Does Watson Know What It Wants to Learn Next?

didnt Watson already win jeapordy...? weeks ago?

@pmagg NOPE. it was just last night, as the article says.

"How fast does a sparrow fly?"

"African or European?"

It was _aired_ last night, but taped last month. And it was supposed to be a secret about the outcome until it aired.

@JohnnyH Yeah, i thought i remembered seeing an article about it on here around a month ago. Just another step towards artifical intelligence

The only thing more powerful than artificial intelligence is natural stupidity ...

Cool :) . I love the fact that IBM is already looking into speech recognition for Watson. Soon we can just ask Watson a question naturally and have it give us a relevant answer :P . No need to type; it listens.

You have no idea how hard it was for me to keep from calling Watson a "him" XD .

-IMP ;) :)

What happens when we put arms and legs, and visual receptors? What happens when it "sees" someone coming at it with an axe? Will it be able to visualize the contortion of the face to denote anger as seen here www.sciencedaily.com/releases/2007/07/070706141909 ?
Will it read through it's vast amounts of information to determine that the persons intent is to destroy it? Will it be able to realize that if the axe strikes it will most likely cause malfunction and scan through ways humans defend themselves from "malfunction" or "injury"? Will it then scan through the millions upon millions of lessons on how to fight and defend itself with those lessons? All before the person with the axe gets close enough to raise it and take a swing?

Skynet is right around the corner! :)

This article makes me want to read "I Robot" by Isaac Asimov again. Great book to read, if you get the chance.

Props to Ken Jennings for the beautiful Simpsons quote. (Kent Brockman: I for one welcome our new insect overlords)

Next: Watson vs. Manny Pacquiao

So they want the supercomputer to take Dr. House's job next? But who will program it's witty insults?

@billyblackhawk
yes, i have read "i, robot". no it is nowhere near that stage. it will only get that far if we give robots enough intelligence to make informed decisions and act on them. for example, watchdog circuits in a drone could be hardwired to guide it back to base if it loses communication. at that point, the circuit could also automatically blow fuses that would render the drone utterly incapable of firing a missile, simply because there is no longer a complete circuit that will do so.

Personally, i am very excited at the implications this has for the medical field. of course, watson's diagnoses would have to be double- checked by both him, and a doctor. given the mistakes he made on Jeopardy, i wouldn't trust him alone. on the other hand, he could be an indispensible addition to various types of research laboratories around the world.

i, for one, do not welcome our new computer overlords.
I, for one, welcome our new computer allies.

the people that script them for the show.

however, just having set insults for it wouldn't be enough. it would have to be able to understand how to piece them together independently, on it's own. i would love to see IBM use Watson to do much more advanced research into the A.L.I.C.E. chatbot. it could turn out some interesting results. especialy if watson could have day-to-day interactions from people to enable it to learn how to understand and speak in natural language. but what i truly want to see is Watson not only pass the Turing Test, but do so with flying colors, and do it a statistically significant number if times. (for the Turing Test @ Wikipedia, go to http://en.wikipedia.org/wiki/Turing_test)

I think that Jeopardy was a great format to display IBM's take on the basis of computer/human interaction. Something that stands out to me looking at the article is that the answer by Jennings should be "Who WAS Stoker" in reference to the long dead author. It seems to me that the distinction is one that Watson would have caught, yet one that has likely passed as a correct question, by type, on the show many times. Maybe Aristotle would chalk it up to "freeing men to do work worthy of men". Maybe we are truly only supposed to create the template for the analogue rather than to dictate every bit(byte) of it's form and function. I think that if I were a newly-awakened AI; one of my first questions might be "why is there so much resistance to fact?", and it's likely these kinds of issues that will become dissension, and the cause of strife between us and our constructs; that are doing exactly what we have told them to do.

If Watson should miss a Jeopardy question, does it miss it a second time? That is, upon hearing the correct answer, does Watson "learn" what the correct answer is? We could find out by asking the same question a second time. I've never seen the question asked a second time; I wonder what would happen. Anyway, that should be what Watson wants to learn next, the anwer to the very question it just missed. We learn, or should, from our mistakes. Does Watson?