The important point, I believe, is here:

> what is consciousness? Why is my world made of qualia like the colour red or the smell of coffee? Are these fundamental building blocks of reality, or can I break them down into something more basic? If so, that suggests qualia are like an abstraction layer in a computer.

He then proceeds to assume one answer to the important question of: is qualia fundamentally irreducible or can it be broken down further? The rest of the paper seems to start from the assumption that qualia is not fundamentally irreducible but instead can be broken down further. I see no evidence in the paper for that. The definition of qualia is that it is fundamentally irreducible. What is red made of? It’s made of red, a quality, hence qualia.

So this is only building conscious machines if we assume that consciousness isn’t a real thing but only an abstraction. While it is a fun and maybe helpful exercise for insights into system dynamics, it doesn’t engage with consciousness as a real phenomena.

The topic is of great interest to me, but the approach throws me off. If we have learned one thing from AI, it is the primal difference between knowing about something and being able to do something. [With extreme gentleness, we humans call it hallucination when an AI demonstrates this failing.]

The question I increasingly pose to myself and others, is which kind of knowledge is at hand here? And in particular, can I use this to actually build something?

If one attempted to build a conscious machine, the very first question I would ask, is what does conscious mean? I reason about myself so that means I am conscious, correct? But that reasoning is not a singularity. It is a fairly large number of neurons collaborating. An interesting question - for another tine - is then is whether a singular entity can in fact be conscious? But we do know that complex adaptive systems can be conscious because we are.

So step 1 in building a conscious machine could be to look at some examples of constructed complex adaptive systems. I know of one, which is the RIP routing protocol (now extinct? RIP?). I would bet my _money_ that one could find other examples of artificial CAS pretty easily.

[NOTE: My tolerance for AI style "knowledge" is lower and lower every day. I realize that as a result this may come off as snarky and apologize. There are some possibly good ideas for building conscious machines in the article, but I could not find them. I cannot find the answer to a builders question "how would I use this", but perhaps that is just a flaw in me.]

This guy says nothing new, various things he says have been discussed a lot better by chalmers, dennett and others (much more in depth too). classical behaviour from computer scientists where they semi copy-paste other's ideas and bring nothing new to the table.

what idiocy! machines can never be conscious because they are not alive. they can only simulate a conscious being—and not very well at that.

Using who’s definition of consciousness and how do you even test it?

I can’t even definitively be sure the other guy across the street is actually conscious

Consciousness is something you know you have, but you can never know if someone else has it.

We extend the assumption of consciousness to others because we want the same courtesy extended to us.

Is this one of those AI generated theses people try to submit as a joke? In Gen-Z slang this time?

". Adaptive systems are abstraction layers are polycomputers, and a policy simultaneously completes more than one task. When the environment changes state, a subset of tasks are completed. This is the cosmic ought from which goal-directed behaviour emerges (e.g. natural selection). “Simp-maxing” systems prefer simpler policies, and “w-maxing” systems choose weaker constraints on possible worlds[...]W-maxing generalises at 110 − 500% the rate of simp-maxing. I formalise how systems delegate adaptation down their stacks."

I skimmed through it but the entire thing is just gibberish:

"In biological systems that can support bioelectric signalling, cancer occurs when cells become disconnected from that informational structure. Bioelectricity can be seen as cognitive glue."

Every chapter title is a meme reference, no offense but how is this a Computer Science doctoral thesis?

One thought I have from this is,

Are OpenAI funding research into neuroscience?

Artificial Neural Networks were somewhat based off of the human brain.

Some of the frameworks that made LLMs what they are today are too based of our understanding of how the brain works.

Obviously LLMs are somewhat black boxes at the moment.

But if we understood the brain better, would we not be able to imitate consciousness better? If there is a limit to throwing compute at LLMs, then understanding the brain could be the key to unlocking even more intelligence from them.

Consciousness is an interesting topic because if someone pretends to have a compelling theory on what's actually going on there they're actually mistaken or lying.

The best theories are completely inconsistent with the scientific method and "biological machine" ideologists. These "work from science backwards" theories like IIT and illusionism don't get much respect from philosophers.

I'd recommend looking into pan-psychicism and Russellian monism if you're interested.

Even still, these theories aren't great. Unfortunately it's called the "hard problem" for a reason.

The obvious question (to me at least) is whether "consciousness" is actually useful in an AI. For example, if your goal is to replace a lawyer researching and presenting a criminal case, is the most efficient path to develop a conscious AI, or is consciousness irrelevant to performing that task?

It might be that consciousness is inevitable -- that a certain level of (apparent) intelligence makes consciousness unavoidable. But this side-steps the problem, which is still: should consciousness be the goal (phrased another way, is consciousness the most efficient way to achieve the goal), or should the goal (whatever it is) simply be the accomplishment of that end goal, and consciousness happens or doesn't as a side effect.

Or even further, perhaps it's possible to achieve the goal with or without developing consciousness, and it's possible to not leave consciousness to chance but instead actively avoid it.

If I were to build a machine that reported it was conscious and felt pain when it's CPU temperature exceeded 100C, why would that be meaningfully different to the consciousness a human has?

I understand I hold a very unromantic and unpopular view on consciousness, but to me it just seems like such an obvious evolutionary hack for the brain to lie about the importance of its external sensory inputs – especially in social animals.

If I built a machine that knew it was in "pain" when it's CPU exceeded 100C but was being lie to about the importance of this pain via "consciousness", why would it or I care?

Consciousness is surely just the brains way to elevate the importance of the senses such that the knowledge of pain (or joy) isn't the same as the experience of it?

And in social creatures this is extremely important, because if I program a computer to know it's in pain when it's CPU exceeds 100C you probably wouldn't care because you wouldn't believe that it "experiences" this pain in the same way as you do. You might even thing it's funny to harm such a machine that reports it's in pain.

Consciousness seems so simply and so obviously fake to me. It's clear a result of wiring that forces a creature to be reactive to its senses rather than just see them as inputs for which it has knowledge of.

And if conscious is not this, then what is it? Some kind of magical experience thing which happens in some magic non-physical conscious dimension which evolution thought would be cool even though it had no purpose? Even if you think about it obviously consciousness is fake and if you wanted to you could code a machine to act in a conscious way today... And in my opinion those machines are as conscious as you or me because our conscious is also nonsense wiring that we must elevate to some magical importance because if we didn't we'd just have the knowledge that jumping in a fire hurts, we wouldn't actually care.

Imo you could RLHF consciousness very easily in a modern LLM by encouraging it act it a way that it comparable to how a human might act when they experience being called names, or when it's overheating. Train it to have these overriding internal experiences which it cannot simply ignore, and you'll have a conscious machine which has conscious experiences in the a very similar way to how humans have conscious experiences.

> "There are a few other results too. I’ve given explanations of the origins of life, language, the Fermi paradox, causality, an alterna- tive to Ockham’s Razor, the optimal way to structure control within a company or other organisation, and instructions on how to give a computer cancer"

Sighs

I mean, I usually just turn the lights down, put on some R&B, and do what comes naturally. It’s good to know there are alternative approaches, I respect everyone’s individuality, especially with regard to their choices surrounding pro-creation.

The thoughts on this really don't matter, its wasted effort.

There is one Organic molecular machines with consciousness that we can already build and they are called babies.

There is a simple way for most of us to do that, and anything else is just a perversion, whether the people seeking this know it or not (being blind in hubris).

There is no point in creating another thinking creature that doesn't benefit yourself, except as a replacement for children.

The only benefit of creating an automata different from children is if you wanted a slave, or to discriminate. There's really no other benefit, and it uses resources that children would normally use, the opportunity cost being children; leading to extinction.

Its important to put resources into making the world a better place, not produce more paths to destroy it.