I think it helps to distinguish automation into two categories. The first, we can call “Specific Automation”. This is when tasks that were once carried out by a person are now carried out by a machine. It’s output is much greater than that of a person, but it does so ‘unthinkingly’. If it can correct errors, it is only in ways that its creator has anticipated. If something goes wrong, it will spit out junk with the same speed and quantity that it would a quality product. If something goes catastrophically wrong, it will destroy itself and, depending on the machine, could pose a danger to any people near by. So you need someone to ‘mind’ it, to be the conscious actor in the system. This is how automation has always been and how it still is today. There’s a lot of money to be saved if you can cut out the conscious actor, if you can make the machine mind itself. Let’s call this “General Automation”.
I’m not gonna beat around the bush, “General Automation” is a pipe-dream. A computer is no different from any other machine; it can carry out specific tasks, reading and writing, with wondrous speeds. And it will crash the instant it’s in a situation it has no instructions for. Sure, you can make the computer write a bunch of instructions for itself, but you still have to predefine a metric by which it prunes those instructions - otherwise you’ve just got a hot mess of code that does nothing. After the pruning, you still have a hot mess of code – but! It yields results! Specific results, only within the parameters it’s been given. And you don’t know how it’s reached those results. And it’s not transferable. You have to go through the process all over again for any task you wish to point it at. Somehow, they say, this process will eventually lead to sentience – a machine that is truly aware of itself. Absurd! Why does anyone believe this? Because
A) It Would Really Save Us Quite a Lot of Money if You Could Deliver, Upfront Costs be Damned
and
B) Computers are Magic Books That Read and Write Themselves and Practically Nobody Knows How They Work, Even Programmers, Sometimes Especially Programmers
it’s a technology for laundering the violence of industrial society through pseudomathematical calculation. if you meet a computer on the road, strike it dead.
It is creating an edge for the surveillance state that makes our own organization more difficuly and will be used to disrupt us later. AI is a capital-intensive project and one where it is difficult for workers to shut it down or appropriate it or withhold their labor to prevent its function. We cannot achieve anything remotely approaching parity through taking over “the AI factory” not bend its use to our own gain in any significant way.
Until revolution, it’s a weapon of the enemy.
Is tool, is okay if used by a friendly hand, not so okay if used by an unfriendly one.
