51–60 of 69 posts
Earlier quoted context omitted.
The problem is the feasibility of auditing, not whether the motivation exists. And no, taking their word for it is not adequate.
Then what do you propose be done? If you accuse someone of having an extremely racist AI you need a way to prove they set out to make it that way. You don’t just get to say “It’s racist” and shut it down.
Earlier quoted context omitted.
> The Bekenstein Bound says that any finite volume of space must contain finite information. Finite information can be fully encoded as a number. Ergo, human consciousness is a number. That would be "many numbers" not "a number." So many numbers that it's possible no digital machine can simulate it.
If human reasoning can fit in say a 2m^3 object, you can (in theory) always build a machine that encodes the information in 2m^3 or less.
> Even lovemaking is at risk, as artificially intelligent robots stand poised to enter the market and provide sexual services and romantic intimacy. This perfectly encapsulates why I'm so skeptical of AI alarmists. They tend to boil down the human experience to such coarse terms that it undermines their legitimate issues. > AI is a collection of passive silicon and metal. It does not speak. It has no mind. Like an oc…
It's not clear to me from your post or essay whether you are saying the current state of AI is absurdly overhyped and is not even remotely close to producing actual intelligence (I agree strongly); or if you are claiming it's not possible with computers even in principle because "numbers", in which case I don't understand the justification at all, as it implies a mystical quality to human intelligence you rightly dis…
As for the second point, I don't think there is a mystical quality to human intelligence. I guess its mystical in the sense that no one really understands consciousness just yet. But I think something else is going on other than the brain being a rote inefficient chemical cruncher. I find the whole AGI, singularity, and AI hysteria reductive
Earlier quoted context omitted.
i believe this is a category error. AI is not sentient or aware, it defintionally cannot “do” anything.
I see no reason why "doing something" requires sentience.
here's the relevant djikstra. To be the agent of an event requires agency. We have it, principal component analysis doesn't.
I'm firmly of the opinion that anthropomorphization of code is a category error and a bad idea.
This was an absolutely fascinating read... > An atomic bomb is a frantic explosion. Yet none of these phenomena is active; they are all passive. They lack the power of cause. To me it’s completely incomprehensible that an obviously intelligent and well educated (MD PhD) person can think like this... How is it possible to deny that a thermonuclear explosion can cause... a lot of things...? I’ve lately come to realize…
I think the author is distinguishing between intent/agency/free will in the case of mental action and the empty, meaningless quality of material "causes". For causality and the slipperiness of it, read Hume.
I don’t think going straight to Hume will be very meaningful for me. Is there anybody who has explained empiricism from a more hypothetical-deductive perspective?
This article completely misses what I see as the biggest problem with intelligence in machines: Corporations can use them to be unaccountable. Insider trading? No, that was our "algorithm". Racist policies? Oops, sorry that was just our bot. Inhumane practices? Sorry, we are still "working" on our artificial intelligence since you know it's very difficult. Where most people see mysterious advanced help in day to day…
> Corporations can use them to be unaccountable. Insider trading? No, that was our "algorithm". Racist policies? Oops, sorry that was just our bot. I would say the opposite. Insider trading? No, here are the inputs to our algorithm. Racist policies? No, here are the inputs to our algorithm.
Earlier quoted context omitted.
> The Bekenstein Bound says that any finite volume of space must contain finite information. Finite information can be fully encoded as a number. Ergo, human consciousness is a number. That would be "many numbers" not "a number." So many numbers that it's possible no digital machine can simulate it.
If human reasoning can fit in say a 2m^3 object, you can (in theory) always build a machine that encodes the information in 2m^3 or less.
Earlier quoted context omitted.
Then what do you propose be done? If you accuse someone of having an extremely racist AI you need a way to prove they set out to make it that way. You don’t just get to say “It’s racist” and shut it down.
You look at the outcome distribution.
Ultimately, transparency and reproducibility are the key criteria I think, like it is in science.
Earlier quoted context omitted.
I see no reason why "doing something" requires sentience.
https://www.cs.utexas.edu/users/EWD/transcriptions/EWD09xx/E... here's the relevant djikstra. To be the agent of an event requires agency. We have it, principal component analysis doesn't. I'm firmly of the opinion that anthropomorphization of code is a category error and a bad idea.
Agency remains undefined, just like "sentience" from your other post. Unless you actually define it, how can you possibly know that we have it and principle component analysis doesn't?
Even if we take a connotative definition of agency, ie. that it is some quality or behaviour that humans exhibit, then it remains to be shown that code cannot possibly have agency in this exact sense.
Certainly not all code will have agency, just like all code does not have a "sorter" property exhibited by sorting algorithms. I also agree with you that current forms of AI don't have what we recognize as agency, but I don't think we're as far off as some think, and I definitely don't see how it's a category error to talk about AI in this fashion. Whatever qualities you think humans have that grants agency, it might not be nearly as complicated as you assume.