George Rebane

IBM’s Watson supercomputer has bested humans at the TV game Jeopardy.  Furthermore, it promises to give a definitive performance on 13, 14, 15 February.  Big deal?  Yes it is; and please don’t confuse this advance in machine intelligence with the 1997 victory over chess master Gary Gasparov by IBM’s then hotshot Deep Blue.  The discovery of Jeopardy questions in response to a short list of ‘answers’ is a much more complex and significant task (more here).

Watson was not developed to give IBM a new source of revenues as a smart contestant on TV quiz shows.  It was developed to displace the next layer of smarter humans from the workplace.  RR readers are familiar with and not surprised by such advances in these pre-Singularity years (please see the posts in the ‘Singularity Signposts’ category on the right.)  Watson is now embodied in a room-sized computer of servers and large memories.   I believe we will be surprised at the rate that Watson’s progeny will shrink in size, power consumption, and cost.

The Watson AI combines code that incorporates speech recognition, speech understanding (including nuancing), and data mining algorithms that have access to a memory the equivalent of over 200 million pages (15TB RAM).  This puts the machine into industrial strength categories required for dredging out obscure patterns in data and inferring the possible relations of these patterns to the problem at hand.  And, of course, it can do that with arbitrary and arcane data sets that are equivalent in complexity to that of a set of data, information, and knowledge items that are common to American 21st century culture – i.e. the facile knowledge base of successful human Jeopardy contestants.

But, all this is already happening today.  Tomorrow and soon thereafter, Watson’s progeny will be able to access and process joined sets of data representing domains of knowledge that are related at various ‘semantic distances’ from each other.  In short, these critters will exhibit expertise in multiple areas way beyond the capability of any one human expert.  And they will be able to draw conclusions and infer new patterns (i.e. improvise and invent) at rates beyond our capabilities.  The point I want to emphasize here is that all this will be available shortly, and long before Singularity actually occurs.

Bottom line – displacing human cognitive capabilities in the work place is already well under way, and now Watson is demonstrating another quantum jump in functionality.  What the impact on the human job markets will be is a political problem that we in the developed world do not broadly acknowledge, and for which we have no preparation in the field of public policy.  Only one thing is certain, the collective response by our standard cohort of political mini-minds is guaranteed to amplify the realworld problems of wealth generation, its redistribution, and the resulting quality of life.

Exit question: I have published my estimate that by 2020 we will have about 70 million Americans (about 1/3 of the workforce then) functionally and structurally unemployed, what’s your estimate?  What should we be doing in preparation?