I was wondering all day:
Can machine learning really teach a machine how to think?
I mean in a real unexpected situation that even a human may be confused , how a trained bot would react?

Machine learning means literally "learning from example". If the machine has never encountered anything similar before (during its training), it won't "know" what to do.

@MeanStreet guess that's true
But is it possible to teach a machine how to think?

We apply such an anthropomorphic view of “thinking” to things. Animals. Machines. Aliens.

It’s highly likely that the way we think is dictated by our biology and our lived experience.

I don’t think we’ll reach “thinking” in machines the way we see it, not unless we truly understand how we do it and replicate it.

It’s more likely that a machine would think in a very machine way.

We may even have achieved machine consciousness, especially if consciousness is an emergent property of distributed computation which the brain seems to suggest.

We’re only able to exert conscious will because we have limbs and the means to control them.

The internet could well be a conscious entity, thinking its machine thoughts, living its best machine life and we’d never know.

Depends on the definition of "think", but in a similar way we humans think, I say no (at least with our current technology).

Basically, it's just advanced statistical models created by generalising examples.

I feel like the name of neural networks in deep learning is misleading, as its structure isn't close at all from a brain, therefore I wouldn't call it "thinking"

I’d also say machine learning and it’s success lends more credence to the consciousness as emergent property hypothesis.

Look at how good it has gotten at insanely complex, seemingly very human tasks, all with a simple informational model.

It could well be that there’s a universal fundamental principle that information flow and processing is at the core of computation and therefore thinking.

It also gives some good reasons to be hopeful about the future of machine intelligence and alien intelligence too.

@Brolls brilliant !
never mentioned this point of view

@R1100 it’s amazing what a few tabs of LSD over the years can do when coupled with the brain of a developer and total sci-fi nerd :p

@MeanStreet you mean being able to create a human brain model is just reaching a perfect AI?

@Brolls :D

@R1100 I’d also say that to your original question the answer is “kind of”.

I think we’ll 100% get a machine to respond the way a human would.

We only have our senses to go on, and the only reason we buy that other people are humans too is because we’re pattern matching in our noggins on things like bodies etc.

This all comes back (as usual) to Plato and his cave thought experiment.

We don’t have to achieve machines thinking like humans to think that they do.

Just look at animals, most scientific literature agrees they’re not conscious. Some people even say that lower animals don’t have personalities. Yet if you have pets you swear differently.

We anthropomorphise so freaking much.

And for the tinfoil moment.

What’s to say we’re even thinking? We could just be chemical reactions playing out according to physics in a linear, fixed dance with no consciousnesses to speak of.

@R1100 No, if we create a perfect brain model, we would have a conscious machine, able to think in the same way we do. But it wouldn't be perfect because it wouldn't be any better than a human.

@MeanStreet so it would be even better than a human just because it would never go on sense and only on logic !

@Brolls in that case we are not even sure about ourselves 😶

@irene +1 for the intelligence != conscious.

I think it’s pretty evident that viruses and other such “life” like cells etc are pretty intelligent.

It may well transpire that when you get down to it, everything is just emergent behaviour from “dumb” components.

We’re probably no different.

Depends on how you define intelligence
Is it speed of your computation?
Or how much you can memorize?
Understanding?
Or even just survival?

Currently reading a related book : Homo Deus from Yuval Noah Harari. It talks about future of society and humanity, and therefore is full of interesting reasonings on AI. Might be interesting if you want to dig deeper into the subject.

I think a machine can completely learn to think like a human, and literally become like a human thinker, I don’t think we have a superior mind. We just have a really complex wiring of neurones that makes some cool stuff