palou said:
I think you're underestimating programs quite a bit, here, and overestimate humans. Machine learning is fully capable discovering patterns/paths to solve problems in manners that the program isn't directly instructed to. Very, very general LSTM machine vision programs are capable of distinguishing a set of 5 newly introduced objects (let's say - cat, dog, mug chair and pencil-case) with 80% accuracy using only 3 specific data samples, after fine-tuning *themselves* to recognize quicker with entirely separate datasets. That's a much better result than any specific recognition program humans could attempt to write, the computer thus able to create efficient methods to work with largely by itself.
Humans are also very much limited to the input we receive (be it sound, audio, touch, etc...), our sensors being currently much more capable than what a machine has to work with. Everything we create stems from this input, one way or another. Yes, we are able to apply it creatively, but again, that's something machine-learning can achieve quite nicely, as well. |
Hmm... well, you have shown alot of knowledge and i only had one year of programming learning, and i don't keep up with the latest. But, even if a machine can recognise patterns if programmed to do so, it should not be able to come up with what to do with them, if not already pre-programmed aswell, from what i know. Basically, they have no initiative of their own. Whatever a program does, even wich parameters it uses to recognise the patterns and make a database have all been pre-programmed.
Of course, at the end of the day, we may discover that we are the same way and our IF's are wether chemical A is more prevelant than B.
My intention was not to say humans are better, but until we know what gives us conscience, we can't replicate it in a machine/program. If we do... yeah time to panic.
Last edited by Nem - on 11 February 2018
