Over the previous few months, I’ve been on a path to grasp extra about AI and machine studying. Right now, as I used to be studying one thing about algorithms and numerous approaches to AI, I had an attention-grabbing thought/realization.
Not too long ago, after a health care provider’s appointment, I had been serious about how imprecise biology is as a science. A reliable physician usually replies with, “I don’t know”, whereas most medical doctors are primarily attempting to guess a particular end result that correlates with the signs they’re seeing. The impreciseness of all of it made me fairly uncomfortable. Biology didn’t appear that totally different from the social sciences, the place there aren’t any clear guidelines, solely frequent observations accompanied with tons of exceptions/assumptions. And it looks as if medication, particularly, is full of confounding variables.
I’ve at all times believed that, in case you really perceive one thing, you possibly can categorical it via a rigorous algorithm. In different phrases, clearly clarify it via a algorithm, inputs, outcomes, and constraints. That’s in all probability what drew me to check physics throughout my undergrad. In physics, we clearly outline the totally different guidelines that specify our world. Physics is gorgeous in its simplicity when you perceive it.
I additionally believed within the converse speculation — something that may be really understood could be expressed via a rigorous algorithm. What does this inform me about social sciences as areas of examine? I’m unsure. Maybe they’re simply inherently unexplainable. Maybe they don’t have any common guidelines, solely micro-explanations. All I do know is that this made me much less motivated to check them.
Again to my thought, or realization, from right now. It occurred to me that in physics, we’re in a position to match the information we see in regards to the world into guidelines, as a result of we’re in a position to compute the information and, extra importantly, perceive the computations. What if our brains are simply not in a position to compute the information obtainable in biology? We all know that machine studying algorithms are discovering new protein folding buildings en-masse, at a charge that will take hundreds of people hundreds of hours. Can they create algorithms to elucidate organic phenomena, algorithms which are as rigorous because the legal guidelines of physics? Can they do the identical for social sciences? Has this at all times simply been a giant information downside, one thing that our brains can not compute? However a deep neural community, with lots of of layers and tens of millions of nodes, is now in a position to compute it?
My realization facilities across the ‘rigorousness’ of varied fields of science. I’ve at all times cherished physics as a rigorous science that explains all the things we see round us. It occurred to me that many different fields could also be simply as rigorous, however our brains are simply not able to the extent of computation required.
My journey to grasp extra about AI has led to an exploration of the character of intelligence, how ‘pondering’ works, the similarities between people and machines in how they motive about issues, what we are able to be taught from each side, what outcomes one is able to that the opposite will not be, and what this tells us the longer term instructions wherein AI can develop. Understanding these is a tall order, and my exploration of them is only a private pastime given the opposite commitments I’ve — although a pastime that I’ve turn out to be extraordinarily captivated with. In my subsequent write-up(s), I hope to enter what I’ve realized about how intelligence works, each in people and in machines.