Does Watson Prove We’ve Reached The Singularity?


Jeopardy has never been so exciting. The US has watched rapt for the past few days as IBM super computer Watson has at first held its own and then gone on to defeat Jeopardy champions Brad Rutter ($10,400) and Ken Jennings ($4,800), raking in $35,734 by the end of  night two yesterday. There’s even been light speculation as to whether its much ballyhooed false “Toronto” answer is some sort computer joke.

Now, consider this in the context of this quote …

“We will soon create intelligences greater than our own. When this happens, human history will have reached a kind of singularity, an intellectual transition as impenetrable as the knotted space-time at the center of a black hole, and the world will pass far beyond our understanding.

We have been here before, the defeat of Gary Kasparov to IBM’s Deep Blue in 1997 was deemed to be such a moment, but of course it was no such thing. So is Watson now proof that we are at the singularity? The short simple answer is … No.

Lets think about this for a moment.

First, what does the term, “singularity” actually mean and where does it come from? (Side note, its also known as the rapture for nerds) It was Vernor Vinge (thats where the above quote comes from) who proposed that the creation of superhuman intelligence would represent a breakdown in the ability of humans to model the future thereafter. He was the first to use the term “singularity” for this notion, it comes from his 1983 article, and a later 1993 article entitled “The Coming Technological Singularity: How to Survive in the Post-Human Era”. Because it was widely disseminated on the World Wide Web, this greatly helped to quickly popularize the idea.

To claim that Watson doing well on Jeopardy means that the singularity has arrived would imply that Watson is a superhuman intelligence. So lets step back and consider that, by asking what Watson actually is.

Watson is an workload optimized system designed for complex analytics. It was developed by IBM, and has been designed to answer questions posed in natural language. In essence, its all about crunching a large volume of data very rapidly by parsing natural speech, then integrating the results into an appropriate reply by using efficient data manipulation …

  • Its not conscious
  • It does not self-replicate
  • Its not alive, it only creates the illusion of conscious intelligence, so while the lights many indeed be on and flashing away, nobody is home
  • It will not evolve, instead it will just do what it has been built to do, which is more of the same, but with more data

Oh come on now, think about it — You are a superhuman intelligence, so in order to dominate humanity your first move is to go on Jeopardy and try to win … seriously!. (That move is fantastic marketing for IBM and came from the mind of a senior IBM exec)

Its other words, its not super-human at all, its just a chunk of tin that has been designed to be a very efficient encylopedia with a natural language interface. The key question here is to ask if such a chunk of tin can ever replicate the biochemical complexity of an organic brain. Right now, the answer is no, we are a very long way from that, but still … watson is indeed an interesting step along that road and so perhaps in say 15 or 20 years we might indeed be a lot closer, but to even get there, we have to first gain a far better understanding of how the brain works and what consciousness actually is, There is still a lot of work to do to even manage that, so much in fact that it may indeed take far longer to replicate because we simply do not yet appreciate the true complexity of the biochemical engine sitting between our ears.

Anybody who claims a date for a singulariry is guessing. So while they might claim that by extrapulation we will hit the target by 2045 (sigh … yea I’m talking about the front page of Time), there are a couple of problems

  • How can you successfully extrapolate and suggest a date by which we will hit a target if we don’t actually know where the target is? (In other words, how can we simulate something we don’t yet understand)
  • Kurzweil hasn’t demonstrated that there is exponential growth at play here, he tends to cherry pick the stuff that supports his hypothesis, and ignores the stuff that does not (PZ makes a good case for that here)

Don’t get the wrong end of the stick here, we are indeed going in this direction. All I’m saying is that we are not there now, we will be closer in 15-20 years, but we will most probably be still some distance away by the time the 2045 calendars have been and gone.

Leave a Reply