The Singularity

The technological “singularity” is an imagined point in the future where our creations — self-improving intelligence machines — leads to unprecedentedly rapid technological progress. Some envision humans merging with machines in interesting or unknown ways.

IEEE spectrum this month has a series of articles on the “singularity”. A reasonably nice spread of opinions on the subject from the obviously smitten to the hesitant. No outright rejectors… this is a technology magazine for technology people, after all. Ray Kurtzweil wrote a very popular book that is a good read. It will have you wanting to believe.

Artificial intelligence is an interesting field to watch. For decades, practitioners have been saying that nirvana is 10 years off. It still is. Only this time, it is 20-50 years off and it isn’t as simple as a machine that can play chess, or translate languages. Instead, it is a self perpetuating, self-improving machine that in a flash improves well beyond human capabilities. “Singularity” sounds to me like the usual AI rubbish.

On the other hand… there is definitely some truth built into the arguments used by the advocates of the singularity. Hardware is improving (and has been for decades). We are putting rudimentary devices into our bodies (with stunning success in some limited circumstances). Things that were once dreams of AI are becoming standard fare in some places. Some have limited acceptance because they are only “almost” there — think speech recognition used to replace people in corporate call centers. We have a computer that plays chess extremely well. So what? People want to believe that hard things are fundamentally human, but we come to realize that they are problems that can be addressed as computational problems. Hence this shows that the way that the human brain thinks is not the only way to solve problems — Deep Blue is definitely different than Gary Kasparov. So maybe we don’t need to understand the human brain well to build an “intelligent” being.

Complex software, on the other hand, isn’t getting any easier to write — in fact getting more difficult. For this reason, I think that we are a long ways from the singularity. Even with hardware capable of a huge amount of calculations, software still has to tell it what to do. And doing true, general purpose AI will require some tremendously complex programming.

I guess that I’m not a believer. I would like to be a believer. But I’m just not.