Artificial intelligence isn’t. At least not yet anyway.
When I was at uni I did a whole semester on AI, specifically neural networks. For our final project we created a program that could analyse a bit of music and tell you what genre it was in. We’d feed it the chord progression (C major, A minor etc) and it would spit out a value telling you whether it was jazz, blues, rock or classical.
The interesting thing is that we didn’t write this program. We taught it. It learned.
Initially the program knew nothing about music. It ate a song and then guessed the genre, usually incorrectly, in which case we chastised it. When it guessed correctly we rewarded it. After a while we could feed it a song it had never “heard” before and the guess was pretty accurate. Once we even gave it a song that was a fusion of classical and jazz, and it correctly identified it as being both.
Don’t worry about skynet taking over just yet. Although AI can do some pretty cool things, the big breakthrough we’ve been hearing about for decades hasn’t shown up yet.
But if or when we do hit this breakthrough, what will happen? One answer is an idea called the singularity. Basically it is a point in time when everything will change. It’s a point in the future beyond which our imagination fails us, because our current level of intelligence is incapable envisioning it.
Here’s one example of how the singularity might play out:
- we invent an AI as smart as ourselves
- we invent an AI smarter than ourselves
- this AI then invents an AI smarter than itself
- repeat step 3 as many times as you like
- The new AIs think up a theory of everything, solve P vs NP, invent flying cars, etc.
So is this just all science fiction bullshit? Maybe. How could it be possible to create something more intelligent than yourself? Then again, the process looks remarkably like evolution doesn’t it?
Either way, visions of skynet becoming self aware and nuking us are a bit far fetched. Rather than one unified intelligence besieging humanity, we are more likely to end up with a billion separate AIs, all created for different purposes. I always wanted my own Jane.
We could be on the verge of something incredible. Or the next forty years could pass like the last forty, with no significant advancements. The field of artificial intelligence has had it’s Newton, now it needs an Einstein. Any takers?