Hacker News Reader

Comment by goatlover | Hacker News Reader

Do you think that ML algorithms are having experiences when they label data?

We label experiences with different words because we have different kinds of experiences. But it's the experiences and not the labels that matter, and the labeling of those experiences does not cause the experience. We say we have dreams because we do have dreams in the first place.

You're proposing that labelling data is the same thing as having an experience, but you haven't explained how data is experience, or why the labelling matters for that.

View on HN

ML algorithms have very beautiful dreams and halucinations https://www.google.com/search?q=neural+net+dreams ;)

More seriously though, if we agree that the brain is described by physics, then all it can do, a turing machine can do as well, so at the root all experiences have to be data and computation.

Perhaps, but keep in mind that physics is a human description of the universe, not the universe itself. So it may also not be data and computation, as those might just be useful abstractions we create to model reality. Same with math.

But that gets into metaphysics, which is notoriously tricky.

That's why i say "if we agree that the brain is described by physics".

If you argue that a brain cannot be described by computation and there is something supernatural like soul, that's a fine theory, and the only way to disprove it, is to create working general ai.

> If you argue that a brain cannot be described by computation and there is something supernatural like soul,

Why are those the only two options? There's lots of different philosophical positions, particularly when it comes to consciousness.

I am not really sure about the value of this philosophical positions, since the ones i have seen, simply suggest to ignore logic for a little bit, to obtain the result they feel is correct.

The suggestion by Penrose that consciousness is connected to quantum state, and so cannot be copied, or computed on any classical turing machine fitting in the universe, is an interesting way to circumvent strange predictions of functionality, but is not very plausible.

And Jaron Lanier from your other comment explicitly suggests to not use approaches like that, because it would weaken vitality argument, when proven wrong.

The pancreas is described by physics. Can a Turing machine make insulin?

Turing machine can make insulin that works on cells created by the Turing machine.

It's only "insulin" and "cells" for someone interpreting the program. The issue here is that if you had a simulated brain, what makes that simulation a brain for the computer such that it would be conscious?

Here's Jaron Lanier on the issue of consciousness and computation: http://www.jaronlanier.com/zombie.html

He's meteor shower intuition pump is similar to the argument that nations should be conscious if functionalism is true. The reason being that any complex physical system could potentially instantiate the functionality that produces a conscious state. It's also similar to Greg Egan's "Dust theory of Consciousness" in Permutation City.

If one is willing to bite the functional bullet on that, then you can have your simulated consciousness, and the universe may be populated with all sorts of weird conscious experiences outside of animals.

There's no scientific way to define "experience". How do you know whether a microphone experiences sound or not?

>Do you think that ML algorithms are having experiences when they label data?

Yes, but not in the way we have. An RNN keeps track of what is going on, for example.

To take in information through sense inputs (eg, eyes) is experience, but in an unconscious way. We get overloaded if the information we take in is not deeply processed before it becomes conscious. A CNN is similar in this manner, but also quite alien.

When we're born our brains have not formed around our senses yet. Our brain goes through a sort of training stage taking in raw sensation and processing it. The more we experience of certain kinds of sensations the more our brain physically changes and adapts. It's theorized if you made a new kind of sense input and could plug it into a newborn brain correctly, the brain would adapt to interfacing with it.

This strongly implies qualia is real, in that we all experience the world in drastically different ways. It may be hard to imagine but my present moment is probably very alien compared to yours, as is with different animals as well as simpler things like bacteria and ML.

Reply on news.ycombinator.com