Responses to "So, IBM Has an AI (Artificial General Intelligence)" -- Part II

Below are some summarized comments on my blog post on IBM's new AI, and my responses...

In answer to one comment dismissing Watson as trivial as compared to software allowing data sharing between robots over the Internet (with the idea of creating a "cloud-based" AI as a result):
There are a number of impressive technologies out there -- the ones I cited above are only a few. But I wouldn't underestimate this step. It's not so much being able "to play a game" -- we had that with Deep Blue. It's being able to interpret and understand human language in such an open-ended way, both in comprehending questions and searching for answers, that make this such a game changer.
Rest assured, there are still major hurdles to overcome to achieving an AI with anything like the full range of basic human capabilities. But this is huge. So much of the progress we've seen so far has related to artificially circumscribed questions -- crunching numbers, inputting pre-formatted data, comparing fingerprints or facial structures, or narrowly interpreting language. Some breakthroughs have already gone beyond this, as I've noted above. This one is so potent because it implies an entity -- if only a huge and incredibly expensive entity -- capable of making complex, open-ended judgments.
I'm sure it has limits, but this is an incredibly impressive baseline to start with, rather than to merely be shooting for.

In response to someone who felt Watson was not really intelligent:

The trick is if you can answer Jeopardy questions with a high level of accuracy in the space of a second or so, you're probably at a level where you can "start following orders" -- even if those orders aren't inputted in highly precise, programming friendly language by a trained expert.
You may still hesitate to entrust an unsupervised machine with a nuclear reactor or loaded weapons, but the ability of this system to act on orders -- including the ability to automatically sift through over 200 million pages of accurate data looking for the answer that is both relevant and correct -- is a game changer.
Oh, it probably won't knock you out of a job in the next six months, but even what they have now is incredibly valuable for organizations struggling with information overload. And just about everybody with more than a 56K modem and an IBM 386 already is...

And further:

This machine may not laugh and dance and weep from the depth of its soul, but its level of problem solving -- at least in understanding and answering questions -- appears to be at least "human comparable." It won't be answering phones at a help desk near you -- or near Bangalore -- anytime soon, not given its size and cost. But there's a host of organizations that would already pay for what IBM has, if they can afford it.
Which means that future iterations will pay for themselves. Which means it will only get better. And smaller. And cheaper.
And continue... to change the game.