Mostly this is a development frustration, but I’m becoming increasingly disillusioned by the idea that any one machine learning algorithm is really enough to produce even an AGI, let alone an ASI.
There are some things you just don’t need machine learning to do. I figured this out by the fact that certain processes are already covered by the existing Linux command line, so you really don’t need any extra ruby libraries for dealing with it.
For example, why use a decision tree to sense temperature, when the sensors command in linux works just fine? Same thing for lighting, or photometry. Just use red shift. I mean my goals is to make something that can directly adjust the lighting, not complain about the lighting.
So I’ve been spending this time trying to find a way to have both basic abilities that don’t have to be learned, and ones one can only be obtained by having the machine think.
It’s really not helpful then, when some discord channels teach only a machine learning approach, and worse yet, focus on one machine learning process or two.
I might reconsider my stance when I do androids, but there are things a computer just knows how to do, without being taught.
Even in a cockroach level intelligence, I need more than just a decision tree to do anything really meaningful.