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

Is artificial consciousness achievable? Lessons from the human brain

arxiv.org

Post reply on HN
Print view View original

31–40 of 604 posts

Re: Is artificial consciousness achievable? Lessons from the human brain

#31

Post by vixen99 »

ben_w wrote: post #12
hiAndrewQuinn wrote: post #3

We don't even know whether other human beings are conscious, man. The only thing which we might say with much certainty is that things which are more "like us" along some metric are more likely to be actually conscious, and things which are less "like us" are less likely to be so. Maybe everything is conscious. Maybe nothing except one's own self. But you'll never truly know one way or another, even if humanity inven…

I know that at least one other human is conscious, otherwise the term would never have been invented. But you have no way to tell if I am as conscious as I claim to be, or if I'm just a large language model trained by humanity :P

In thinking about this perennial problem it's worth bearing in mind that human beings pick up and process an immense amount of data on a continuous basis, that is currently unavailable to any LLM.

Re: Is artificial consciousness achievable? Lessons from the human brain

#32

Post by dist-epoch »

Dibby053 wrote: post #18

So far, the concept of consciousness is basically metaphysics. It doesn't have a role, it can't be measured... If I may suggest a starting point to get over this hurdle: let's create a "consciousness captcha": some task that is easy for conscious beings, but hard for algorithms. Consciousness evolved, therefore it must have provided an advantage. We just have to find it.

> It doesn't have a role ... Consciousness evolved, therefore it must have provided an advantage. We just have to find it.

That's a pretty wild claim, considering that the only animal proven to have a consciousness is the apex-predator of the whole world.

Re: Is artificial consciousness achievable? Lessons from the human brain

#34

Post by QuadmasterXLII »

Dibby053 wrote: post #18

So far, the concept of consciousness is basically metaphysics. It doesn't have a role, it can't be measured... If I may suggest a starting point to get over this hurdle: let's create a "consciousness captcha": some task that is easy for conscious beings, but hard for algorithms. Consciousness evolved, therefore it must have provided an advantage. We just have to find it.

This is a puzzle with a known solution, although we have obviously not found an algorithm that passes. Casually isolate the being/algorithm from known conscious beings, put it with a bunch of its friends instead, and see if it starts arguing with them about qualia. This will of course require creating an AI self-training algorithm strong enough to invent language independently (otherwise you get LLMs which are likely parotting human arguments about qualia), but it shouldn't require that much compute compared to training and running an AGI for other tasks- humans invent language when you put 30 children together for 30 years (see Nicaraguan Sign Language). This highlights that we have no idea how to train a transformer such that 30 copies of it left on a minecraft server and started from random weights would begin to meaningfully communicate . On the other hand, if 400B parameter versions of openAI's hide and seek AI start arguing with each other about whether ramps are really phlornge or if phlornge is just a property in their heads, we aught to believe that they have qualia.

Re: Is artificial consciousness achievable? Lessons from the human brain

#35

Post by pas »

hiAndrewQuinn wrote: post #3

We don't even know whether other human beings are conscious, man. The only thing which we might say with much certainty is that things which are more "like us" along some metric are more likely to be actually conscious, and things which are less "like us" are less likely to be so. Maybe everything is conscious. Maybe nothing except one's own self. But you'll never truly know one way or another, even if humanity inven…

Joscha Bach's model of "evolution of a shared consciousness, but compartmentalized into separate bodies" seems to be the one that makes the most sense currently

Re: Is artificial consciousness achievable? Lessons from the human brain

#36

Post by pas »

seydor wrote: post #13
xvector wrote: post #6

I really do hope ASI gets us to gradual replacement-based uploading. Having something as glorious as sapience trapped in a delicate and temporary bag of flesh kinda sucks.

I m still conflicted what we do with the bag of flesh after the fact. Are we bag holding forever?

what do you mean? the bag is useful for going around, but it's frail and needs replacing every few decades.

Re: Is artificial consciousness achievable? Lessons from the human brain

#37

Post by vixen99 »

xvector wrote: post #6

I really do hope ASI gets us to gradual replacement-based uploading. Having something as glorious as sapience trapped in a delicate and temporary bag of flesh kinda sucks.

Nothing particularly glorious about sapience trapped in a permanent bag of flesh especially one with a rather fixed view about the world about it. Never mind dictators; how about:

'The President of Global Enterprises Inc. is today 150 years old and currently holds the record for the longest serving president in the Company's history'.

A cause for celebration?

Re: Is artificial consciousness achievable? Lessons from the human brain

#39

Post by visarga »

grishka wrote: post #10

For all we know, it might be that consciousness is not fully contained in the physical structure of the brain. It might as well be something that partially exists on another layer of reality we have no idea about yet.

> For all we know, it might be that consciousness is not fully contained in the physical structure of the brain. It might as well be something that partially exists on another layer of reality we have no idea about yet.

Yes, but not another layer. Just the environment around us - physical and social. Every sensation comes from the environment, our perceptions are trained on this data stream, every value is dependent on environment, our emotions reflect it, language and society are part of the environment, we base our thoughts on language and our actions on other people. Our brain is made from environment signals, just like GPT-4 is made from its language corpus.

The unsung hero of consciousness is actually the environment with its reach data stream and feedback. Consciousness, language, genes, internet, LLMs and the evolution of intelligence are all social processes. They don't make sense individually, only as part of an evolutionary population. They can only evolve if there are many agents.

Now, I know this doesn't sound as sexy as quantum consciousness, but it is a more parsimonious and better grounded position. It accounts for the data engine that actually creates consciousness. Don't be looking for consciousness inside the brain or in exotic physics when the magical ingredient is outside.

Re: Is artificial consciousness achievable? Lessons from the human brain

#40

Post by bondarchuk »

Dibby053 wrote: post #18

So far, the concept of consciousness is basically metaphysics. It doesn't have a role, it can't be measured... If I may suggest a starting point to get over this hurdle: let's create a "consciousness captcha": some task that is easy for conscious beings, but hard for algorithms. Consciousness evolved, therefore it must have provided an advantage. We just have to find it.

You're talking about it right now in the physical world, though.
Post reply on HN