SUO: Re: Industry takeover
Jean-Luc,
Your observations are correct:
> What matters today is *acceptance*.
> Acceptance shows up as the number of users
> and it comes from either monopolistic enforcement,
> mediatic frenzy or any mix thereof, covering
> the spectrum from Microsoft/IBM/Java/ICQ/Napster.
But I would add two qualifications:
1. The word "today" is inappropriate, since hype
and frenzy have always been the most important
determiners of acceptance. For evidence that
they are universal, I suggest you look at the
history of religious movements.
2. But sometimes people buy something new when they
recognize that they have a problem that is not
being adequately addressed by the current solutions.
An example is the rapid success of Google over
the established search engines.
The importance of new problems in motivating change
is illustrated by the slogan that was popularized
by IBM salesmen:
You have to sell the problem
before you can sell the solution.
In any case, I don't have any illusion that a bunch
of axioms such as SUMO or OpenCyc will gain any
acceptance in the marketplace, even with the blessing
of the IEEE. The best it can do is to help some
companies win a few government contracts, which have
been the mainstay of Cyc for the past 19 years.
But Cyc itself has been a failure in the commercial
markets -- not because of a lack of hype (which it
most definitely had) but because it doesn't solve
any problems that anybody needs to solve. And I
certainly don't believe that a smaller version of
Cyc, such as SUMO or OpenCyc, will be successful.
But I also don't believe that a bigger ontology
(such as Google's new acquisition of Applied
Semantics) will do anything more than help Google
improve the precision of information retrieval.
That's not a bad thing to do, but it still won't
address reasoning, NL understanding, software
development, or a host of other problems.
The next question is whether IFF or the lattice of
theories could be the "magic ingredient" that would
solve, by itself, the problems that Cyc has not been
able to solve. The answer, of course, is no. I
consider it to be an important step in the right
direction, but it must be incorporated in a much
more complete package to be useful.
I am now working on a paper that addresses some
of these issues. But the short answer is that
a richer, more flexible framework is necessary.
An ontology is one component, something like
IFF or the lattice of theories is another, more
flexible reasoning methods, including both
deductive methods and analogies, are another,
and support for natural language is another.
Finally, you can't just dump all these things
into the same pot. There has to be a systematic
way of relating them, using them, building new
software with them and on top of them, and
most importantly providing a migration path
that will enable people to get from where
they are today to where you want them to go
tomorrow. For better or worse, that migration
path has been Microsoft's major (perhaps only)
advantage. Anything that ignores the migration
issue is doomed.
John