaches have severe limitations." This is unclear to me. I think that "either" is meant for "both", but before rewriting, this pre-bold author-respecting newbie thought i'd ask. If it's meant to be "both", i feel this should be explained in the article. Thx, "alyosha" 22:55, 15 December 2005 (UTC)

[edit] 5th Generation

This article omits what is probably the most important series of events in AI during the 80's and 90's. The creation of the Japanese 5th Generation Project, whose aim was to produce computers "as smart as a man" programmed in Prolog, with their computing performance measured in LIPS (Logical Inferences per Second). The US started a project to compete with this, headed by Navy Admiral Bobby Inman. Both projects tanked horribly, and the AI community totally abandoned attempts to mimic human thought, and re-invented itself as researchers in expert systems and fuzzy logic. Marvin Minsky is extremely critical of what is called AI today, with its robot wars, and Roomba, the intelligent vacuum cleaner.

Deep Blue was not an artifical intelligence exercise, but a demonstration that massively parallel computing with custom ASICs could do by brute force what heuristics and AI failed to accomplish - play chess better than a human.

I strongly suggest someone find a copy of "The Fifth Generation" by Edward A. Feigenbaum and read it. The article currently reads as if Artificial Intelligence has been a rousing success, when in fact, research attempting to produce real artificial intelligence like HAL in 2001 spent billions of dollars, crashed and burned horribly, along with pretty much all of the AI movement. What exists now has much lower expectations, and no longer has any resemblance to the AI of yesteryear.Hermitian 10:59, 10 January 2006 (UTC)

I would support your creating a new subsection (possibly to be expanded into its own article) on AI failures and/or hype. That is certainly a part of its history. However, there is a reason this page seems enthusiastic; it lists only a subset of the succesful projects, and I think / hope people edit down the vain claims of researchers who pretend they have achieved more than they have. The truth is that we have learned a lot about intelligence from trying to build it, so know why the early hopes were ludicrous, and have a better understanding of our own capacities as well now as a result. And we are making progress, even if it is slower & steadier than we initially hoped. I am sure you could find a lot of learned people from just a century ago that wouldn't have believed a machine could do mathematics or play chess, even though now we know recognizing faces is just as hard & wonderful.--Jaibe 17:18, 14 September 2006 (UTC)

[edit] Recent DARPA grand challenge

Personally, I think this page could do with a well-written blob about Stanley and the other competitors who finally actually completed the desert race? It seems like something that will be somewhat of a milestone in history.


[edit] Query quote

Is this quote true: Douglas Hofstadter, in Gödel, Escher, Bach, pointed out that this moving of the goalposts effectively defines "intelligence" as "whatever humans can do that machines cannot".

Can anybody refer me to the chapter/section? --moxon 20:15, 11 January 2006 (UTC)

[edit] linking dates

An editor recently deleted all the date links on this page. Personally, I always thought that date links are silly & wouldn't scale (so I don't mind!), but I just thought people interested in the page might want to weigh in on this change. Personally I'd say the page looks more professional now. But I want to make sure we are following the norms. --Jaibe 17:10, 14 September 2006 (UTC)

[edit] Cleanup Tag - References

I added a cleanup tag for the chronologic history, surely we can find some references for these events. Bugone 00:14, 9 March 2007 (UTC)

[edit] Steady progress?

I'm not sure that the claim that AI has made "steady progress" can be justified. There have been at least two periods when AI seemed to falter: the so-called "AI Winters" of the late 70s and late 90s.


It seems to me a more neutral statement would be something like this: "Some believe that the quest for artificial intelligence has made steady progress since at least the 1950s[1] while others claim that progress in A.I. been slow or non-existent[2]." But I need to find the references. Dreyfus or Winograd might do for the second case, but they are kind of old now. For the first reference one could use Kurzweil or Moravec


The article should mention the discovery and resolution of major problems in the history of AI: Minsky & Paperts criticism of Perceptrons and how Hopfield and backpropagation resolved it. The discovery of intractability ([1]) and how "scruffy" strategies attempt to address it. The discovery of the "common sense knowledge" problem and how expert systems and CYC use very different strategies to overcome it. The devastating effect of the Lighthill Report and the ALPAC report, the emergence of commercial success in the late 80s and then the most recent 'winter' of the 90s and where they are now at Google and elsewhere. It seems to me these are the central events and would capture some of the boom-and-bust nature of the history of AI. For reference, there is AI: The Tumultuous History of the Search for Artificial Intelligence by Crevier and I can look for a newer one. If I have time I'll look into it. CharlesGillingham 22:20, 10 June 2007 (UTC)

[edit] Rewrite in Progress

I am preparing a complete rewrite of this article. It's at User:CharlesGillingham/History of AI. Once it's finished, I will move most of the information on this page to a new "Timeline of artificial intelligence" page, archive this talk page, and replace the entire article. I should be done before the end of this week. CharlesGillingham 21:53, 22 July 2007 (UTC)