In their current iteration, our machine learning things are mean-regression machines. If you want something average, these are your best bet.
Folks say “I wrote a screenplay and the A.I. made it way better. It’s not a mean-regression machine!” but it is. The screenplay was so far below average, regressing to the mean improved it.
If you ask a big giant trillion dollar mean-regression machine to just do all your accounting work, you’re going to get some pretty average results. That’s the whole point!
I’ve used A.I. to make python scripts. My python scripts are basically “put an excel sheet in, get different excel sheets out”. They all serve as “find the errors” short-cuts. When you have a big giant trillion dollar mean-regression machine, using it to make python scripts is pretty good. I didn’t ask which scripts I should make, I didn’t ask which errors I should sniff out. I didn’t ask which steps the python scripts should encode, I merely used the A.I. to serve as a python code interpreter, because I knew exactly what I wanted.
Soon enough the A.I. will just do my whole job. you’ll know you don’t need to hire a human when the auditor is also an A.I. won’t that be nice? An A.I. will talk to another A.I., and they’ll decide together whether or not to take money from you. At that point I’ll be useless to society, and so will you. Bear this in mind you ask me to hurry up my work by using A.I. more.