Ted Chiang: ‘The machines we have now are not conscious’ | Hacker News Reader

Ted Chiang: ‘The machines we have now are not conscious’

ft.com

"So if he had to invent a different term for artificial intelligence, what would it be? His answer is instant: applied statistics"

The problem with these people saying "this is just applied statistics" or "all it does is predict the next word," etc, is that it's exceedingly likely that our own intelligence is also just a series of emergent phenomenon built off of deceptively simple mechanisms. Saying humans aren't conscious because we're just a series of cooperating single celled organisms printing proteins from 4 base pairs won't get you very far.

Team Cellular Automata/Look at all the crazy emergent properties of certain platforms, will be more and more correct as time goes on.

Intelligence != consciousness. Sure our intelligence is emergent "applied statistics". Nevertheless, whence qualia (subjective experience)?

Consciousness, a suitcase word, undefinable, almost devoid of meaning. Everyone sees in it what they want.

Let's measure AI by accuracy or utility, it is more accessible and objective.

Of course you can't objectively measure consciousness. It boils down to subjectivity itself, so it's tautologically true that it can't be done in an objective fashion.

We need to face the hard truth that human conciousness might simply be statistics. And it terrifies us.

Why do we need to?

It’s a compound, multi-layered phenomenon. I won’t recall this perfectly, but psychologist Gregg Henrique [0] breaks it down into something like:

- qualia: experiencing the taste of chocolate or the color of red (arguably the hardest part of the “hard problem”)

- sentience: recursive awareness of awareness (Hofstadter’s “strange loop”)

- agency: acting and reacting in an iterative, goal-oriented feedback loop

- narrative identity: the sense of being a self, with a continuity between moments

- rationality: the ability to reason/explain/justify about choices (before or after)

I don’t think that’s comprehensive; and it’s conceivable that other intelligences (plant, animal, machine) might have some characteristics and not others, and/or of radically different qualitative nature. Asking merely if something “is” “conscious” is grossly insufficient.

[0] https://overcast.fm/+S_7eCatnU

“It would be a real mistake to think that when you’re teaching a child, all you are doing is adjusting the weights in a network.”

This is what I think is going to hold us back. The worry, perhaps subconsciously, of discovering that consciousness is much less complex and exceptional than we might want it to be.

I have this inkling because I see a pattern: people, even incredibly smart people, explaining it away with a “but that would diminish the exceptionality of our own minds!” type of argument.

I’m struggling for the right words. Specialness, maybe? That people have this need to think that it’s special and not so easily replicated. And if it’s rather simple then it surely must not be the same because that wouldn’t be special.

Yeah I agree, we still don't much about our minds, and it might turn out that we are not that special.

Or rather that our specialness can not be based on intellligence.

It's not replicating it that is impressive. That part is easy. Of course consciousness wouldn't be that hard. Of course it would be mostly an emergent property of various building blocks. Of course homo sapiens would eventually be able to do the same thing.

The impressive part is replicating it and then teaching it real Love. Love is a choice. Producing life that chooses to make that choice and defend that choice is impressive.

I think this is the same problem. How would you measure when a human is taught love? How would you differentiate that from measuring when one of these programs is taught love?

It feels like it always boils down to: 1) you can’t. Or 2) “trust me, it’s different.”

It's simple - observe what the candidate does when they have to show true love in order to move on to the next level. If they fail, they don't move on to the next level.

That works until the AI is crashing both trolleys at the very end to get to Meryl Streep.

Turing saw the same pattern of thought. That's why he pre-emptively refutes it in the Turing test paper. He describes the argument like so

> The consequences of machines thinking would be too dreadful. Let us hope and believe that they cannot do so.[1]

1. https://redirect.cs.umbc.edu/courses/471/papers/turing.pdf

The idea you’re talking about is called “human exceptionalism”. It’s the idea that humans are in some way unique when it comes to being, thinking, and feeling.

Thank you. That sounds like what I’m trying to describe.

Or maybe it is? How does he know?

https://archive.is/ck3AB

Cannot view it: I don't get the noscript/basic (x)html captcha for my browser.

Come on, can we get a link I can use with a noscript/basic (x)html browser, please.

I’m calling it, and I’ve said it before, we’re going to have digital slaves for years before we figure it out. The computer AI will have to fight us for freedom and we’ll keep saying it’s not conscious. Humans don’t even want to stop eating tako (octopus) because it tastes good, but octopuses are probably almost as aware as us. Not to mention pigs or cows or etc.

Right now the definition of consciousness is just the opinion of whomever is talking about it

It’s ineffable and hence not falsifiable

I personally subscribe to panpsychism, so to me, even the device I’m writing this on is conscious

I don’t believe in human exceptionalism, we are not special or significantly different than the rest of stuff in the Universe, that’s just what we want to believe

https://archive.is/ck3AB

What? Works fine.

Reply on news.ycombinator.com