1. One of the articles I read around the time of the competitions said that Watson buzzed in with a mechanical thumb–it had to actuate a motor in order to press the button. Was this article mistaken/did I misread?

2. I don’t think that IBM would have had an easy time collecting the relevant primary sources for their databases back when the Internet was still so immature. I want to point out that one of the developers of Watson mentioned in a lecture that long-standing formalized hierarchical knowledgebases were looked at for potential inclusion in Watson but were rejected as parallel searching of an immense body of unorganized knowledge could be done sufficiently quickly while making it possible to introduce far more information.

3. I think physical compactness is an irrelevant question in a world where all one’s processing and storage can be done in the cloud. The better question to ask is “how long until cybernetics advances enough that every person within range of a wireless router becomes a walking jeopardy champion?” And a better question still: “What would such a world be like?” It’s unfathomable. My guess: less than 50 years.

4. (The next great milestone for AI)
I like the idea of the computer that can be given a (formalized) English description of a board game, time, and a number of good opponents, and become a reasonably good player. That is probably a good 10 year project.
For 20 years, a computer that can watch a video of a person and report 1) what they are doing, 2) their current emotional state 2) their underlying reasons for doing it and their probably internal planning process and 4) a prediction of what they are going to do next. In other words, a computer with a theory of mind. What would a good publicity stunt be for this? I don’t know, but I’d love to see it. For 50 years, an android fully capable of all the speed, flexibility, and energy of human motion can play a sport that requires this sort of agility, even if it requires a nearby supercomputer and an array of cameras to control it. Robotics tradition says that sport should be soccer, but that’s not an individual sport and so doesn’t show off the abilities of the android well enough. Tennis seems like the right idea here, but I would be more impressed with racquetball due to the increased number of possible trajectories, and the add difficulty of having to avoid running into your opponent.
And for 100 years? That’s easy. A working quantum computer.

There have to be people smarter than me who have worked with some rigor along the lines of the notion that passing a Turing test is dependent not just on performance of the AI, but also on the state of testing technology at that time.

If anyone could direct me analysis in that direction, I’d appreciate it.

The topic this week on Dick Lipton and Ken Regan’s weblog is Happy St. Patrick’s Day—Again Again. This topic has provided a long-awaited venue for a end-to-end complexity theoretic / quantum informatic / medical roadmap.

The post on Gödel’s Lost Letter and P=NP is the penultimate step toward migrating these long-in-preparation questions onto TCS StackExchange and MathOverflow, and I am hopeful that that this migration can take place with a maximum of creative participation and enjoyment for everyone.

That’s why comments and criticisms from Shtetl Optimized readers too would be sincerely welcome and greatly appreciated.

Scott, I am pretty sure that your post is referring to Hartmanis’ monograph Feasible computations and provable complexity properties (1978); this is a terrific monograph (IMHO). Moreover, Hartmanis followed-up with a similarly wonderful roadmap article Observations about the development of theoretical computer science (1981).

Certainly you and I entirely agree that Hartmanis deserved both his Turing Prize *and* his honored place on your “Timeline of Computer Science.”

My roadmap database has long included several juicy TCS-related quotes from Hartmanis’ 1981 article, but to date, there has been no suitable occasion to deploy Hartmanis’ quotes in the blogoshere … oh well, as Dick Liption says.

It is both instructive and enjoyable to read Hartmanis’ early work side-by-side with (for example) a modern textbook like Arora and Barak, for reasons discussed on Gödel’s Lost Letter and P=NP. In particular, a recent highly-rated result on TCS StackExchange, namely Emanuele Viola’s proof of the undecidability of runtime estimation in P, is a special case (AFAICT) of a theorem (it is Theorem 7.4) that Hartmanis proves in his monograph.

Obviously, it’s no surprise that a non-expert like me was unaware of Hartmanis’ early work … what is surprising (to me) is that no-one on TCS StackExchange recognized it … including the 26 recommenders of Viola’s Theorem.

One lesson (perhaps?) is that it pays to read the roadmaps of Turing Prize winners. And it’s true also, that Hartmanis’ 1981 roadmap for TCS makes very congenial and inspirational reading, for systems engineers like me. Good!

Comedy is one of those disciplines whose practitioners tend to improve with age … I have reason to be hopeful that quantum systems engineering will prove to be similar.

As for funny computers … it seems to me that kindly comedy is the most sophisticated of all forms of cognition. In particular, we will know that strong AI has arrived, for sure, when our computers begin to chuckle every time they speak with us.

-you have great taste in comedy, but I’m surprised you chose such old comediense? Are they your age cohort, because I thought you were much younger.

No one knows how to write down a discrete version of the laws of physics.

No one knows how to simulate the laws of physics on a computer.

Where by “laws of physics” I mean those that have already been established, in particular the Standard Model. I could well imagine that the first problem above is solved but not the second (there is still something called the Fermi sign problem to overcome). I could also imagine that both problems are solved. But, at the present time, both of the above statements are true.

http://www.fqxi.org/community/forum/topic/897

I’ve been wondering about this.

But a drive-by quote-pasting isn’t going to cut it. Read one way, what Pattee writes is obvious and unobjectionable (DNA, text, computers, brain, etc. obey “higher-level organizational laws” that can be mostly separated from the underlying laws of physics); read another way, it’s just as transparently false (nothing in the physical world is “entirely free” of physical laws). In neither case do I see anything even remotely resembling an argument against AI: a computer can be governed by “higher-level laws” just as surely as a brain can, and Pattee even says that!