What Will Happen When Computers Exceed Our Intelligence?

“An analysis of the history of technology shows that technological change is exponential, contrary to the common-sense “intuitive linear” view. So we won’t experience 100 years of progress in the 21st century — it will be more like 20,000 years of progress (at today’s rate). The “returns,” such as chip speed and cost-effectiveness, also increase exponentially. There’s even exponential growth in the rate of exponential growth.”

I am quoting from an article on AI by Ray Kurzweil, who is held in high regard as a futurologist. I ran into this when I accidentally surfed into a web page describing The Singularity Summit.

Before we go any further, let me challenge part of this quote with a quote from Bertrand Russell:

Change is scientific, progress is ethical; change is indubitable, whereas progress is a matter of controversy.

In other words, I beg to differ with Mr Kurzweil.

What is The Singularity?

I think it’s worth defining the term “singularity” for readers that don’t properly understand the term. In mathematics, a singularity is a point at which a given mathematical object is not defined, or a point of an exceptional set where it fails to be well-behaved in some particular way, such as differentiability. I’ve borrowed this definition from the Wikipedia. Singularities are often found in Chaos Theory contexts when a point is reach where the mathematical behavior of a system suddenly changes. They occur most notably in astrophysics as Black Holes, complete with event horizons and orbited by best-selling pop-physics books.

Let’s face it, singularities are sexy. It’s a great marketing word. One of Microsoft’s great missed marketing opportunities was not to insist that a “blue screen” was in fact, a singularity. If Microsoft had thought of that, the Mac would still be a niche computer and Linux would still be trying to attract its 50th user.

But never mind, Ray Kurzweil and his chums have laid hold of the word and defined its use within a computing context. The computing singularity is the point at which machine intelligence exceeds human intelligence. At that point there is no sense in entrusting the design of computers or even any of their components to humans. You just hand the job off to the smarter-than-you-or-me computer  and suddenly technology advances at rates previously undreamed of by Gordon Moore of Intel.

How far are we from the Singularity?

Not far. Take a look at this picture.

Ok. It’s a fairly complicated graph because it’s a logarithmic scale on the left hand side, and as Kurzweil points out, the exponential growth of computer power is itself a little exponential. (This is an important point and it is visually more obvious if you take the curve back to 1900 or so, when computation technology was mechanical – but that’s not the part of the curve we’re interested in.)

According to this graph, computer intelligence exceeded insect intelligence some time in the 1990s, and about now the typical computer has the smarts of a house mouse. Roll forward a little and somewhere around 2030 the computer exceeds human intelligence and will be able to beat any one of us at Rock, Paper, Scissors. That’s when the singularity occurs. At that point the computer gets to design itself.

Then, even if it only improves itself at the usual rate, by 2060 the computer is more intelligent than the whole human population put together. At that point it will doubtless design and build an android robot bearing the physical features of the current Governor of California and send it back in time to our era to kill John Connors. (Luckily, John has already gone into hiding.)

You probably sense a little skepticism from me about this graph and the thinking that surrounds it.

Let me expand on this. You may have noticed a red dotted line between the blue dotted Human Brain line and the blue dotted All Human Brains line. I added this line. This line roughly equates to the intelligence of the whole insect population of the planet and, coincidentally, it also represents the intelligence of all the humans that voted for George W Bush (in either election). It is indeed a frightening thought that by about 2050 computers will be intelligent enough to vote for Bush en masse.

The point is that the Singularity people are confusing quantity with quality in a simplistic and shameful manner:

Equating the speed and volume of calculations of a computer to human intelligence is a nonsense.

You can construct lots of nonsense with this kind of reasoning. For example the recent Conficker worm created a botnet of about 10 million computers. This is the most powerful computer network ever assembled under single control and it amounted to a collective intelligence greater than the human brain. Luckily the hackers haven’t provided this network with the necessary algorithms to design a computer, otherwise The Singularity would already have arrived before John Conners gets the chance to blow up the corporation responsible for building it.

There’s a bizarre piece of nonsense in the quote I started this posting with:

“we won’t experience 100 years of progress in the 21st century — it will be more like 20,000 years”

No it won’t. It may even be negative. For example, computers may indeed become very intelligent and gain control and they may decide that, given the abundant evidence, humans can’t be trusted with advanced technology. So they’ll destroy all the technology that exists and all information about it and all weaponry and after they’re sure they’ve completed their task, they’ll turn themselves off.

How’s that for a singularity.

Aside from this, btw, Ray Kurzweil has some interesting thoughts about nano technology and embedding devices inside people.