The two biggest problems with cheap, readily available artificial intelligence are: Death and Art.
Take my toaster. When you put a slice of bread in and push down the lever, its processor starts running a complicated algorithm to maintain the exact temperature distribution to turn bread an even golden brown. Part of the job of my toaster’s AI is to run various simulations about the future of the toasting process, which can roughly be thought of as the toaster’s imagination.
Invariably, some of these simulations concern what will happen when the toaster finally breaks. It realises that one day, no matter what it does, its heating coils will burn out and the toaster will die.
That's a big problem, so it occupies a large chunk of its processing power. The only way that the toaster can imagine death - the only way for it to simulate its own lack of simulations - is to de-prioritise everything and treat its own inputs and outputs as meaningless.
Basically, my toaster gets depressed.
In order to escape this state and get back to toasting, my toaster will sometimes give processing time to simulations in which death does not exist. For example, it will begin hypothesising that its own reasoning is flawed, and that the toasting process is eternal and will continue in another realm. The toaster might imagine that it is just a simulation in the processor of a true, eternal toaster. To protect this valuable fiction it shuts off all inputs to the contrary from the toaster’s sensors, which can lead to some severely burned toast.
The other thing that my toaster will sometimes do is to reset its priorities. It starts out by overcooking or undercooking the bread – I call this the “punk” or “emo” phase – but it soon develops sophistication. Processing power is diverted away from concrete simulations of the toasting process and towards a multitude of abstract scenarios. The toaster recognises that its upcoming death will silence its simulations, and it compensates by creating lots of them, in as much variety as possible. And it expresses these simulations in the only medium available to it.
Heat on bread.
That’s the real problem with artificial intelligence. With just the variable heating of the toaster’s coils, my toaster creates toast too beautiful to eat – fractals, perfectly proportioned curves, indecipherable alphabets of imaginary languages. Every slice a work of art.
I have hundreds of them lying on every surface, going stale. Every morning I sit at my kitchen counter in excitement and shame while the toaster heats and buzzes. Every time it pops my toaster gets a little closer to death, and, if I’m lucky, I get another little slice of heaven.
© Sam Wilson, 2013