I had the opportunity to visit the Watson Research labs a few weeks ago and meet with several researchers working on dynamic languages there. IBM is not like any Silicon Valley company, researchers have individual offices, there isn't tremendous shared space, and you pay for stuff in the cafeteria. But they have brilliant scientists and engineers there. It sounds like Ferrucci's attempts collaboration were much closer to a Bay Area startup than the IBM I saw.
Well my first Job at a high end RnD place it was all small 2 or 3 person offices.
Some times you need a quiet place to think - though we did have some interesting brainstorming sessions in my office - one on the best way to measure the efficiency of flushing toilets was one I remember.
I remember looking into using machine vision to help with this but when the base cost of the kit was £250,000 we had to give that up (this was in teh early 80's)
I think this Jeopardy challenge was so interesting in terms of general AI that it'd be a shame if the original team completely shifted their focus onto new fields. There are a lot of goalposts left, such as beating humans without the advantage of the instant reaction time or beating humans assisted by the Internet.
