First it was chess. Now it is a popular TV game show. IBM scored its first victory against human champions in the practice round of Jeopardy.

Robot Apocalypse Draws Nearer With IBM’s Jeopardy! Victory

IBM’s Computer Wins “Jeopardy!” Practice Round

How soon will Skynet take over?

It’s somewhat narrow AI…but not as narrow as the programs that are embedded in our everyday lives.  It’s a victory for parsing, definitely.  Anybody who says there is not much going on with Watson should realize that Watson was using up to 4 hours of time to answer questions a few years ago.

Also check out Jetty’s posting in our business new area:

http://www.chatbots.org/conversational/agent/ibm_watson_jeopardy_tv_contest_quiz/

Actually, we should integrate the forum an business news areas a little bit more…, future, priorities ..that sort of stuff..

Erwin

Could this be a Google killer?

I don’t think so. It’s a beautiful semantic engine, that’s true, but I definitely sure Google will be able to build this kind of technology as well. And even better. It’s just a matter of time…

@Victor that’s an excellent point. I guess it depends on how well the technology can be scaled up to the level that would be needed by an application like Google but whatever level that may be, it could only be a matter of time.

If you look at queries people tend to type into Google it almost mirrors the question format for Jeopardy, i.e. lots of keywords about a particular item.  It could be an improvement or complementary technology for Google.

Andrew -yes, from what I read about Watson, it requires millions of dollars of equipment… yes,  “pocket change” for Google.  Also agree with you Erwin - google is fully capable I would bet, and time is all that it requires.

I wonder - will it be a big corporation like IBM or Google that will first develop a Turing-Test passing computer program?  I really wonder how much work Watson would take to take it from a “glorified search engine” into a system that can fully understand and carry a conversation?  I’m sure all of you have wondered that.  Exciting times within the next 5 years I’d say !!!!!!!!

Go Watson! Now kick butt on TV next month!

The thing about Google is that people have sort of been trained by the algorithm by now. That is, they formulate their google searches in terms that make google more accurate at bringing good results to the fore by drawing on their previous experience interacting with search engines. So if the algorithm were to change significantly—even if in a way that better supports natural language input—I could see a backlash from users who might get worse results for the type of queries they make.

Victor: I’d definitely be interested in seeing Watson become more generalized. There’s a lot of potential there. I’m wondering—what part of Watson’s speedup was a matter of more clever coding and how much was improved hardware? It might not bode well for our own chatbot attempts if getting Watson-level behavior requires millions in investment. Yikes.

@CR I do believe Google has trained us.  Interesting question you bring up is generalization.  I’ve seen programs get *less* generalized; I’d like to see a program get more generalized and automate that process.  I don’t know what the best bots are running on hardware-wise but with exponential speed increases that may not be a problem by the time some of us get our bots kicking butt.

As a Dutch citizen, it’s not possible to watch this show live. Does anyone have tips how to stream US-TV to my local computer screen in the Netherlands?

Any help appreciated…. sometimes it’s hard to be a European

CR & Toby, yes you bring up 2 very important points.    It will be seriously disappointing for us ‘very low budget’ bot developers if the only way to deal with natural language processing in any real time or near real time way, is to use massively parallel machine (Watson, using hardware with THREE THOUSAND cores!!!—according to https://www.ibm.com/developerworks/mydeveloperworks/blogs/benchmarking/entry/benchmarks_for_200?lang=en ).

I don’t know about all developers on chatbots.org, but that is slightly out of range of my chatbot development budget !!!  I may buy 2 or 3 other machines and try parallel processing, or a system with 2 cpu’s, each quad core.  Also, spending a lot of time on algorithm optimizations.  Then there is the human-hours of effort difference.  I can only put about 16 hours per weekend in my bot, while probably a dozen people working full time at IBM !  Even 5 people @ 40 hours a week = 10,400 hours per year, compared to my 16x52 = 832.  How much time does everyone else on chatbots ‘get to’ work on there bot?

Well, these days at most I can clock in an hour in the evenings about 4 times a week. And maybe a few hours on weekends. It’s also a matter of “quality” work time—by the point that I really get into my programming groove, it’s bed time. :(

How much does hiring a team of IBM programmers cost these days?

I usually spend an hour or more between reviewing notes from my last session, then planning the day of what to code, test, document and backup/verify backup.  So at least an hour of ‘administration’ without any coding.

Would anyone like to help the guys from IBM? Voluntering probably… I could make a phone call on behalf of a few of us. I guess they will be interested…

Ha ha, that would be great fun Erwin, but I don’t have time enough for the things I’m paid to do.