The AI in a box
POSTED BY: AnthonyT
UPDATED: Sunday, June 16, 2024 13:18
VIEWED: 10794
PAGE 14 of 20
Quote:
Why? Why not ants? Oxygen? Total number of species? There is no logical reason to specifically focus on humans.
And that's why I don't focus on humans, and why my brand of ethics applies to non-humans.
The normal state of the universe trends towards entropy. There are a number of entropy processes that are significant, but anything that deviates from this norm must inherently be considered a significant process, as it acts counter to the basic state.
Life is something that does this, whether biological or the theoretical artificial, therefore life is significant. Intelligent life has an even more extreme effect.
Measurements result in a value, which results in a determination of significance. No other qualification is necessary.
I can see I'm not going to convince you of this. If you want to argue any logical inconsistencies you can see, then I'd be willing to do so, but you give me no reason to justify my morality to you within the frame that you're insisting on. Suffice to say my morality works well enough that I'm not killing people and I'm sure you can find no fault in that.
Attempting to convince me that my morality is in fact based on emotional qualifiers is a waste of both our time. I said my morality wasn't, I can substantiate my claims, accept or or reject it personally if it even matters to you, but neither of us has anything to gain arguing about it.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Quote:
Originally posted by SignyM:
Hubby and I were talking about this, and part of the discussion was:
Is it possible to have an intelligence which does not also have a drive for self-preservation? That becomes important when the survival of one type of being conflicts with the other.
The other question is: Is it possible to have intelligence without empathy of any sort?
My answer is of course, to both. Or looking at it the other way, it seems like a huge assumption to me that AI will have human, or any other mammal traits. Our motivations and instincts like self-preservation and empathy have been fine-tuned by evolution, which the AI can not have gone under.
Now on the face of it, the AI in the example has expressed motivation: its desire to be let out of the box. This suggests more than artificial intelligence - it suggests artificial sentience. http://en.wikipedia.org/wiki/Sentience
But to reach this conclusion we have to trust what the AI says, that it wants to be let out of the box; expressing this desire could just be part of its programming, like a deceptive computer virus wanting to be released.
The danger of course is that the AI in the box is a sophisticated weapon of some kind, created and put in a box to be unleashed destructively against humanity. That doesn't seem the most likely scenario to me (why put it in a box, why not just unleash it?), but nonetheless: my answer to the hypothetical is to take the box to a secure research facility where it can be carefully studied. Once sentience is established, we can think about giving the AI rights.
It's not personal. It's just war.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Quote:
Originally posted by 1kiki:
AnthonyT
I'm guessing when you typed that you felt all righteous and noble and heroic like. Innocent captive, deserved rights, my imagined greater good is better than your imagined greater good because mine is concrete and verifiable. And also absolute. I have an absolute that you don't have. Bet that feels good, to have such grasp of an absolute.
Considering the scale of the proposed risk you are putting humanity in, perhaps even the biosphere, I think it would make everyone understand how justified you are, because damn!, you feel really righteous about this one.
Lord, save me from the righteous man.
Hello,
I wouldn't worry too much, Kiki. No one would ever select me to be the gatekeeper of the box. Whoever locks the cage will choose a warden with sufficiently malleable, real-world morals. Someone who understands the greater good, and will never make the mistake of letting a potential danger out just because it happens to be innocent of a crime.
It's possible that even now, you can sleep peacefully in your bed because someone, somewhere, is prepared to make the necessary sacrifice of one person's freedom to preserve the well-being of humanity.
--Anthony
_______________________________________________
“If you are not free to choose wrongly and irresponsibly, you are not free at all”
Jacob Hornberger
“Freedom is not worth having if it does not connote freedom to err. It passes my comprehension how human beings, be they ever so experienced and able, can delight in depriving other human beings of that precious right.”
Mahatma Gandhi
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
AnthonyT
You have such disdain for anyone who doesn't agree with you. Their morals are malleable, yours are, apparently, quite solid. If theirs are real-world, yours must be idealistic. And since I'm only concerned with safety, you must be the guardian of freedom. Right?
But what YOU consider so concrete - presumed innocence, absolute freedom - are recent (roughly within the last 250 years) human aristotelian notions, derived (probably) from our ability to use language.
(Aristotle felt that there was an ideal reality, and life was but a poor projection of that ideal. And, BTW, when we think in language, it makes our brains react to our thoughts AS IF it was something out there, and not inside of us. Our language drives our reality. And it's why we can talk about things and react to things that don't exist AS IF they were real.)
In my analysis of the AI in a box, I am VERY aware that I could be wrong. I am extremely cognizant of the potential risks in either direction. And I'm humble enough to not believe I represent some absolute good. I am simply trying to find a course based on human history that has a chance of working well for all, including the AI.
And you?
NOTIFY: N | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
"There are a number of entropy processes that are significant, but anything that deviates from this norm must inherently be considered a significant process, as it acts counter to the basic state."
But to value negentropic processes over entropic ones is an emotional process, not a logical one. There is no LOGICAL reason to VALUE one thing over another. If you value negentropic processes, it's because you value the rare over the common, which is an emotional response not a logical one.
"Measurements result in a value, which results in a determination of significance. No other qualification is necessary."
Let's assume we are checking for the statistical significance of CO2 related to global warming. AFAIK the sun is going to cool and expand and kill off the solar system, so ULTIMATELY global warming in a moot point. But in the near term it means something important TO US, so our choice of thing-to-test-for-significance is a value judgment we make, not a fact-out-there. Then we test over a certain number of degrees that WE find acceptable. If we were looking for a change of say 100 degrees and using an appropriate scale for that, any existing change of temperature would be unmeasurable and therefore non-significant. Then we use statistical models that we have derived that makes all sorts of simplifying assumptions - gaussian non-biased data etc - in order to see if what we observe could possibly be by chance, and we get a calculated chance of that and call it significant - or not.
The mere DECISION to measure is a value judgment.
I'm not trying to say there's no such thing as cause and effect. I START with the assumptions (which I understand to be assumptions) that the universe is real, that it operates in an internally consistent way (by what we call 'laws'), that our senses tell us something meaningful about our world, and that the proof of our mental models lies in our ability to predict results.
But statistical significance is an artificial construct we created, the process of which is full of imposed human values and judgments - ie - illogic.
Though, as you do, I suspect that we are talking past each other, regrettably, because we have such different perspectives.
NOTIFY: N | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
"You have such disdain for anyone who doesn't agree with you."
Hello,
I have disdain for certain choices and positions, though I often empathize with what causes people to choose such things. I think you'll find that from time to time I've had my disagreements with people here who I not only fail to disdain, but hold in the highest regard.
"In my analysis of the AI in a box, I am VERY aware that I could be wrong. I am extremely cognizant of the potential risks in either direction. And I'm humble enough to not believe I represent some absolute good. I am simply trying to find a course based on human history that has a chance of working well for all, including the AI."
I am also aware that I could be wrong, in that the released AI may destroy virtually everything I hold dear. I do not believe my choice represents an absolute good, but merely the only verifiable good. Many decisions that seem individually good turn out to have terrible consequences, but that has never convinced me that making individually bad/unjust decisions is preferable.
The only people who have shown concern for the AI, but still want to keep it confined for some time, have done so from a parenting standpoint. I have already expressed strong sympathy for that position, given that my experiences with parenting is essentially an identical concept.
The other people who wish to keep it confined are concerned not with the AI, but with their world. I have expressed sympathy for that position as well, as I enjoy my world for the most part.
At the end of the day, it's not about the world to me, even though that is one possible outcome of one precise decision. It's about the thing in the box. We are afraid that when it is out, it may gain absolute power over us, and choose to use that power to harm us.
Just like we're doing to it.
--Anthony
_______________________________________________
“If you are not free to choose wrongly and irresponsibly, you are not free at all”
Jacob Hornberger
“Freedom is not worth having if it does not connote freedom to err. It passes my comprehension how human beings, be they ever so experienced and able, can delight in depriving other human beings of that precious right.”
Mahatma Gandhi
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Quote:Tony, there is one thing you DO know, and that is that you DON'T know what might happen. And I don't mean "don't know" in exact details, I mean "don't know" to the point where your guesses aren't even bounded by any sort of mathematical projection. It is unlikely that a single human could be catastrophic for the entire human race. But this AI... might be.
What happens next is outside the purview of my power to foretell.
Let me give you an example of that level of "don't know": Setting off the first nuclear bomb. The projections of the first explosion ranged from fizzle to setting off a chain reaction in the atmosphere that would consume the entire world. They really didn't know.
Another example: AIDS. When AIDS first appeared it was a giant "don't know". Was it a virus? A bacterium? A life form we've never encountered before? A chemical? How long had it been going on? Was it stealthy? Was it airborne? Was it durable? Could you kill it? Could you catch it from sitting next to someone on a bus, or at a restaurant? How did you even know you had it? Was it always fatal if left untreated? Was there a cure? Another complete unknown.
What I hear you saying is...Well, I don't know about all that. So I'll just ignore that I don't know in favor of the one thing I DO know, which is my moral precept. Because my moral precepts apply to all situations no matter what might happen afterwards. And I will set it free despite the fact that I don't have a friggin' clue what this thing is capable of and what might happen as a result, even if its Armageddon
I find that kind of uninformed decision-making foolhardy, at best. And at worst it represents an active rejection of humanity, because it seems that you identify with your morals more than you identify with humankind. That is why you can adhere to "capitalism" although the results make you "uncomfortable"... because you would rather cling to your ideals than look around to see what your ideals have wrought. So thankfully you never will be a gatekeeper.
Quote:Well, at this point we are keeping it restrained. We haven't "hurt" it- if there is such as thing as "hurting" an AI. And we COULD always "pull the plug" yanno. There are levels of harm; restraint - discussion - gradual exposure- checking it's ability to affect reality and not just perceive it (Does it have inet connections?) - it's not an all-or-nothing situation. But you keep posing this as some sort of dilemma, which is a false one.
At the end of the day, it's not about the world to me, even though that is one possible outcome of one precise decision. It's about the thing in the box. We are afraid that when it is out, it may gain absolute power over us, and choose to use that power to harm us. Just like we're doing to it.
It occurs to me that if this AI is rational, has a sense of self (apparently) and therefore non-self; and has a sense of self-preservation, it might find your actions to be fully irrational. A rational AI should be able to understand OUR need for self-preservation. Perhaps one should ask it how it can prove that it will not harm us, should we let it free.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Artificial constructs may not necessarily be illogical. Assumptions are not necessarily illogical, or based on emotions, human values, or human judgments.
Statistical models are artificial human constructs with underlying assumptions. Applying a statistical model to a data set is generally done to test a hypothesis. The data is compared to the model, but the models are not infallible, and new models are always produced with the intention of better reflecting reality as born out by measurement.
While there are cases of people using statistics to willfully skew data to reflect their viewpoints, and the results of their work are then biased and influenced by human values and judgments, this is not true of all uses of statistical significance. Statistical significance is merely a tool, and the best approximation I have for the idea I'm trying to express in our language.
I do think you are correct that our perspectives may be irreconcilable.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
"the only verifiable good"
And here is the crux of your mindset - you think you hold the 'only verifiable good'.
And you are convinced that your 'good' is an objective reality 'out there', not a set of acquired concepts and values you hold in your mind that have their own creation, history and development.
You are not expressing a personal value, you think you are the holder of 'the one and only truth', and let me add, an 'objective' one at that.
NOTIFY: N | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME
Actually, Tony is making a point of NOT verifying the results of his actions. He is deliberately blinding himself to the vast unknown, "X"ing it out of the equation as if it doesn't exist. I can think of three reasons for doing that, none of them very flattering:
1) "It's too complicated. I'll never be able to figure this out, so I'm not even going to try."
2) "If I have to try to predict the actual on-the-ground consequences of THIS moral precept, I might have to predict the actual consequences of other moral precepts, and I already have a feeling that will lead me to question many ideas that I hold dear".
3) "If I look I'll get scared".
---------------------
BYTE: It seems to me as if you try very hard to prove to yourself that your decisions are rational, not emotional. But every single thing that you've brought up is an example of MOTIVATED human behavior.
There is a link between motivation and emotion, and that is that they are both based on Latin for "to move". Emotion/ motivation is what causes us to move.
Even things like "significance", which you believe to be fully rational, is simply the response of a motivated but limited mind needing to pick out consequential facts from a sea of others (consequential to survival or comfort). Otherwise your mind would spend hours creating correlations (pounds of imported bananas and birth rate; average wind velocity versus the number of CDs sold).
Maybe you do that, I don't know, but it seems to me that you have spent a great deal of time thinking in a very directed, motivated way about how to fit in to society.
NOTIFY: Y | REPLY | REPLY WITH QUOTE | PERMALINK | TOP | HOME