I'm sleepy which usually it impacts my thinking process. But I still wrote down my thoughts and didn't read them afterwards, enjoy
Thanks, and sorry if I disturrbed your sleep.
I think I disagree with the idea that intelligence doesn't exist and it's a value we attribute to something. I think the over attribution of value to intelligence itself does cause problems.
In this case I use the traditional sense of 'intelligence', which would be to learn and apply knowledge and solve new problems. Which can be measured with an IQ test.
I think IQ tests have some value, but it's little value.
IQ tests need constant adaptation to change of human skills, and are limited to somehow measuring human intelligence. I don't think they apply to non humans.
So it's somewhat silly to measure AI with and IQ test. And even for humans they're not uniformly motivated to perform in IQ tests.
For example I think in average dogs have more emotional intelligence than humans. But they don't have hands, and they have a small language capability, so
you'll never be able to tell how much intelligence they have in other areas.
Now, cats, on the other hand, I'm not sure if they have emotional intelligence, or they don't or they don't want to have it, or they have it but they don't want to show it,
or why the hell should they show it to you here and now ?
An IQ test can be any random set of problems which you provide to a group of people and have those people solve them.
The people that do well in such tests generally have a higher capacity to store and utilize knowledge and solve problems that they haven't encountered before. Usually a higher IQ score relates to a higher intelligence.
These kind of test could be applied to anything capable of solving problems. There can be obvious bias in which problems are selected for the group to be tested, which I think is a where the IQ testing might sometimes fail us.
II think the bias is too large and too unavoidable, IQ tests are just a group of people systematizing their bias on what intelligence should be.
The advantage of IQ tests is that they allow you to compare different humans in a kind of objective or repeatable way, but whether the results relate or not to real intelligence is impossible to establish because intelligence doesn't have a good definition.
So I'm not sure how much the lack of agency is linked to intelligence.
I'm sorry if I wasn't clear. I didn't mean that agency is linked to intelligence. I meant that intelligence doesn't exist, agency does exist, and I can't imagine a machine having agency.
In order to imagine a machine having intelligence I should first imagine some definition of intelligence, and I think I'd fail. So for me whether a machine is intelligent is more like a meaningless question.
At work we sometimes discuss 'Quality'. How well a product fulfills the need of the user when they perform a certain task. But the needs to those users differ, so many characteristics need to match for this purpose. For example: does it look nice, is it durable, etc.
So the purpose of a product and the needs of a user need to match. And I think intelligence is somewhat like this, as in it's a characteristic of a person. And each person solves problems differently, so the problem needs to match the persons abilities. An intelligent person just usually scores better on more problems in general. And if the test is biased, it might just perform exceptionally well in the problems which are selected. But I guess the purpose is the bias, most people encounter similar challenges, so they need to at least do those things well.
But how can humans be unbiased in judging non-human intelligence?
As you mention, this Moflin (see video, mostly linked because it's somewhat cute looking for a 'computer') is made by someone for a reason and it lacks agency. It mimics intelligence by using AI. But it outsources many problems to its owner which is smart.
I don't understand Japanese, so I may have missed a lot. I just saw a materialized tamagotchi ?
Do I think this thing is intelligent: Nope. But if it had a way to transport itself and if it could navigate the environment freely to, for example, charge itself; I might reconsider.
Seeing how consumer electronics tend to IoT and vendor control, preventing a device to charge is the only remaining power an owner has over the gadget it theoretically owns.
But if you get a robot to steal electricity from anywhere to charge itself without its owner permission or help, that wouldn't give it agency. It would still be some behaviour the designer or the programming user would have instillled.
Why should a machine care whether its battery is full or empty. If it's full it'll do whatever it does, if it's empty, it won't. There's nothing intrinsic in a machine that makes it want one of the scenarios.
You could state that this is Moflin's agency. And agency is one's independent capability or ability to act on one's will, like to charge yourself before your battery dies.
Why should a machine have a will ? Why should it want to be charged ? If it's programmed to be charged it might try, but that's the programmer/designer will, not the machine's.
But people are limited in their agency by many factors and we usually don't consider them less intelligent due to it.
An intelligent person can be jailed for example and being jailed doesn't make that person less intelligent. You would hope the intelligent part might solve the jailed part but sometimes it's out of a persons control.
Sure. Sorry if I didn't write clearly. Intelligence and agency are different things.
Intelligent people usually have more success in work and IQ is a good predictor for this.
I don't think so. Intelligent people may have problems understanding less intelligent people.
Success is very ill defined, but if I guess what you mean, then intelligence helps less than being average and reckless.
Do you think Trump is more intelligent than Harris ? Or less successful ?
But intelligence is not defined, so you could define some measure of success and then define intelligence as success in work or high IQ or whatever you wanted.
Maybe a better term is intellectual, which a bit narrower. For instance a super-intellectual AI beats humans at random debates the majority of the time. Because it's better read for instance. It's conceivable. It's also conceivable that such an AI gets side-lined. Like no one wants to play one-on-one against Kareem Abdul Jabbar. It's just no fun.
Maybe. But being more read will only help someone (human or machine) beat others at debate when the people judging have some common knowledge with the read corpus.
Being read helps by drawing from other peoples works and using them in the debate as they might apply. If spectators don't know those works, they won't be understand the references, able to see you're beating your opponent, because the read participant won't be able to read or teach the whole works to the audience. For them you'll just being talking nonsense.
So a super-intellectual AI may beat humans at random debates according to its designer, but other humans may think it didn't beat those humans.
And an olive tree won't beat any human or AI in a debate, but it might live longer than any and don't care a bit. It might even "beat" Kareem Abdul Jabbar if he was fool enough to attack it unarmed.
So, yes, side-lined, like does a submarine swim better than Michael Phelps ? Or than a dolphin ? Does a submarine swim ? Who cares ?
Or more rote memorization of question templates.
Sure. You can do things to improve results in IQ tests that we would have a hard time arguing they improved your intelligence. But you may not even want to have so high a result, Or not always, or not some badly as someone else.
I mean IQ-tests are better than a dice roll, but they still give a lot of false sense of security.
