OK, by now everyone's familiar with "Watson", the IBM Jeopardy machine with all the hype and drama associated with Artificial Intelligence and what it means for the future of humanity.  Is it HAL or Skynet?  Not by any definition.

So what's the point?  At present, as a proof of concept, "Watson" has demonstrated that it is possible to use algorithms to produce some understanding of natural human speech.  I don't place much stock in "Watson" winning at Jeopardy since that seems as much a function of how quickly the buzzer can be pushed as it is in having an answer.

A recent article and various sources have suggested that this is the first step in improving the future of medicine.  However, even this is still far from reality, since there is no process (at present) by which other diagnostic data can be readily processed (i.e. x-rays, MRI's, etc.).  This is still a significant hurdle when it comes to diagnosis, since any natural human speech (if you want to call a doctor's speech "natural") is only part of the process in diagnosing diseases.

There have also been articles raising the concerns about what such machines will mean to the future of humans being employed in various knowledge-based jobs.  While I continue to remain skeptical, I don't see any issue that "Watson" or such machines are just over the horizon. In truth, we've seen several decades pass where "expert systems" and "neural networks" were going to be the path to artificial intelligence and revolutionaize the role of "knowledge-workers". Then despite Deep Blue's win against Kasparov, it has been over a decade since any other such process has been introduced, although chess playing programs have clearly gotten significantly better.

While not trivializing the gains in processing natural language, the rest is simply a large search engine.  Perhaps some will be offended at my using the word "simply", but this system isn't anything except a proof of concept.  Consider that the failure of "Watson" to answer the airport question under the "U.S. Cities" category on the second night, resulted in an answer of Toronto.  While this may not seem overly significant, it is more of a big deal since it indicated that somewhere in the depths of the algorithms employed, a simple qualifier like "U.S." was completely misinterpreted resulting in an incorrect answer.

There is no doubt that many of these problems will be solved.  Algorithms will get better; technology will get cheaper.  Everything will ultimately improve on this, or a comparable, system and produce ever more sophisticated systems with increasingly sophisticated capabilities.

What is important to remember is that we don't even know what we don't know.  We use terms like "intelligence" and "understanding" as if we have a working definition that we can attempt to address with technology.  There have been numerous machines and technological innovations over the decades that always look interesting and promising, but in the end, we invariably end up with a hugely sophisticated machine that is capable of only doing one thing.  While it is a valuable exercise in engineering and computation, let's not be dusting off Asimov's three laws of robotics just yet.