[FoRK] W3C sponsor, AI / supercomputing Ph.D. program, Silicon
Stephen D. Williams
<sdw at lig.net> on
Thu Oct 4 09:14:07 PDT 2007
Jeff Bone wrote:
> My recommendation:
> The Bay is overbought. I'm long-term short the Bay. ;-)
I wouldn't even think of buying real estate in the Bay Area. I'm
renting, plain and simple. I'd prefer a house, but my few days in the
Spring didn't find anything reasonable to rent.
As for the area as a source of work, education (in various senses),
opportunity, venture capital, etc., it remains the magical kingdom. Say
it ain't so...
Salaries are probably still depressed, the commute can be a killer,
etc., but in my situation I'll be immune to most of those problems
barring a disaster of some kind.
> Except for the increased personal existential risk, which is probably
> less than, say, driving a car --- I'd say that the prospective
> Washington gig sounds more intriguing; that doesn't satisfy the child
> / education issue, though. Hard to say what to do.
Been there, done that, have the suit and tie (well, I usually avoid
that)... There are definitely great things about the DC area. It
remains my second favorite all around place to be.
On the other hand, too much work here is semi-mundane application of
technology invented elsewhere and frequently there is a distaste for
> In general, I'm not much in favor of CS PhDs; they're generally less
> than useful --- Russell's one of the only guys with one that I know
> that can actually still write code. Seems to often ruin you for
> that. And the world needs another treatise on a typed lambda calculus
> with "novel" concurrency mechanisms like it needs a hole in the head.
I will always write code, although I have been working my way out of a
cold period, brought on by too many architecture-level projects.
Loosing coding skills is definitely something to avoid. Any Ph.D. work
I do will involve a lot of code.
> More useful than an MBA or yet another JD, but that's beside the
> point... ;-)
> AI at Stanford, MIT, or CMU --- probably the biggest exceptions to
> that rule. GMU is interesting; perhaps there's some kind of
> multi-agent / economics / AI thing might fly, given the pool of
> advisors to tap there. IMHO, once everybody else gets over the
> novelty of the "multi-core" / parallel programming thing that we've
> all known was coming for over 20 years and realizes Erlang / join- or
> pi-calculus is the wellspring for the next 700*, then the intersection
> between game theory, economics, practical game playing, and
> heterogeneous multiple agent systems stuff is going to become the hot
> topic. GMU's as ripe for groundbreaking work in that area as any
> place outside of Hebrew University in Jerusalem...
Yes, and there's a guy leading the Econ department who used to do AI,
among other things. I could get into that somewhat, although I'm not
quite ready for full immersion in Econ. If I wanted to end up on Wall
Street and make money that way, which has some interesting
possibilities, that might make sense. It's not an immediate choice however.
> (Understand all this educational advice is from a guy that didn't have
> the patience to finish his undergrad... :-)
;-) I'm mostly self-taught, except for the Masters work, so there you
go. I caught the bug at IJCAI/AAAI/IAAI 2003 and have the momentum from
the Masters, but I know that it is of dubious worth in the average
case. I'm not average and, as you concur, AI at Stanford/MIT/CMU is a
whole different thing. I see a boom / bubble coming, probably in two
waves, and since I cannot at the moment stomach a raw startup, this is a
different path into that space.
> Get admitted, fire up the ole' grant-proposal-writing grammar / text
> generator, feed it a bunch of relevant things from citeseer,  and
> fund the W3C participation that way. Or: write a decent poker bot,
> buy some rack space in Anguilla, and fund it that way...
It appears that being a student at Stanford will be enough to get member
access to W3C and I can get there by taking a NDP class (Non-degree
program), I think.
> $0.02, YMMV.
>  start here: http://citeseer.ist.psu.edu/zlotkin89negotiation.html
> * I reserve the right to be wrong about this. Indeed, I *hope* I'm
> wrong about it. I hope that the compiler writers can actually make
> concurrent Haskell a practicality for real-world systems. It would be
> lovely if the argument for explicit concurrency was as moot 10 years
> from now as the arguments against "high-level languages like C" and
> the use of compilers was by 1980.
I did Haskell... It would take a lot, a super-Eclipse or something, for
Haskell to be fun.
> But Haskell isn't the right vehicle anyway; the language designers
> need to get over the novelty and generality of monads, come back to
> earth, and realize that uniqueness types are the only way truly
> functional languages are going to make it in the mainstream. Paul
> Graham be damned, there's more to be gained by doubling the
> productivity of the lower 90% than by increasing the productivity of
> the elite 10% by 10%.
The right answer will be simple, fun, clear, semantically scalable, and
powerful. We're not there yet. A lot of crap results from
pseudo-solutions like the old requirement at NCR that Pascal functions
not be more than 10 lines of code. Try finding all of that
"non-complex" complexity spread over thousands of functions, files, and
modules and class hierarchies 20 deep. Blech.
swilliams at hpti.com http://www.hpti.com Per: sdw at lig.net http://sdw.st
Stephen D. Williams 703-371-9362C 703-995-0407Fax 20147 AIM: sdw
More information about the FoRK