[<< | Prev | Index | Next | >>]

Thursday, September 05, 2013

An Open Letter to Friendly AI Proponents

An unfortunate number of intelligent folks who ought to be working their asses off to understand and implement AI asap have sadly opted for the much easier task of arm-chair philosophizing about the dangers of AI and the merits and methods of making sure it has indelibly benevolent intentions towards us.

This disappoints me every time I'm reminded of it, which unfortunately is nine times out of ten that I meet someone with a supposed interest in AI.

While I can't fault the intention, actually spending more than half an hour at this particular navel gazing is insidiously self-destructive. The insidious part being that it's going to take down everyone else too. That may sound backwards considering the seemingly good intentions, but let's be realistic:

No law or even good idea is going to stop various militaries around the world, including our own, from working as fast as they can to create Skynet. Even if they tell you they've put the brakes on and are cautiously proceeding in perfect accordance with your carefully constructed rules of friendly AI, that's just their way of telling you you're stupid.

There are basically two outcomes possible here: They succeed in your lifetime, and you are killed by a Terminator, or they don't succeed in your lifetime and you die of old age.

I suggest choosing option three: Have one last party with your navel, then get off your sofa and grab a computer or a pad of paper and start working on solving AI as fast as you can. Contrary to singularity b.s., the AI you invent isn't going to rewrite the laws of physics and destroy the universe before you can hit control-C. Basic space, time, and energy limitations will likely confound your laptop's ambitions to take over the world for quite some time--plenty of time for those who best understand it to toy with what it really takes to make it friendly. That's assuming it's you and me, and not SAIC. And maybe, just maybe, if we work together and make enough progress in our lifetimes, that AI can help us live long enough to live even longer still...

But it starts now, and the first step is admitting that AI is hard and accepting that you have no fucking clue how to do it. If you can't do that, you'll never be able to leave that sofa comfort zone. Have an idea? Try it. Code it up. Nothing will teach you more about what you do (and mostly don't) know than that. Share your results, positive or negative. Look for more ideas. Don't be attached to anything--wear failures with pride. Today's good idea is tomorrow's nonsense, and two years later may prove the solution after all. Stir the pot and dive in. Make it happen.

I promise you that by default the next thirty years of your life will go by in a blink and you will look around you horrified at how little progress has happened--and you'll wish you'd been working on the other side of the equation.

(No coffee beans were harmed in the production of this entry.)

[2013-10-20 addendum: Just noticed Ben Goertzel penned a similar rant three years ago... Nothing changes...]

[<< | Prev | Index | Next | >>]

Simon Funk / simonfunk@gmail.com