While I generally adhere to the course curriculum and the spirit of learning, I couldn’t help myself on this one.
Instead of uploading clear photos and testing my model, I threw a little science fiction at the model and look what came out!
59 and 17 were considered human.
But 44 and 52 were horses.
I think the model needs to be trained on a few books from the Lion The Witch and the Wardrobe before it understands that these are centaurs. Would that work? lol
Next step is to put a couple of those photos of people wearing those horse masks through the model. I am willing to bet… they would be horses.
yeah nice imaginative mind. The neural net will learn what you feed to it and as you might know it learns patterns, it has no idea if the pattern is a horse or human. It also learns from what its fed its impropable to come up with decisions it has has not seen before in one or another way.
In this way, we can control the machines. Only teach them so much as is good for them and keep them oblivious to the other things that they shouldn’t know–like centaurs (just to make sure that a person horsing around doesn’t get misclassified ).
When the Google project learned to play Go at a championship level there was a lot of press about it ‘learning on its own’. But if you looked far enough under the covers you found a human-authored objective function that told ‘it’ what was good Go and what was not, which was then applied to thousands of game scenarios. Until the machines are capable of and motivated to author their own objective functions and decide for themselves what is Good and Evil, we’re ok.
Interesting story but it makes sense to me. Airtificial intelligence is not really intelligence it is learning things (lets say patterns) about what its fed, about past occurences. It seems intelligent but its just memory really. Real intelligence would not be predictable otherwise would become memory. Also there cannot be any line dividing the so called good and evil, what is good now maybe become bad tomorrow or even after an instance. In zen they say the source of intelligence comes from “no-mind” and this is significant, it basically means, its undefinable, unpredictable from the logic’s point of view.