comments 6

The silliness of smarter-than-you machines

SkyNETAhh, the silliness of people.

John Kelly, Head of IBM research, said according to the Economist: “We are at a moment where computers and computer technology now have approached humans“. The moment of singularity is near, and the next president/prime minister will be SkyNET. Blah blah blah . . .

Apart from making great headlines, the debate is just plain silliness.

Think about it for a moment.

Computers have surpassed human abilities for decades – initially in calculus, then in memory (as in remembering stuff), and later in chess and now Jeopardy. But the fundamental principle of computers haven’t changed – they are precision calculators working increasingly fast. Nothing more, nothing less – so to compare it with our brain is just silly. For example, research has shown how the relative small brain of a bumble bee is able to solve the ‘travelling salesman problem‘ – i.e., find the shortest route for a given set of destinations – an impossible task for a computer.

The experts make claims based on contests with humans where the computer is specially designed to perform that task, and that task only (Watson can’t even play chess to same degree as DeepBlue). Secondly, the selected contests only measured the performance of the conscious mind – ignoring the remaining brain activities supporting our senses, keeping our body working (organs, immune system etc), feeling emotions and laughing at jokes during the contest. Our intelligence is the sum of our brain and nervous system – and extraordinary complex system consuming roughly 25% of the body’s (sugar) energy and 20% of the oxygen. Our brain ‘grows’ and constantly adapts in response to the tasks we are trying to do or master – i.e., our brain is an expert in environmental adaptation, not calculus.

The computer is doing none of that.

A computer is an expert in precision calculations, and as a result, computers play chess or Jeopardy in a completely different way to the way a human plays. A contest between the two is similar to a race between a standard, small, propeller airplane (brain) and a Bugatti Veyron (computer) – as done by the TopGear guys. And yes, the car did win, but only because of the German autobahn – another race between, say, Nice and Bethlehem would probably have worked out in favour of the plane, because the car is a specialist at going really fast on straight, well maintained, well protected roads.

We invented computers, because our brains are not good at performing large amounts of precision calculations. Computers are here to make our lives better, easier, and simpler (although not always successful) – they are scary the same way insects are scary when supersized in old Hollywood movies.

For people to design (or create in a God like fashion) another self-aware intelligence would require a deep understanding of what makes us ‘conscience’ (which I am yet to see any evidence off) while simultaneously performing all the tasks that we humans perform without conscience thought (see, hear, feel, emotion, being social, digestion, heal, adapt etc). If we ever do reach singularity then, I suspect that, we will be designing that computer using a technology dramatically different to our current computer/software technology….

Further reading: Artificial intelligence: Smarter-than-you machines | The Economist.

Filed under: complex systems, technology

About the Author

Posted by

Dad, husband, software architect, blogger, and more... I am a software architect by trade (but not one with a coding focus).. I try to apply the "keep it simple" principle to all aspects of my life. But as I have found, it is KIS and not KISS, because there is nothing simple or stupid about keeping it simple - the second law of thermodynamics really does apply to you as an 'isolated system' :)

6 Comments

  1. Paul Rusk

    Agreed.

    Last I heard the best inference engines aren’t as capable as a typical 12th grader.

    Since humans decide in the right brain, indeed I think an entirely new computer design would be needed.

  2. MT

    Re: Travelling Salesman Problem

    The Concorde TSP solver has been used to find provably optimal tours for TSP problems as large as 85900 points, running on relative clusters of commodity (1990’s) hardware. I think you’re slightly overestimating the abilities of bees. Instances of the problem that are solvable by bees are generally solvable by brute force algorithms in exponentially less time than the bees take. What is interesting in the bee article is that the bees are able to do it at all.

    But yes, the whole singularity idea is pretty silly – it’s like saying that when motor cars reached the point of being able to run at 80 km per hour, cheetahs became obsolete. It’s a handy shibboleth in AI debates for isolating the people who have literally no idea what they are talking about.

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s