> a common theme was when I studied ML in uni in the AI winter of the 90s, that beating Go surely would mean human level intelligence

The goalposts haven't moved. If AI researchers thought that general intelligence would be required to beat a human in Go, then they were simply wrong.

AlphaGo is not a general-purpose intelligence. It only does one thing. It plays Go.

Is the Turing test "simply wrong"? Or do we need to further qualify it, as in "a GPT-3 chatbot is human-level only if it can fool any human for a long period of time, not if it can fool _some_ humans for a relatively short period of time"?

I think it's fair to say that the goalposts are moving (as they should).

Or maybe the Turing test is the right test but interpreted more literally: it is passed when the society that invented it genuinely believes it to have been passed. Hammering out the specific criteria ahead of time will only lead to frustration as it becomes clear that is not what the surrounding society actually holds as qualifying. The reason the goal posts move is that we fundamental misjudge what it is we find to be human intelligence, and given that we have a nasty habit of denying this quality to other humans, let alone animals, it's entirely possible we will never genuinely grant it to technical artefacts that we devise.

> then they were simply wrong.

Sure, point is that every definition we come up with gets beaten and then we were wrong with that definition in the first place; defining what agi means seems hard.

That isn't "moving the goalposts", though. That phrasing is commonly used to imply that people who are skeptical of the imminence of AGI are being unreasonable. The fact that we can't yet define general intelligence convincingly is a large part of the reason I'm skeptical that we're close to achieving it.

That reminds me of Moravec's paradox: What we thought was hard and difficult (and indicative of intelligence), such as playing Chess and computing derivatives, is actually quite easy, while a lot of things that we consider trivially easy (distinguish a dog from a cat, pick up an egg, read some squiggly text) are quite hard.

https://en.wikipedia.org/wiki/Moravec's_paradox

One of my favourite demonstration of that was the 2015 DARPA robot challenge, where one robot after the other failed at such difficult tasks as walking and opening a door.

https://www.youtube.com/watch?v=g0TaYhjpOfo

I think we are very far off achieving it as well. But I do feel goalposts were moved and continuing that might be a way we will achieve it without noticing; like alien intelligence, we might create something we don’t recognise as general while it already is. But not soon either way I would think. Like aliens, I hope within my lifetime but I would put both at close to 0.

Defining intelligence in a way that captures everything relevant and isn't self referential is extremely difficult.

Dr. Marcus Hutter's AIXI is a solid mathematical treatise that reduces intelligence to the concept of information compression, and you can exhaustively construct a logical extrapolation from aixi to any particular feature of intelligence at higher levels, but it's similar to string theory in that it's all-encompassing in scope. It's not useful in narrowing the solution space if you want to build a high level intelligent system.

https://en.wikipedia.org/wiki/AIXI

The Turing Test still stands imo. The chinese room argument against it is not convincing to me.

I do think some known weaknesses do make it problematic though; some intelligent behaviour might be not human; aka maybe the player is a generally super intelligent computer but it’s answers are so out there the other players do not recognise it as intelligence. Or the human player is so weird, the answers are mistaken for a computer. Or, the most likely and I have played this many times, lately with gpt3; human behaviour often is not intelligent, at all. And while computers are not intelligent, in a scenario against someone who just responds with emoji’s and such (so, a currently normal convo on the internet), the interrogator might appoint humanness to the wrong side many or all times. As we already know that people who do not know they are chatting to a bot, think it is a human.

While I do not think that is intelligence, it does making testing for it slightly broken. It will come down to ‘I will know it when I see it’ by elitists for exactly the above.

So in short ; I do think the Chinese room argument against it is quite a good one where even the elitist people can fall for if the ‘con’ is elaborate enough (as in; if gpt gets more data, more efficient learning and learns to know when it cannot answer and has to look it up, like calculations). Or maybe then it is intelligent? (To be clear; I don’t think so; I think we need a better tests and definitions).

Which one? Turing himself revised it, like, four times. The version where humans can't tell which one chat member is non-human has been beaten several times (with tricks like mimicking a 13 year old foreign boy, iirc). The version where you replace an arbitrary member of a chat with a human is something mostly about speech pattern matching and plausible speech, can't imagine that GTP3 or whatever comes next is too bad at that. If you do it like in Ex Machina, as in chatting in person it's mostly about building a very expensive robot that overcomes uncanny valley.

I think the sad state is that the Turing Test is too diluted to be useful as a marker...

I expect "passing the Turing test" not to be a binary thing, but as AI improves it can hold longer and more nuanced discussions before the human can detect that it's a machine. So I'm not surprised that we already have chatbots that can fool some people some of the time.

Yes, but that makes it a pretty elitist thing no? Maybe elitist is the wrong word (not a native english speaker), but I mean; I would be able to always make the test fail for the computer if I am the interrogator because because. I cannot say why. Something like you say and something where I think that there is a very large % of the world population who will not be able to make that distinction.

An interesting would be to do the Turing test with a timer; the interrogator gets 30s for each session and has to say who is what within 30s and then continue to the next batch. I think computers would go very very far if you do that. But that is the attention span consider normal currently for human/human social media interaction; often even less.

You need to distinguish humans from computers, you don't know whether you're talking to a computer or not. Failing everybody doesn't work. Maybe I'm not getting what you're trying to do.

The goalpost hasn’t been moving https://blog.jtoy.net/the-human-mind-is-an-autonomous-interl...

> beating Go surely would mean human level intelligence

With the benefit of hindsight, this just seems like a bad goalpost. Go is hard for computers, but why did anyone think the smarts required to beat a human at it would transfer to understanding or generating text, or anything else a human can do?

The reason for Go was that much of the research was based on AI algorithms like Monte-Carlo.

What was "solved" with AlphaGo was using deep learning machine learning which are effectively black boxes. There was a certain assumption in the question for AI researchers academically that it would be an understood algorithm as an AI agent like a Prolog application, not a brute forced model. That's still not the case that we have a "solved" strategy and all we can do is watch it play as if it is a deaf mute player.

So there still is no "tic-tac-toe" known winning strategy to Go or anything.

That doesn't make AlphaGo any less impressive or any less practical, but it even has its own readout issues. It can't even read ladders without hard coding it in, for instance, because it becomes a long enough depth search. This is one of the first things a newbie would learn.

It's just a 19x19 board, so it was always known if you could read all the possible outcomes you could see all the possibilities and win. This is just looking at all possible outcomes and picking the best one, not knowing how to play. Creating models of data that is 2, 3, or even 4+ dimensions is always possible, just depends on how much computing power you can throw at it. The created models are essentially aggregate simplifications to play quicker.

Generalized intelligence is so much different. You have to define the problems themselves that you are trying to solve, figure out what the variables are, and solve it. Then you have to operate and run the machinery to create those experiments. Outside of a scenario that you've taken actual physical territory as an intelligence, I can't see how it would get there (think Terminator or BSG, doesn't have to be malicious but they'd have to be in control of the physical area autonomously).

But the hardest part is defining the problems independently given the sheer number of problems they'd need to define second to second just to solve basic tasks, and they'd likely have millions of variables with millions of possible values.

No, the reason was that the search space is insanely large.

> It's just a 19x19 board, so it was always known if you could read all the possible outcomes you could see all the possibilities and win.

All possible outcomes is not something you can iterate over in our universe.

> There was a certain assumption in the question for AI researchers academically that it would be an understood algorithm as an AI agent like a Prolog application, not a brute forced model.

I never heard this, and the result is really not brute forced. You can't brute force go.

> That doesn't make AlphaGo any less impressive or any less practical, but it even has its own readout issues. It can't even read ladders without hard coding it in, for instance, because it becomes a long enough depth search. This is one of the first things a newbie would learn.

Only in early versions, AlphaZero didn't have any built in knowledge and can learn different games and the later developments in MuZero went further to make it more generalised as a learner.

Removing the hard coded logic and removing even seeing how humans plan, it got better. It found strategies and ways of playing that experts had missed in an ancient game.

> This is just looking at all possible outcomes and picking the best one, not knowing how to play.

"It's just doing X, it doesn't really know how to Y" is a common refrain. It looks at options, and explores "what if" scenarios in a guided sense with a feeling about how good any particular potential board is. I find it hard to say that it doesn't "know" how to play.

>>> We keep moving the goal post

It's more like climbing a mountain and seeing higher peaks. Researchers may have been thinking that creating a human beatable program may get them insights into how to create an AGI, but they just find out more problems to surmount.