🪷 on Warpcast

Sam Iglesias pfp

nah https://i.imgur.com/xgoO1Nd.jpg

4 replies

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

I'm down for defaulting to treating AI like it's conscious. Historically, we've made way more mistakes by defaulting to dehumanization.

2 replies

0 recast

0 reaction

Sam Iglesias pfp

Suffice it to say IMO it’s likelier that assuming consciousness in AI will lead to more mistakes (and more harm) than not.

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

ooh interesting. Can you give the tldr?

3 replies

0 recast

0 reaction

Sam Iglesias pfp

I can try. Consciousness names a biological process like digestion or photosynthesis. Anything seeking to produce consciousness would have to duplicate its causal mechanisms, which are poorly understood. An artificial heart needs to pump blood. A flying machine needs to operate with drag and lift.

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

So you're defining consciousness as biological? Couldn't you do the same thing for flight? "Flight is something that birds do so you can't have a machine do it." Ok that's fine, what do you want to call it when a machine does the thing which we call flight when it's a bird? What about the thing we call consciousness?

1 reply

0 recast

0 reaction

Sam Iglesias pfp

I didn’t say it is biological, I said it needs to duplicate the causal mechanisms. Artificial hearts and wings don’t use biological tissue, but they are implemented using an understanding of the physics of the phenomenon.

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

Ah ok. I think I would have understood this better if you said that it needed to duplicate its causal effects not its mechanisms. So, do you think we won't engineer consciousness by accident because we don't understand what it is?

2 replies

0 recast

0 reaction

Sam Iglesias pfp

You’d have to be committed, for example, to the position that a system of water pipes that implemented a program passing the Turing test using mechanical energy, *must* be conscious. I personally find that implausible.

2 replies

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

why in the world is that implausible?

1 reply

0 recast

0 reaction

Sam Iglesias pfp

Ok, what about a system of beer cans? A pack of well trained dogs running the program? The population of china? A program is just a set of formal instructions. Get it out of a box of silicon. Consciousness as a phenomenon would have to be the only thing in the universe with this special functional property.

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

Yes. I'm saying that consciousness is fundamentally computational, and so could come to have a sense of self whether it emerged as a pattern of earthquakes on a volcanic planet, a system of reflective crystals in a photosensitive gel, a scurry of neutrinos on a neutron star, or a pile of amino acids on earth.

1 reply

0 recast

0 reaction

Sam Iglesias pfp

What’s your justification for asserting that consciousness is fundamentally computational? If biological brains cause consciousness and cause minds, how do you get from there to stating that programs and Turing machines do the same thing?

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

The same way you can know I'm conscious without once asking me to distinguish whether I run on neurons or transistors.

1 reply

0 recast

0 reaction

Sam Iglesias pfp

I infer your consciousness on the basis of shared biology and behavior, roughly, as a background assumption. if you ended up being a bot or a fictional invention of multiple people I would simply take it that I was mistaken. That inference as far as I can tell has nothing to do with computation.

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

Then ultimately you've defined consciousness as biological. Which is fine. What name would you like to use for the thing that looks and acts like consciousness in non-biological systems?

1 reply

0 recast

0 reaction

Sam Iglesias pfp

I’ll stress again. You can in principle make an artificial brain out of something inorganic. It needs to duplicate the processes in the brain that cause consciousness. Biology points the way because brains cause minds. The depiction in Ex Machina of a blue glowing thing is apt.

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

Ah ok. But then which processes specifically does it need to replicate? What if it merely needs to develop priors over key and query vectors and Attention is truly All You Need? Are axons and dendrites prerequisites for consciousness? Do you need myelin sheaths for a sense of "my"? Or Golgi bodies to contemplate God?

1 reply

0 recast

0 reaction

Sam Iglesias pfp

We don’t fully understand the mechanism yet. The larger issue is what abstractly computation is relative to the physical world and physical processes. Molecules of water arrange in such a way that they are wet or frozen. A computer simulating water can be neither wet nor frozen. So to with molecules being conscious.

2 replies

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

Ultimately, I think it comes down to these questions, which I don't think are easy for either of us to answer: If we meet a silicon creature that seems to be conscious: 1. how can you change your mind and decide it is conscious? 2. how can I change my mind and decide it's not? I don't have a good answer for either one

1 reply

0 recast

0 reaction

Sam Iglesias pfp

I think ultimately a better understanding of the relationship between the brain and consciousness will settle these questions. I also think deep meditation on the nature of simulation vs duplication could clarify a few things too.

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

Simulation vs duplication seems like a syntactic distinction. If the simulation is faithfully duplicating the dynamics then the two are isomorphic, and that's sufficient to say they are, up to isomorphism, the same.

3 replies

0 recast

0 reaction

Sam Iglesias pfp

Is killing a sufficiently realistic AI enemy in a video game murder?

1 reply

0 recast

0 reaction

Connor McCormick ☀️ pfp

Connor McCormick ☀️

The ethics of AI are different because the causal model is different. Deleting an AI's weights yes might be murder. But no killing them in a video game probably isn't, just like killing your avatar in a game isn't murder.

1 reply

0 recast

0 reaction

🪷 pfp

0 reply

0 recast

0 reaction