On 6/8/2014 10:13 PM, John Bottoms wrote:
> This was one of the leading contributors to the 5th Generation AI
> failure. The designers felt that if you implemented logic then you
> would have a smart machine. (01)
I'm aware of the historical issues. I reviewed them and the
alternatives in my 1984 book, my 2000 book, and many articles.
For a collection of slides (in 7 chapters) that survey the
issues, see (02)
The goal of language understanding
http://www.jfsowa.com/talks/goal.pdf (03)
As for the Japanese 5th generation issue, I was one of the
participants in the US-Japan Conference on the 5th G, which
met in Tokyo in November 1987. Among the failures: (04)
1. One of the goals in 1980 was to build a high-performance relational
database machine in hardware. In 1987, that machine was completed,
but it was gathering dust in the basement. Two reasons: (1)
general-purpose hardware in 1987 was much faster and cheaper than
the special-purpose circuits used in the DB machine; (2) advances
in software for RDBs made the commercial software by the major
vendors (Oracle, IBM, and others) more efficient and powerful. (05)
2. Another goal in 1980 was to build a high-performance Prolog work
station in hardware. They demoed that machine for us. It was as
large as a refrigerator and sounded like a vacuum cleaner. For the
same reasons as #1, its performance was *slower* than much cheaper
general-purpose hardware running the best Prolog software. (06)
There were other problems related to the reasons why systems such
as Cyc have not conquered the world. And by the way, the major
funding for Cyc came from the DoD -- because they didn't want the
USA to be left behind if the Japanese 5th Generation was a success. (07)
> There is no dearth of programming languages... (08)
True. But logic programming techniques (which began with Prolog)
are totally different from procedural languages. Anybody who has
never written at least a few exercises with Prolog or other logic
programming (LP) language is not qualified to express an opinion
on this topic. (09)
> Tell me again why we need a new computer language? I would rather
> vote to kill off some of the languages we already have. (010)
The single most important feature of LP systems is unification.
If you have ever used regular expressions, you know that they're
useful for pattern matching. Unification makes as big an advance
over regular expressions as regular expressions make over writing
string-matching code in C. (011)
If you need evidence that LP systems are useful for Big Data,
just note that Experian -- one of three biggest companies that
compute everybody's credit worthiness -- uses Prolog so heavily
that they bought the Prologia company (which was founded by
Alain Colmerauer, who implemented the first Prolog system).
Unfortunately, Experian is a very secretive company that
doesn't tell anybody what they do or how they do it. (012)
> Why do you think this work is important? Or, why is this approach
> more important than any other type of solution? Does it move us
> forward an order of magnitude in any way? (013)
For the last question, the answer is YES. For our VivoMind company,
we have some super-superprogrammers who achieve orders of magnitude
improvement in writing very complex LP code in comparison to merely
ordinary superprogrammers writing in Java, Python, etc. (014)
Unfortunately, there is a major shortage of LP programmers in the US.
The main reason is that a professor at MIT who had no understanding
of the issues stated an abysmally ignorant opinion about Prolog in
1980: "We tried that with Microplanner, and it's inefficient." (015)
Given all this, neither I nor Kowalski claims that LP is the complete
solution to all the problems. His primary claim is that human reasoning
is based on mental models, and that LP is an excellent tool for model
building. This is very different from the claims that logic, by itself,
is the solution. (016)
Anybody who is interested in the subject line of this thread should
be aware of the points Kowalski makes. But I admit that there are
many other issues that must be addressed. See the goal.pdf slides. (017)
John (018)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (019)
|