f13.net

Been a long time comin', but finally a supercomputer will be playing against humans (Ken Jennings and Brad Ratter btw) in a few matches of Jeopardy next week.  Its name is Watson and it's from IBM of course.
Just recently was a special on NOVA explaining the technology.  And no, it's not a cloud computer like Wolfram Alpha (which was a bit of a disappointment imo).

http://www-03.ibm.com/innovation/us/watson/

Having seen the demo of it:


Funny, because it seems to be designed to narrow down google results. That sentence above is like calling a janitor a custodial engineer.

The sentence has technobabble fluff but don't underrate the difficulty of getting an AI to rapidly and *accurately* narrow down Google results in response to a natural language query. (With or without the weird formatting involved in Jeopardy.)

A sufficiently sized rules engine? Not anything technologically new or astounding from what I can see.

If it could provide an answer to a question that we don't have an answer for, that would be interesting.  It's what I do every day and I've always said that a trained monkey computer could do my job.

I bet this tech could be applied to medicine... hmmm.

(http://t1.gstatic.com/images?q=tbn:ANd9GcRDrMrQ4IkkiEOSCQsXPGhp1_bcxrGrJ-lu1NlJ2cBnFnN1gRy7&t=1)

Diagnosis is one of the classic examples of an "expert knowledge system" in AI.  If you take out the whole "natural language" part it's actually really easy; it's not hard to write a piece of software where you basically check boxes for symptoms and have it crank through a database of a few thousand different diseases and tell you which ones match your symptoms.

What makes this particular tech impressive is that it's not working with a neatly formatted question in the form of a series of tick boxes that correspond to some neatly formatted data that it's going to work with.  It's working with an English sentence as its "question" and what I presume is the entire Internet as its data, and it's got to decipher all of that to come up with a useful response.

I'm pretty sure a lot of diagnosis is already moving towards automated systems, at the very least for checking symptoms.

Watson isnt a cloud computer.  It's self-contained, offline, and curated.  It doesnt have the benefit of Google or weighted search engine results like something like Alpha has since every piece of data was manually (errr, I'm sure it was automated somehow) entered. 

Actually I'd say the tough part of a good diagnostician is being able to properly translate the patient's declared symptoms into the appropriate medical terminology and also to read the patient sufficiently to get a basic idea of their condition. Of course I've no idea how good most GPs are at that this but it would seem the tough part of writing a diagnostic medical programme would be ensuring it got the right input, which is really what most GPs have been reduced to.

Yeah, though to continue with the digression, I think it's pretty well understood at this point that what we've done to GPs combined with long-standing weaknesses in how we train doctors in the first place has seriously compromised diagnostic effectiveness. E.g., we're getting to where an expert system can stand in for GPs in some ways, but only because we're at a suboptimal point anyway. Really good diagnosis involves a lot of emotional and social skill, getting people to talk about what they're feeling in a way that brings out the relevant details and information.

Its an impressive machine, no doubt about it, but frankly, I think people are making slightly too big a deal about it.  Yes, its language ability is not to be under rated, but I've seen so much "SkyNet is NOW" stuff the last couple days  (with only a little exaggeration), that I feel like people just need a bit more realistic picture of what Watson can do.

Apparently IBM feels that Toronto has succeeded to the US?  :oh_i_see:

*insert pithy 51st state comment here*

The demo seemed to be more impressive than the actual performance but that might have been due to the fact that I was under the impression that Watson was actively listening to the announcer and not been sent a text with the transmission. The machine is impressive but after watching the first two nights it's readily apparent the machines won't be taking over anytime soon. Well, Cleverbot is worrying me a bit...

I haven't been watching on TV.  Is the chatbot making a good go of it or is it just embarrassing?

If you mean Watson ,he has stomped all competition.

Honestly, I'd rather see Watson in a pure trivia competition.  Jeopardy is nice and all, but since it's so 'punny' and the questions sometimes awkwardly phrased or based on wit and humor, it's tough to guage how smart the machine actually is.  I bet if it competed in a standardized academic competition we'd really see him shine.

Writing software to decipher the Jeopardy "questions" is, to me, a much more interesting challenge than just packing a database full of academic trivia questions.  I mean, if you want to see a machine doing things that machines are obviously good at, why not enter it in an arithmetic competition?

Yah, the challenge wasn't really to make a machine that will win trivia competitions.  It was to make a machine that can decipher natural language with all of it's puns, metaphors, hyperbole, etc etc.

It's still doing remarkably well, but it's still struggling dealing with some of the more difficult linguistic parts of the show.  It also has the limitation of being deaf and blind, so it can't respond to information given by the answers of others. 

Yeah, the phraseology of Jeopardy questions was exactly the point of this test.  Think of it as a stepping stone on the way to having audio computer interfaces like Star Trek instead of having to phrase them in exactly the correct structure like Google.

I'm generally against this sort of training for computers.  Next thing you know they'll get all
(http://www.pamil-visions.net/wp-content/uploads/2009/04/skynet.jpg)

on us.

I used to be worried about it.. then I realized that's just the next stage of evolution if it does happen.   Of course, I'm a big fan of cybernetics and artificial parts vs organics.

If music didn't happen, I was originally going to go into biochemistry with the aim of creating cyborgs for deep space travel (for real, I even did a semester while the band was on hiatus).

Skynet, indeed  :drill:

I'm just kidding around.  

People worry about computers taking over, e.g. Dune and Terminator, but I suspect that over time computers and people will sort of meld together, just as you're suggesting.  

Edit:  I always thought that this would be a cool book to read:  singularity (http://www.singularity.com/)

I've never gotten around to it, since my free reading time has been significantly curtailed in the past 5 years. 

Shitty book, but cool concepts.  Kurzweil himself is a carpetbagging charlatan though.  Peddling his "futurism" all over the country for exhorbitant costs.  Nothing he posits is really new and he rarely explains exactly how these technologies will come to pass.  I wouldnt hate on the man if it didnt cost a small fortune to merely be in his presence.  That, and one gets much more out of spending time with a hardened astrophysicist or comp. engineer than guys like Kurzweil... and you can do it for free most of the time.

That is unfortunate.  I guess I'll avoid it with that review.  It's not like I have time to read much now anyway.

You can do a google search on "singularity" and get a large chunk of what he writes about in the book, from his own and others' perspectives.  Might fit better into limited time as well.

I've done that.  My father in law and I were just discussing the subject last night.  I believe that 2045 or so is when skynet will become self-aware, haha.  As for the book itself, I was probably just suckered in by the clever blurbs on the dustcover. 

It's totally okay though, as Bunk mentions above, when Watson finally decides to nuke the US he'll hit Toronto????? instead.

Nah, by then I'm sure Toronto will have "succeeded" back to Canada.

(I for one welcome our new computer overlords)

Is Watson really "smarter" or just quicker with the buzzer?

There were some questions on the second night where Watson had a fairly positive idea of the answer, but didn't buzz in first. It appeared that questions that didn't have straightforward and obvious keywords took Watson a little longer to sort through. I missed the first night, so I didn't get all the details. Is it running off of an inhouse db, or is it actually webcrawling?

And yes, I know succeeded was wrong (who would ever intentionally use that word in reference to Toronto?)

Hes a contained system. No web crawling.

Jennings actually wrote that as part of Final Jeopardy last night.  Was pretty funny.


For certain definitions of smarter, it's smarter.  (In the same way those high-functioning Autistics who can spit out huge math equations are smarter)   In the end it doesn't matter, because computers WILL wind-up smarter than us within our lifetimes.   The edge we have (as has been pointed out in a few articles this week) is creativity.  We can't figure out how to make machines more than processors.  Yet.

I only watched parts of it - seriously?  They didn't pull from the web at all?  That just seems... dumb.  They basically programmed a talking encyclopedia that gets shit wrong then. 

(http://cdn04.okcdn.okmagazine.com/wp-content/uploads/2011/02/Watson_Jeopardy_Feb16news.jpg)

Hey, if it doesn't work 100% perfectly on the first time out it's clearly a failure and waste of resources.   :awesome_for_real:

They likely did it a favor by not exposing it to the web. Can you imagine? They let anyone and anything on the web. Every third answer would have been shitcock.

That is a gross misstatement of the technical subtleties involved. It's a self-contained mirror of large, pertinent parts of the internet, pre-selected based on the jeopardy category and indexed similarly to the way google crawls the real internet. Don't discount the amount of technical sophistication, because developing algorithms and the backend to provide near realtime indexing made google billions upon billions upon billions of dollars.

It isn't just a talking, fallible encyclopedia. It's really more like an entire search engine server farm, only smaller, because the subset of pertinent data is less and it only has to parse one query at a time.

Watson wasn't connected to the Internet. It had its own database.

The human mind works context based and associative.

The big step forward for me lies in the natuarl language processing capabilities of Watson.

Human Language is very context sensitive. I'm not even certain that a grammar for any human language would even lead do decidable/computable problems.

Watson has to actually parse a spoken question decide on the most likely context that question is phrased for and then use some for of associative memory to infer the answer.

If you would ask the question to a human operator that knew the right context and were able to then search for the question on google the most work would have already been done once the operator enters the search terms.

So I'm less impressed by the data mining capabilities of Watson (that could be brute forced if necessary) but by actually arriving at the right terms - given a deliberately obscure question - as input for the search given only a spoken query in human language form.

Parsing human language and context selection is one of the hardest comp-sci problems around (PSPACE even, IIRC) and many pure computing approaches are unfeasible because it either takes to long to arrive at the 'correct' context and sentence or the grammar is even undecidable.

So that Watson is able to understand human language well enough to actually play Jeopardy! is an achievement in its own right. The associative memory and efficient lookup is just the icing on the cake.

(http://i215.photobucket.com/albums/cc251/Lorekeep/Meme-1.jpg)

Ok, that actually made me lol here at work.  Now there on to me that I might not actually be working on the web site.  Thanks!

Plus, and this is very important, it doesn't just give one answer really. It was giving just one on Jeopardy because that's what the format requires, but it can output a list with weighted probabilities, which in a lot of ways is much more useful. There's a lot of interesting applications for this.

Well, if it was getting the answer correct, it was still impressive.  Think about Google's "I feel lucky" button giving back a good answer to a every natural language query.