Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Artificial Intelligence Is Already Weirdly Inhuman

nautil.us

Post reply on HN
Print view View original

71–80 of 84 posts

Re: Artificial Intelligence Is Already Weirdly Inhuman

#71

Post by h0l0cube »

There is one important point that seems to be lost every time an article uses adversarial examples to justify why neural nets are deficient: not even a human is perfect at understand an image at a blink. When we see an image we decode a stream of impulses. Saccades will follow observing in detail number of areas of the image at different orientations, and so our interpretation of the image will come from numerous samples, not a single image.

IMHO, it's actually quite amazing that such primitive software neural networks can understand an image in a blink, in one 'sample'. Conversely, it's not inhuman to see pictures in clouds, Rorschach tests, or even static.

Re: Artificial Intelligence Is Already Weirdly Inhuman

#72

Post by chriswarbo »

ectoplasm wrote: post #63
teddyh wrote: post #61

Earlier quoted context omitted.

Even humans can’t explain their own reasoning. We can only rationalize our intuitive decisions after the fact, but it’s pure hokum – all decisions are made intuitively using no rational thought whatsoever, only afterward can we make up rationalizations for it. It follows that it should be perfectly possible for an A.I. to do this, too. (Edit: I now realize that this rationalization process was theorized by Douglas Ad…

Let's say you want as much money as you can get. This is probably not rational. Now let's say you have a choice of being given $5 or $10, which do you choose? Why was that not a rational choice? Because it's predicated on an irrational desire?

It's trivial to brute-force an answer to your question: if I take $5 I'll get $5, if I take $10 I'll get $10. I prefer $10 to $5, so I take the $10.

Intelligence is needed when there are too many possibilities to brute-force. We need to make guesses about which actions may give better results; we have to spot patterns, generalisations and simplifications to allow re-use of previous experience; we need to identify the crucial aspects, in order to narrow down the possibilities; we need to allocate resources efficiently, and know when to give up; and so on.

These are the things we would like machines to do well, but we have no satisfactory theory to explain, measure or compare such things. Introspecting our own thought patterns is not an effective way to reverse-engineer these proceses in humans, as the parent says.

Re: Artificial Intelligence Is Already Weirdly Inhuman

#73

Post by chriswarbo »

seiji wrote: post #25
chriswarbo wrote: post #23

Earlier quoted context omitted.

I'm a firm believer that it's a mistake to ask machines to do X in the hope that they'll do Y. We can't train an ANN to classify images then ask it to explain its reasoning; that's not the task we trained it for! How might we train an ANN to explain its reasoning? One approach would be to learn programs: have the ANN write programs which classify images. Then we have a classifier (run the program) and an explanation…

that's not the task we trained it for! Yes, exactly. We're not training consciousness networks. That's not even a goal of all this research. We're training dumb high-speed classifiers. How might we train an ANN to explain its reasoning? There are networks that exist today that'll happily explain an entire scene to you (the whole "this picture contains a pizza sitting on an oven in a kitchen and there is a dog in the…

> There are networks that exist today that'll happily explain an entire scene to you (the whole "this picture contains a pizza sitting on an oven in a kitchen and there is a dog in the corner")

I was thinking more along the lines of: "This picture contains pizza. This is because there is an ellipse which appears to be covered with cheese and pepperoni. I say that because the dominant colour is yellow and there are elliptical patches of a more reddish colour. ..." and so on, down to the pixel level.

We can kind of get this by running the networks backwards, but it's not really deductive reasoning; for example, we can't correct the network by saying "that yellow ellipse covered with red ellipses is actually a pile of vomit", and watch that knowledge propagate through the weights. Instead, we have to re-train with more examples of pizzas and vomit.

Re: Artificial Intelligence Is Already Weirdly Inhuman

#74

Post by noir_lord »

minikomi wrote: post #66
idlewords wrote: post #21

Earlier quoted context omitted.

This is particularly evident in endgames. Forced mates have been discovered that require hundreds of apparently random and pointless moves, until suddenly the king is cornered and mated. There's a nice discussion of some of these on Krabbé's chess blog: http://timkr.home.xs4all.nl/chess/perfect.htm

Wow.. do the games change when it's computer vs human as opposed to computer vs computer?

Once you are inside the horizon of an endgame tablebase (currently 7 pieces on the board or fewer) then the end is known, at that point the computer plays the game like a hypothetical god would.

Re: Artificial Intelligence Is Already Weirdly Inhuman

#75

Post by ilaksh »

We are trying to emulate some human capabilities. In doing so we have created a complex system which can be almost as hard to predict as a human. That's not surprising, that's hust physics.

Having a few odd classicifations is not surprising either. It hasn't been trained like a human and is being asked to select a class where 'randon squiggly' isn't an option.

To get human-like intelligence we need to develop them more as virtually embodied agents. Sort of like kids.

Re: Artificial Intelligence Is Already Weirdly Inhuman

#76

Post by noskynethere »

crazypyro wrote: post #33
noskynethere wrote: post #29

Earlier quoted context omitted.

> They try, usually with simple probabilistic techniques and input element-wise transforms, to mimic some function that produces approximations for a given set of inputs and outputs It's my understanding that this is basically how the brain works. My personal theory is that enough of these "dumb" inputs, wired correctly together, leads to emergent behavior that is consciousness.

I imagine the brain more like hundreds (thousands, millions, I'm not sure the magnitude) of different specialized neural networks. So you have a specific neural network for picking out colors and that feeds (along with a bunch of other inputs) into the neural network for picking out object boundaries and that feeds into the neural network for object recognition and so on. In comparison, most neural networks that are…

I'm outside my depth here, but isn't that what hierarchical learning is? (I think it's popularly called "deep learning", which I assume means the neural nets have depth?)

From what I've read, we aren't going more than a few dozens of levels deep. But it also sounds like this technique is very successful in image recognition.

Am I incorrect in my understanding?

Re: Artificial Intelligence Is Already Weirdly Inhuman

#77

Post by kordless »

eru wrote: post #68
kordless wrote: post #43

What if you showed the algorithm/network/whatever itself? You could 'take a picture of it' and then train it to know that's itself, which would then change what it looks like. Keep doing that until it doesn't change or you fail trying.

You could get into a loop. Though, what are you trying to accomplish here?

Consciousness?

Re: Artificial Intelligence Is Already Weirdly Inhuman

#78

Post by pokkanome »

ThrustVectoring wrote: post #57
pokkanome wrote: post #56

Earlier quoted context omitted.

A transistor is not doing computation. It is flipping a bit. Those bits are flipped in binary patterns with logic gates to do the computation. And a transistor is not a neural network node.

That just answered why I think that human brains are doing computation. The neurons fill the same role as transistors, and the patterns of neural connections fill the same role as how the transistors are wired together. I'm not saying that these are simple computations, or ones that are easy to understand, or ones that can be done in reasonable timeframes on silicon. For more useful discussion, I'd like to hear what…

I am using computation in the strictly mathematical sense. As in dealing with numbers. I do not think that our minds operate through a constant stream of numbers that become thoughts.

In that way, a computer and a human are fundamentally different. You cannot stimulate human thoughts as pure numbers. I think we need some extra layer of yet-to-be-invented abstraction to achieve that goal.

Of course, we could go the route of trying to create a new model of thought based around numbers, but that is proving to be difficult to understand. It would not be a good idea to try to build an intelligent system that we cannot completely understand because then all we could do is hope it works as we intended.

Re: Artificial Intelligence Is Already Weirdly Inhuman

#79

Post by ectoplasm »

teddyh wrote: post #69
ectoplasm wrote: post #63

Earlier quoted context omitted.

Let's say you want as much money as you can get. This is probably not rational. Now let's say you have a choice of being given $5 or $10, which do you choose? Why was that not a rational choice? Because it's predicated on an irrational desire?

No, that is not what I meant – the choice itself is not made rationally, even though a theoretical rational actor might have made the same choice. The most we humans can do it to make an intuitive choice, allow the post-hoc rationalization to occur and then check its rationality by analyzing its logic. If the logic is bad, we go back and try another intuitive choice.

By your own logic, you just made an intuitive decision about whether I have a point, and now you're just rationalizing it. Not much point in trying to have a rational conversation.

Re: Artificial Intelligence Is Already Weirdly Inhuman

#80

Post by ThrustVectoring »

pokkanome wrote: post #78
ThrustVectoring wrote: post #57

Earlier quoted context omitted.

That just answered why I think that human brains are doing computation. The neurons fill the same role as transistors, and the patterns of neural connections fill the same role as how the transistors are wired together. I'm not saying that these are simple computations, or ones that are easy to understand, or ones that can be done in reasonable timeframes on silicon. For more useful discussion, I'd like to hear what…

I am using computation in the strictly mathematical sense. As in dealing with numbers. I do not think that our minds operate through a constant stream of numbers that become thoughts. In that way, a computer and a human are fundamentally different. You cannot stimulate human thoughts as pure numbers. I think we need some extra layer of yet-to-be-invented abstraction to achieve that goal. Of course, we could go the ro…

The map is not the territory. "Computation" is fundamentally an abstraction for talking about that which various algorithms have in common. Algorithms themselves are a high-level description of a series of well-defined tasks. Computers aren't literally doing computation in the sense you are describing. What they're doing is simple physics with lots of voltage levels. The "computation" is a useful high-level description of what the computer is doing.

I agree that there's a missing abstraction for talking about human thought - it's a terribly complicated subject that isn't well understood. That doesn't mean that the human brain is doing anything that's different on a fundamental level than what computers can do. We don't have a high-level description of how human though works like we do with a computer, but it doesn't mean that human though has some kind of magic.

Post reply on HN