I probably should have said "important" instead of "interesting".

How many of those companies are doing research that can be compared to what came out of Bell Labs in the 60's and 70's - the transistor, the laser, ... ?

IBM's work on advanced semiconductors and Watson certainly qualify. The later will probably be seen as one of the most important developments in the history of humanity and it wouldn't have been possible without the former.

Outside of Google, Palintar, and some of the other machine learning or niche consulting companies, I see little to no research on problems as hard and potentially influential as those that IBM or HP work on. The big reason you don't see what they do is that they don't work on consumer facing products very often.

The companies whose work I see at CS conferences and in patent searches for core algorithmic techniques are mostly IBM, HP, Microsoft, and Google. The majority of companies that are in the top links on HN on a given day do not do interesting research of the type IBM is doing.

Further, there is a lot of reinventing of the wheel going on in the startup scene these days. Just look at "new" things like in memory databases, column stores, event driven frameworks (Node.js) and others. All of these things are many decades old and do not count as new research even though they are newly written implementations. They may be more approachable to the common developer than the decades old implementations, and that has value, but this is not "research".

Check your link? That URL returns a server error.

Survives ? From what I hear it's already being applied to medicine.