Washington University researchers collaborating with a private software company (Allen Institute for Artificial Intelligence) reported results from a system called GeoS that uses a combination of machine vision, natural language processing, and a geometric solver to solve SAT geometry problems. The system achieved an average math SAT score of 500. Results were reported at the 2015 Conference on Empirical Methods in Natural Language Processing a couple of weeks ago.
Not even a single reference to IBM or Watson even though IBM claims to be the world leader in natural language processing. Unlike IBM, they didn't spend $billions to develop new hardware and they didn't work on the problem for 15 years.
But more importantly, they aren't interested in meaningless stunts like winning Jeopardy on television.
If IBM was serious about AI, they would stop playing publicity games and participate in the real world of academic research and contribute something. Then maybe they'd gain a real reputation and real customers would pay money to use their products.
But instead they focus their attention on publicity stunts and PR and fabricate fake customers by giving away systems as "First of a Kind" prototypes or "co-funded" development projects.
And when the stock price falls they shift the compensation from options (which end up out-of-the-money) to restricted stock that always has value. Then they raise the amount of restricted stock they receive to target a specific value in a specific time frame. And if that costs too much, they just fire more employees to cover the cost.