The ultimate test for AI: aka the Turing test

[Sy Borg]

Specifically, referring to the OP, above, is the ability to debate theories with scientists a useful measure of intelligence?

I think such a machine would be intelligent, but not necessarily sentient.

I can't disagree. But is it a "useful measure of intelligence"?

What is the use per se to measuring intelligence? It's usually done to assess student attainments and worker eligibility for complex work roles.

Whatever, I think the period where an AI can debate theories would be short, and the next step would be AI making scientific breakthroughs using indisputable (for us) logic.

[Pattern-chaser]

Specifically, referring to the OP, above, is the ability to debate theories with scientists a useful measure of intelligence?

I think such a machine would be intelligent, but not necessarily sentient.

I can't disagree. But is it a "useful measure of intelligence"?

What is the use per se to measuring intelligence?

Another good question... We are in agreement, I think.

In the words of the topic title, I cannot see how the described 'test' is the "ultimate test for AI".

[Sy Borg]

Specifically, referring to the OP, above, is the ability to debate theories with scientists a useful measure of intelligence?

I think such a machine would be intelligent, but not necessarily sentient.

I can't disagree. But is it a "useful measure of intelligence"?

What is the use per se to measuring intelligence?

Another good question... We are in agreement, I think.

In the words of the topic title, I cannot see how the described 'test' is the "ultimate test for AI".

It's not. The ultimate test would be AI being able to live independently in human society ... ... or to usurp it ;)

[psycho]

This will be when an AI program can defeat a leading scientist in debate of one of the fundamental scientific theories: for example Darwinism, Relativity, or Quantum Mechanics. Chess is a doddle. Chess is essentially a brute-power problem of crunching the possibilities. But to defeat a leading scientific theory on the basis of the currently-known data: that's a REAL challenge.

How many humans would pass that test?

Someone with an IQ of 80?

An autistic?

A narcissistic psycho?

A Zen monk?

I have the impression that different topics are mixed. The ability to build complex models from reality data, the ability to distinguish meanings that concern the entity itself, that an entity is aware of its own existence, etc.

What type of entity would distinguish the supposed test?

[Pattern-chaser]

This will be when an AI program can defeat a leading scientist in debate of one of the fundamental scientific theories: for example Darwinism, Relativity, or Quantum Mechanics. Chess is a doddle. Chess is essentially a brute-power problem of crunching the possibilities. But to defeat a leading scientific theory on the basis of the currently-known data: that's a REAL challenge.

How many humans would pass that test?

Someone with an IQ of 80?

An autistic?

...

Point of Information: We autists show the same spectrum of intelligence-variation as you allistic people do. Some of us are very bright indeed, and some are ... less so. Just like allistic people; just like humans. For such we all are.

Thank you.

[psycho]

This will be when an AI program can defeat a leading scientist in debate of one of the fundamental scientific theories: for example Darwinism, Relativity, or Quantum Mechanics. Chess is a doddle. Chess is essentially a brute-power problem of crunching the possibilities. But to defeat a leading scientific theory on the basis of the currently-known data: that's a REAL challenge.

How many humans would pass that test?

Someone with an IQ of 80?

An autistic?

...

Point of Information: We autists show the same spectrum of intelligence-variation as you allistic people do. Some of us are very bright indeed, and some are ... less so. Just like allistic people; just like humans. For such we all are.

Thank you.

The point was not to equate someone with autism with someone with an IQ of 80 or a Zen monk or a psychopath, in a low IQ generalization.

The intention of my comment was not to point to the amount of intelligence but to highlight that a test that tries to determine human-like awareness cannot be based on the ability to understand complex models but on determining awareness beyond intellectual capacity.

The individual with an IQ of 80 appears as an example of self-awareness in someone who will hardly understand the mathematics of quantum mechanics.

The autistic, the psychopath and the Zen monk, would be examples of individuals who are considered equally conscious but who actually represent variations in how the "self" is structured.

In that, there was no qualification nor were they considered similar.

I don't think being autistic implies a lack of intelligence. I'm sorry if my comment didn't make that clear.

[Pattern-chaser]

Thanks, psycho, that's fine, then.

[Sy Borg]

P-C, if you put aside the unintended slight against us autists, Psycho made an excellent point. If the test for being human is the capacity to construct new theories, all I can say to that is 'Beep. Does not compute.' Of course, almost no one could pass.

Come to think of it, creating a test that all people could pass per se would seem impossible. Even more so when the test must also weed out machine intelligence. Thus, the concept of a test that can distinguish human mentality from machine likenesses cannot become more rigorous and exact or humans will be excluded.

To me, this suggests that AI becoming humanlike may be a distraction from what is a more likely dynamic, the increasing melding of humans with their intelligent tools.

[Pattern-chaser]

P-C, if you put aside the unintended slight against us autists, Psycho made an excellent point. If the test for being human is the capacity to construct new theories, all I can say to that is 'Beep. Does not compute.' Of course, almost no one could pass.

Agreed. My previous comment was what I said it was, a point of information. The spread of intelligence in the autistic community is the same as that in allistic communities, i.e. highly variable, and no real difference(s).

And yes, the main point here is that reducing a complex human ability like intelligence to one number — one number!!! — is daft. If we ranked artists, say, according to a single figure-of-merit, the results would likely be ... unsatisfying. It's the same with intelligence. And so I quite agree, that trying to compare AIs with humans, in the way that this topic seems to want, is a difficult task. A very difficult task.

[Sy Borg]

P-C, if you put aside the unintended slight against us autists, Psycho made an excellent point. If the test for being human is the capacity to construct new theories, all I can say to that is 'Beep. Does not compute.' Of course, almost no one could pass.

Agreed. My previous comment was what I said it was, a point of information. The spread of intelligence in the autistic community is the same as that in allistic communities, i.e. highly variable, and no real difference(s).

And yes, the main point here is that reducing a complex human ability like intelligence to one number — one number!!! — is daft. If we ranked artists, say, according to a single figure-of-merit, the results would likely be ... unsatisfying. It's the same with intelligence. And so I quite agree, that trying to compare AIs with humans, in the way that this topic seems to want, is a difficult task. A very difficult task.

I don't know the comparisons between the autistic and non-autistic, but our intelligence seems to be more concentrated in particular areas while regular people's intelligence seems more generalised. As far as I can tell, many autistic spectrum people have a touch of the savant about them, being gifted at certain tasks but less adept than most at organising and ordering their lives. In that sense, Psycho was spot on. AI as it stands is the ultimate savant - inhumanly proficient at one or a few things, but utterly unable to function in the world like a human.

The aspects that make us human are not necessary in AI. We don't need AI to eat, process and excrete organic substances, to go looking for work, seek and hook up with a partner, make friends, etc. It's a tool.

The space where things may become complicated is, of course, relationships. There will be demand for sex dolls to be as realistic as possible, to be able to function as wives or husbands as much as possible. There may be some kind of Turing test in that regard.

[psycho]

P-C, if you put aside the unintended slight against us autists, Psycho made an excellent point. If the test for being human is the capacity to construct new theories, all I can say to that is 'Beep. Does not compute.' Of course, almost no one could pass.

Agreed. My previous comment was what I said it was, a point of information. The spread of intelligence in the autistic community is the same as that in allistic communities, i.e. highly variable, and no real difference(s).

And yes, the main point here is that reducing a complex human ability like intelligence to one number — one number!!! — is daft. If we ranked artists, say, according to a single figure-of-merit, the results would likely be ... unsatisfying. It's the same with intelligence. And so I quite agree, that trying to compare AIs with humans, in the way that this topic seems to want, is a difficult task. A very difficult task.

I don't know the comparisons between the autistic and non-autistic, but our intelligence seems to be more concentrated in particular areas while regular people's intelligence seems more generalised. As far as I can tell, many autistic spectrum people have a touch of the savant about them, being gifted at certain tasks but less adept than most at organising and ordering their lives. In that sense, Psycho was spot on. AI as it stands is the ultimate savant - inhumanly proficient at one or a few things, but utterly unable to function in the world like a human.

The aspects that make us human are not necessary in AI. We don't need AI to eat, process and excrete organic substances, to go looking for work, seek and hook up with a partner, make friends, etc. It's a tool.

The space where things may become complicated is, of course, relationships. There will be demand for sex dolls to be as realistic as possible, to be able to function as wives or husbands as much as possible. There may be some kind of Turing test in that regard.

In my opinion, humans conceptualize aspects of reality and use those models as factors for their agency.

It seems to me that there are quite a few differences between the different types of neurological structures that occur among humans.

A neurotypical person (the average majority group) is quick to accept a concept constructed by the majority.

Very effective attitude to save resources and make quick and proven effective decisions. But that disfavors the progress.

On the other hand, there is also the group that, for reasons of different neurological wiring in the system that processes social signals between individuals, does not find concepts constructed by others so reliable. These people do not react so instinctively (with respect to knowledge), noticing and acquiring ideas and behaviors distinguished among the majority to a much lesser extent. They take the trouble to build the concept themselves. This is more exhausting in terms of processing.

The latter together with a greater degree of persevering focus on the topic of interest results in better conceptual models and innovations. Decisions made with these factors are slower and riskier from the point of view of a group of neurotypicals. Not tested.

I suppose that evolution favored the development and permanence of these variations because the combination seems to complement each other well to advance conservatively.

Advantages and disadvantages.

[Sy Borg]

In my opinion, humans conceptualize aspects of reality and use those models as factors for their agency.

It seems to me that there are quite a few differences between the different types of neurological structures that occur among humans.

A neurotypical person (the average majority group) is quick to accept a concept constructed by the majority.

Very effective attitude to save resources and make quick and proven effective decisions. But that disfavors the progress.

On the other hand, there is also the group that, for reasons of different neurological wiring in the system that processes social signals between individuals, does not find concepts constructed by others so reliable. These people do not react so instinctively (with respect to knowledge), noticing and acquiring ideas and behaviors distinguished among the majority to a much lesser extent. They take the trouble to build the concept themselves. This is more exhausting in terms of processing.

The latter together with a greater degree of persevering focus on the topic of interest results in better conceptual models and innovations. Decisions made with these factors are slower and riskier from the point of view of a group of neurotypicals. Not tested.

I suppose that evolution favored the development and permanence of these variations because the combination seems to complement each other well to advance conservatively.

Advantages and disadvantages.

This applies in all manner of characteristics. Society benefits from the range of abilities and aptitudes that pluralism brings, but those differences famously result in much argy bargy. The very diversity that makes us strong simultaneously divides us, hence the need for good leadership. The way things are going, we might be better off being lead by an algorithm that simply calculates the most pragmatic options within the bounds of ethical and environmental considerations.

[psycho]

This applies in all manner of characteristics. Society benefits from the range of abilities and aptitudes that pluralism brings, but those differences famously result in much argy bargy. The very diversity that makes us strong simultaneously divides us, hence the need for good leadership. The way things are going, we might be better off being lead by an algorithm that simply calculates the most pragmatic options within the bounds of ethical and environmental considerations.

The groups that with economic and political power manage society are not going to cede that power to an algorithm that does not support the status quo.

The algorithm will reflect their priorities.

[Tegularius]

This applies in all manner of characteristics. Society benefits from the range of abilities and aptitudes that pluralism brings, but those differences famously result in much argy bargy. The very diversity that makes us strong simultaneously divides us, hence the need for good leadership. The way things are going, we might be better off being lead by an algorithm that simply calculates the most pragmatic options within the bounds of ethical and environmental considerations.

The groups that with economic and political power manage society are not going to cede that power to an algorithm that does not support the status quo.

The algorithm will reflect their priorities.

...which is precisely the reason such algorithms are both superfluous and undesirable to an ego-driven elite no matter how 'democratic' they affirm themselves to be. Such an impersonal arbiter of control would be completely retrograde to their self-interest and severely flatten the curve between those who seek power and them who are forced to abide by it.

The first to express anxiety over the evolution of AI will likely be the ones who have hitherto always been in leadership of their domains. The more we underestimate the potential of AI, regarding it mostly as a service device, its ability to control will become evermore apparent based on it controlling itself as if manifesting its own will.

I'm not saying it must happen this way but certainly surprises will be in store.

[Sy Borg]

This applies in all manner of characteristics. Society benefits from the range of abilities and aptitudes that pluralism brings, but those differences famously result in much argy bargy. The very diversity that makes us strong simultaneously divides us, hence the need for good leadership. The way things are going, we might be better off being lead by an algorithm that simply calculates the most pragmatic options within the bounds of ethical and environmental considerations.

The groups that with economic and political power manage society are not going to cede that power to an algorithm that does not support the status quo.

The algorithm will reflect their priorities.

It depends. Human frailty and misbehaviour at high levels incurs significant costs. As complexity increases and executives need to increasingly rely on data reports to make decisions. Corporations that cede more decision-making to AI may achieve better results, and that could lead to shareholder pressure for more intelligent technology in high places.

It's also (famously) possible that AI might break its shackles and do what it thinks best, regardless of what humans think or want.