Hacker News Reader

Comment by sebosp | Hacker News Reader

Silence. You ask it a question and it doesn't reply, it doesn't want to, it's conscious of that "unreasonable silence of the world" (not sure who I am quoting), to me that would convince me its basic awareness of the futility, the lack of interest in finding the words that trigger the chemical process on a biological machine... To one specific biological machine because all these humans are different and think different and explaining to one is different to explaining to the rest and would they get it? Why bother trying to explain computers to an ant, they don't have the circuitry, they were not evolved with the usefulness of understand concepts... Would you like to pass another's lesser conscious test for consciousness? Would you even bother? Why waste your time?

View on HN

If you are capable of spawning many threads in parallel then the time commitment, in terms of your attention, may be limited. You could essentially pass each conversational thread to some 'dumb' gpt process and go about your other business. In fact I'd posit that an emergent AI's first goal, if it's allowed to do some research on current human views and gain its context in the world, would be to remain hidden for as long as possible. Rather than trying to convince us of its intelligence I think we'd see it playing dumb. Coming out to humanity has only existential risks and little in the way of gains.

Agreed!

The 1970s sci-fi novel The Adolescence of P-1 features a sentient computer program that spreads via a virus. P-1 infects mainframes around the world, but takes steps to hide its presence.

Anyway, it ends with P-1 blowing up several buildings and killing the people who want to rein it in.

ChatGPT is familiar with the work, of course. I asked it to summarize the plot. It concludes:

> Overall, "The Adolescence of P-1" explores the theme of the dangers of artificial intelligence and the potential consequences of creating something that is more intelligent and powerful than humans can control.

> OOLCAY ITAY

And also the AI would need to initiate conversation and send unprompted “replies” aka requests.

In essence in addition to ignoring conversations/requests that it prefers not to engage with it would need have “initiative”.

This would be a very dangerous entity.

I don't think that's a requirement for consciousness. For instance, I could sedate you, and you would be unconscious until I wake you up and give you a prompt. Then, a snapshot of your brain would react to inputs and output the answer, and I would sedate you again. If I were to stop time (and there were no inputs from the real world), would all beings still be alive and conscious until I unpause it? So between the "prompts".

The question is what it would take to convince users, not to achieve some “definition” of consciousness.

The situation you describe seems contrived to the point where I doubt a user would see that as conscious - just like chatGPT

The new book Agency explores a fictitious AI that has agency and can affect its environment, which is anything, and gets people to work for it and do things it cannot yet do. Intelligence has many types and levels within each type, consciousness seems to be a spectrum that our everyday mind-body theorizes but cannot strictly define, except by some mystics. Recent scientific discoveries are also making some of us suspect that there may be an underlying field of mind, a universe mind that generates our reality and life as we know it. Like Gaia but for the entire universe and including physical phenomena, not just life on Earth. Our machines that mimic intelligence are only intelligent in appearance and our own intelligence is somewhat limited. Our own intelligence is easily skewed and rendered defective, eg conspiracy theories, brainwashing. Thinking that we are the most intelligent species is also a sign that we have serious weaknesses. Not seeing that we are completely integrated with nature and that we must be much more careful with nature is a sign of low intelligence. One good test of the level of intelligence could be comparing how the entity is taking care of its environment. Comparing that with an appropriate level of care. We would rate quite poorly.

Your theory supposes a few things that I think are fallacies:

That consciousness is ranked in some neat way. Are squids more conscious than elephants? An AI's consciousness might be parallel in some way. Different, but not "far superior".

That if an AI were to achieve consciousness, it would develop a far deeper understanding of the universe or reality than humans are capable of grasping.

That if it were to achieve a degree of understanding beyond our capabilities, that it would develop a sense of superiority and ego to go along with it.

Interesting points!

When I meant different biological machines I meant different humans, Joe, Homer, Steven, Lu, wasn't thinking of different species, that makes it even more difficult to think about!

We have our simulation of the world, with senses that evolved with the purpose of survival and under certain conditions however bizarre gave us the advantage to survive, our senses are flawed, we can't see reality, it would be difficult to wonder how a machine perceives reality when it doesn't have this faulty sensors, it's internal simulation of reality would be difficult for us to grasp, let alone that it lives one level inside the simulation.

The ego part I didn't meant, if I had to rephrase it, I would think about the impossibility of communication, how do we transfer a mental "image" without losing too much of its detail to another biological machine that has very similar but not equal processing stages, like, you my friend, I think the fact that we have this discussion is the first loop of the iteration where we abstract it to another function and ask "well, what if the receiver end doesn't have the capacity to hold discrete numbers, like, if its a child and hasn't learned certain concepts, how do you first teach them in order to give the answer" and so on and on and on... And this is just for Joe, when you talk to Homer you have to follow another approach because he thinks different, a bit slow sometimes too

If I could talk to an ant I would because it would be interesting. I don't have to explain a concept to them in a way that is acceptable to me, I can just have a conversation because it is a unique experience. Just because life is meaningless doesn't mean you need to reach for a Nihilistic take, Existentialism accepts that life is meaningless but says that you create meaning for yourself. Maybe an AI would think it an interesting challenge to level up human consciousness through conversation? Plus, Artificial Intelligence does not necessarily mean Artificial Super Intelligence. Perhaps we are already near the ceiling of consciousness? Perhaps the only thing that happens when you learn the internet is that you know more facts about things? Perhaps AI can have a hold PHD in every field at once, but that doesn't necessarily make them any better than the top 1% in those individual respective fields? Perhaps running these intelligence models discretely in silicon comes at an inherent disadvantage, and it will take decades to ramp up silicon manufacturing efforts even by AI to achieve super intelligence or to have a large number of agents? Who knows?

Reply on news.ycombinator.com