Hacker News Reader

Comment by cryodesign | Hacker News Reader

Ignorance is bliss so they say. This is not about a computer beating humans in trivia questions, but about computers learning to understand humans. This is a major milestone in natural language processing and imagine what this can do when it will be available on a wider scale.

Watch this please:

Building Watson - A Brief Overview of the DeepQA Project http://www.youtube.com/watch?v=3G2H3DZ8rNc

View on HN

Yes, but that's the point, it doesn't actually "understand". When it answers a question in the category of "rhyme time" it's not because it actually understands what a rhyme is, but because it's been programmed to. If Jeopardy renamed the rhyme time category "time rhyme" (requiring the contestant to rhyme AND invert) watson would be up the creek, whereas someone like Ken Jennings could probably even figure it out without Alex Trebek explaining the category, and you or I would figure out the scheme after flubbing the first one.

So what exactly then is the definition of understanding, vs being programmed to act like one understands? Do I, for instance, understand pain when I remove the hand from the hot water, or is it because the reflexes in my brainstem are programmed to? Or do I understand mathematical concepts like the monads I use in Haskell when I probably will fail horribly to descripe them adequately or manipulate them in a proper mathematicial setting?

Appreciate what you're saying and I did get that Watson doesn't truly "understand" us yet, but still, what has been achieved so far is pretty impressive and would have useful applications even in its current form. Curious to see what IBM will do next with this.

Reply on news.ycombinator.com