Talk:Eurisko

Just noting - according to the Engines of Creation link, Eurisko was developed at Stanford, not MIT. So I changed it. -- Ben-Arba 01:24, 26 August 2006 (UTC)

Importance

It would be good to have more information on this. It's one of the emblematic AI projects. --Eulen 07:39, 13 March 2006 (UTC)

Eurisko Source Code

Does anyone know if Dr. Lenat had published the Eurisko source code in any form? Failing that, has anyone seen it in person, and is willing to comment? Given the fame of the experiment, surprisingly little has been written regarding the details of how the program actually worked. --Asciilifeform (talk) 13:03, 18 January 2008 (UTC)

I'm curious about the source code too. As a taxpayer, I'd really like it if the Eurisko source code were placed somewhere in the public domain where I could easily find it. I did find the following in the endnotes for Engines of Creation:

Developed by Professor Douglas Lenat ... and described by him in a series of articles on "The Nature of Heuristics" (Artificial Intelligence, Vol. 19, pp. 189-249, 1982; Vol. 21, pp. 31-59 and 61-98, 1983; Vol. 23, pp. 269-93, 1984).

That wouldn't be source code, but it might be enough to reconstruct the program. - WillWare (talk) 15:54, 19 January 2009 (UTC)

I'm afraid the publications by Lenat apparently shed little insight for any would-be implementor. This, at least, is my impression from the AI literature parsing Lenat's work; ie. Ritchie & hanna, "AM: A Case Study in AI Methodology".

I can send you a copy if you email me. (I also have handy "EURISKO: A Program That Learns New Heuristics and Domain Concepts" & "Why AM and Eurisko Appear to Work" by Lenat.) --Gwern (contribs) 01:35 20 January 2009 (GMT)

Bad Link

The link with text “Traveller TCS” appears to be pointing to something completely irrelevant; the dates involved (mentioned here and on that other page) do not appear to match at all so they can't be the same thing. --Donal Fellows (talk) 16:29, 10 November 2009 (UTC)

What do you mean? The first listed Traveller edition is several years before the listed Eurisko date, and the second edition came some years after that tournament and with, the article says, major rule revisions (which certainly is plausible if Eurisko found good exploits). --Gwern (contribs) 20:56 10 November 2009 (GMT)

Conflicting reports

I had the same recollection about the composition of the Traveller fleet. I've got two sources, and there seems to be a primary-source article both are drawing on, one incorrectly. Or perhaps two conflicting primary sources, an interview and a published paper. Any help? Josh Parris 06:11, 10 January 2010 (UTC)

Perhaps they're talking about different victories; Eurisko won twice, after all. (The Engines of Creation describes the second as being "a very different fleet".) --Gwern (contribs) 19:53 10 January 2010 (GMT)

That could explain it, but on checking it doesn't (see below). Is there a third secondary source we can go to, or a primary source? Googling shows lots of parroting of the NewYorker text in blogs; but gbooks, gscholar and gnews don't show anything. In the academic papers Lenat published was there any mention of this - or the second - tournament? Josh Parris 04:09, 11 January 2010 (UTC)

I've run into physical limitations trying to acquire supporting documentation:

• http://www.eric.ed.gov/ERICWebPortal/custom/portlets/recordDetails/detailmini.jsp?_nfpb=true&_&ERICExtSearch_SearchValue_0=EJ304678&ERICExtSearch_SearchType_0=no&accno=EJ304678 - anyone got access to Scientific American, v251 n3 p204-09,211-13 Sep 1984, so we can see what Computer Software for Intelligent Systems (by Lenat, Douglas B.) has to say? You can't buy 1984 editions from the publisher.
• Lenat, Douglas B. "Eurisko: Discovery of Heuristics by Heuristic Search" Palo Alto Calif.: Heuristic Programming Project, Stanford University, Jauary 1982 -- which I'm lead to believe is only available at the Stanford University computer science department in paper form (and apparently Carnegie Mellon University Libraries too). Can anyone friendly with Lenat ask him to republish on the web (assuming he still has a 28yo paper)? Alternatively, can someone fetch, scan and arrange for electronic publication of the original paper copy?

Josh Parris 01:15, 12 January 2010 (UTC)

I'm looking at my Eurisko papers, and the description seems to be right. The first victory was with a swarm of relatively slow ships armed with tons of missiles, and the weird little lifeboat; the second victory was with a swarm of very fast unarmored ships.

I'm using "Eurisko: A Program That Learns New; Heuristics and Domain Concepts, The Nature of Heuristics III: Program, Design and Results", from Artificial Intelligence. (I also have "Why AM and EURISKO Appear to Work" and "AM: A Case Study in AI Methodology"; contact me for copies.) --Gwern (contribs) 02:03 12 January 2010 (GMT)

Current sources

“ Eurisko thought differently. “The program came up with a strategy of spending the trillion on an astronomical number of small ships like P.T. boats, with powerful weapons but absolutely no defense and no mobility,” Lenat said. “They just sat there. Basically, if they were hit once they would sink. And what happened is that the enemy would take its shots, and every one of those shots would sink our ships. But it didn’t matter, because we had so many. Lenat won the tournament in a runaway.” ” — NewYorker

This is explicitly about the first victory, for the second victory the article mentions that that competition measured "fleet agility" as a victory criteria. Josh Parris 04:09, 11 January 2010 (UTC)

“ Largely as a result, Lenat wrote later, Eurisko's fleet evolved into a highly unconventional one, "consisting almost entirely of small, swift attack vessels rather like PT boats and including one ship so fast and so tiny that it was virtually indestructible." ” — Understanding Computers: Artificial Intelligence, pg 83

The single superfast ship was described as the keystone ensuring at least a stalemate. This quote was explicitly about the first victory, in comparison the second victory is glossed over in this book, with no description as to the fleet composition or the rule variations introduced. Josh Parris 04:09, 11 January 2010 (UTC)

"Too technical"

@Quantum Knot: Which concepts in this article are "too technical" for readers to understand? Jarble (talk) 16:29, 28 April 2024 (UTC)

Put yourselves in the shoes of someone who doesn't already know what "Eurisko" or a "discovery system" and read the article. There is zero explanation other than "it consists of heuristics, i.e. rules of thumb, including heuristics describing how to use and change its own heuristics". Which is a really opaque explanation if you don't already know what it is. Quantum Knot (talk) 11:33, 1 May 2024 (UTC)