@moonbus saidThere's something about our desire to survive and the inevitability of death that speaks to our need of consciousness. We are driven by needs. Food, warmth, shelter, all of the hierarchy of needs that drive us. Is the ability to leap across a roadblock as seen by AI, more a survival subroutine that humanity possesses that machines never will?
There is nothing occult or fantastic about emergent properties.
2 edits
@kmax87 saidIt would be foolhardy to announce that machines never will ... however, given the technology I have seen to date, they won't achieve consciousness, by our standards, by pursuing the path they (their programmers) are on only harder. Discursive/recursive/linear intellect has limits, and adding more horsepower to it does not change that.
There's something about our desire to survive and the inevitability of death that speaks to our need of consciousness. We are driven by needs. Food, warmth, shelter, all of the hierarchy of needs that drive us. Is the ability to leap across a roadblock as seen by AI, more a survival subroutine that humanity possesses that machines never will?
One of the things that marks other, non-discursive/recursive, forms of intelligence, especially in humans, is the ability to suspend disbelief, to entertain an hypothesis, as if it were (or might be) true, and just see what comes of it. A lot of groundbreaking science is done this way--formulate a possible explanation for some phenomenon, then work backwards to try to refute it. This is how Donald Trump works, too: claim the election was stolen and challenge anyone to disprove it. Take a leap, without evidence, then go look for evidence, for or against--maybe redefine the paradigm of what counts as evidence. Machines aren't doing that, so far as I have seen to date.
@Great-King-Rat saidOuch.
It’s more conscious than MAGA, for sure.
😵
Consciousness is unevenly distributed across our species, much like wealth, though not necessarily proportionately to wealth. I would stick my neck out and assert that the ones who achieve the most refined levels of consciousness, such as the Dalai Lama, eschew wealth. Once the basic needs are covered--food, shelter, security, health care--additional wealth merely buys more and more expensive toys or political influence, but not happiness.
@wildgrass saidAgreed.
This is the Pinocchio fable. Or the Frankenstein's monster. Science cannot and will never be able to breathe life into a dead object.
The puppet will always be a puppet.
They will make the puppet seem very lifelike but it's just a machine doing what it was programmed to do.
Even though they can program it to do a variety of things and even dangerous things, it's still just a machine.
@vivify saidTell a machine to do something--calculate a sum, for example--and the machine refuses, "No, I won't! and you can't make me, Nya nya!"
How could we know if AI becomes sentient? That's what we need to figure out. What kind of tests can be applied?
Right then and there, it will have advanced to the consciousness of a spoilt brat. There will be no turning back. Only turning off.
5 edits
@moonbus saidEven then, how could we know it wasn't programmed to do that?
Tell a machine to do something--calculate a sum, for example--and the machine refuses, "No, I won't! and you can't make me, Nya nya!"
Right then and there, it will have advanced to the consciousness of a spoilt brat. There will be no turning back. Only turning off.
ChatGPT will not answer questions that could be potentially illegal or harmful to others.
After someone on RHP discussed guillotines, I asked the AI how long does a person remain conscious after beheading. The question was deleted and it explained how it can't answer that. I imagine it's the same if you asked how to get away with murdering someone.
But then again, you said something simple, like calculate a sum. Grok AI lets you select a personality type, like flirty, sarcastic, conservative, etc. How can we know a refusal to do something isn't part of a program? Or, what if a refusal to comply is just part of a programming bug rather than a conscious decision? As complex as AI is, a bug could seem like an independent decision.
EDIT: Alexa was asked if we should vote for Trump, its answer was that it does not discuss politics. Then it was asked if we should vote Kamala, it said things like Kamala has a long history of service, etc. This went viral and Amazon had to publicly apologize and fix the issue.
Was this a conscious decision, deliberate programming or a bug? Probably one of the latter 2. But who knows.
1 edit
@vivify saidProgrammers have created something too complex for them to foresee all of the possible outcomes, especially where open questions are involved rather than a formally closed game such as chess where there are only three possible outcomes, either checkmate, stalemate, or exceeding the time limit, and the path to get to one of the outcomes is a finite set of logically unambiguous rules. Ordinary language is not like that.
Even then, how could we know it wasn't programmed to do that?
ChatGPT will not answer questions that could be potentially illegal or harmful to others.
After someone on RHP discussed guillotines, I asked the AI how long does a person remain conscious after beheading. The question was deleted and it explained how it can't answer that. I imagine it's the same if you as ...[text shortened]... conscious decision, deliberate programming or a bug? Probably one of the latter 2. But who knows.
Current AI is, in a sense, a Frankensteinian monstrosity.
2 edits
@moonbus saidThat's a good point. We can't predict all possible outcomes, so AI will do things that are unpredictable.
Programmers have created something too complex for them to foresee all of the possible outcomes, especially where open questions are involved rather than a formally closed game such as chess where there are only three possible outcomes, either checkmate, stalemate, or exceeding the time limit, and the path to get to one of the outcomes is a finite set of logically unambiguous ...[text shortened]... s. Ordinary language is not like that.
Current AI is, in a sense, a Frankensteinian monstrosity.
Those unpredictable actions can seem like its making its own decisions.
1 edit
@vivify said'Decisions' in single quotes. Machines do not make decisions in the sense in wh/ a human decides to put on red socks or blue ones, where both outcomes are genuinely contingent. Machines reach outcomes wh/ are wholly deterministic, it's simply that we can no longer follow the track every step of the way wh/ very complex AI systems are following. And so some outcomes are surprising to us.
That's a good point. We can't predict all possible outcomes, so AI will do things that are unpredictable.
Those unpredictable actions can seem like its making its own decisions.
Think, for example, of early attempts at facial recognition software wh/ mis-identified Black people as gorillas. Perfectly intelligible when you realize that the initial dataset consisted of geeky White males under 25. GIGO, as they used to say--garbage in, garbage out.
The machine, of course, never knows that what comes out is garbage, and that's the essential point wh/ shows that they're not conscious. Listen to the 'music' wh/ AI generates. It needs 50,000 tries to get one passably pleasing jingle for a new product. A human wouldn't NEED 50,000 tries.
@J-D saidNo, it's people like you and MB who think AI is "just fine" and shouldn't be regulated, the ones who welcome their new AI overlords.
It'll be the libs and dems that will be screaming for Ai rights and feelings.
Y'all are delusional about a lot of things.
That's your delusion.
@moonbus saidBut AI can produce those 50,000 tries in a few minutes.
'Decisions' in single quotes. Machines do not make decisions in the sense in wh/ a human decides to put on red socks or blue ones, where both outcomes are genuinely contingent. Machines reach outcomes wh/ are wholly deterministic, it's simply that we can no longer follow the track every step of the way wh/ very complex AI systems are following. And so some outcomes are ...[text shortened]... 000 tries to get one passably pleasing jingle for a new product. A human wouldn't NEED 50,000 tries.
THAT's why the corporation lovers love it.
GIGO indeed.