The Rise Of The Machines...

[macdoc]

What's the big deal with "human intelligence" ....it clearly is inferior to other intelligences in various frameworks.

[Brian Peacock]

What's the big deal with "human intelligence" ....it clearly is inferior to other intelligences in various frameworks.

What "various frameworks" are you talking about?

A computer can solve complex non-linear algebra problems more quickly than a human. But although that might mean computers are 'superior' to humans at particular tasks it doesn't make them intelligent does it - let alone a superior intelligence?

Your point runs the risk of becoming somewhat circular.

[L'Emmerdeur]

AI is only 'intelligent' if you ignore the meaning of the word. Check any source -- the concept that will be used to ground the word is 'understanding.' There is no evidence that AI understands things in any genuine sense. See the article and paper I posted upthread.

[macdoc]

Way too anthropocentric. "Genuine sense" what's that? alternate facts?
Do ants understand?
https://www.scientificamerican.com/arti ... wrong-way/

Non-human intelligence (NHI) encompasses any form of intelligence that is not human. This includes artificial intelligence (AI), extraterrestrial intelligence (ETI), and the intelligence exhibited by various non-human species on Earth. The study of NHI also explores the potential for communication with and understanding of these diverse forms of intelligence.

https://nautil.us/why-darwin-admired-th ... rm-361515/

[Brian Peacock]

AI is only 'intelligent' if you ignore the meaning of the word. Check any source -- the concept that will be used to ground the word is 'understanding.' There is no evidence that AI understands things in any genuine sense. See the article and paper I posted upthread.

Aye. The history of a definition of "intelligence" is unavoidably bound to the projects of late 19th and early 20th century social Dawinists and Eugenicists - with all the pitfalls one might expect. Many pioneers in psychology proposed definitions of intelligence that correlated certain abilities and traits ostensibly associated with a 'good' education (such as an understanding of mathematics and music, verbal comprehension and reasoning, and general knowledge etc) to create hierarchies of intelligence, so that someone scoring highly would necessarily be considered more intelligent than someone who scored less well. Unsurprisingly, by these descriptors the well-educated were always deemed more intelligent that the poorly-educated for that reason.

The IQ-test we all know and love (!) is an example of a hierarchical description of intelligence that persists to this day. And while the IQ-test may give some relative measure of a person's intellectual abilities and traits (at the very least their ability to do the IQ-test!) it is, as mac notes, entirely anthropocentric and says very little about what intelligence actually is or might be in a general sense. Nonetheless, IQ-test data is still being trotted out to 'prove' that black, brown, yellow, and poor people are dumber than educated white folks - if only for those who are already invested in that spurious idea to begin with.

This kind of conception seems to inform a commonly repeated view of intelligence as being a function of an agent's capacity to perform tasks that are assumed (or simply asserted) to be significant to the determination intelligence. This is what I was hinting at when I suggested that this kind of definition runs the risk of becoming circular, particularly in the context of machine learning systems and large language models.

[rainbow]

of autonomous, independent agents with intelligence/cognition, knowledge and motivation etc etc.

and you it's not how?
When it does something you cannot imagine or ever do yourself ....how is that not independent and intelligent.

Can it unblock a toilet?

I can't imagine doing it myself.

[macdoc]

Such delicate sensibility

[L'Emmerdeur]

AI is only 'intelligent' if you ignore the meaning of the word. Check any source -- the concept that will be used to ground the word is 'understanding.' There is no evidence that AI understands things in any genuine sense. See the article and paper I posted upthread.

Aye. The history of a definition of "intelligence" is unavoidably bound to the projects of late 19th and early 20th century social Dawinists and Eugenicists - with all the pitfalls one might expect. Many pioneers in psychology proposed definitions of intelligence that correlated certain abilities and traits ostensibly associated with a 'good' education (such as an understanding of mathematics and music, verbal comprehension and reasoning, and general knowledge etc) to create hierarchies of intelligence, so that someone scoring highly would necessarily be considered more intelligent than someone who scored less well. Unsurprisingly, by these descriptors the well-educated were always deemed more intelligent that the poorly-educated for that reason.

The IQ-test we all know and love (!) is an example of a hierarchical description of intelligence that persists to this day. And while the IQ-test may give some relative measure of a person's intellectual abilities and traits (at the very least their ability to do the IQ-test!) it is, as mac notes, entirely anthropocentric and says very little about what intelligence actually is or might be in a general sense. Nonetheless, IQ-test data is still being trotted out to 'prove' that black, brown, yellow, and poor people are dumber than educated white folks - if only for those who are already invested in that spurious idea to begin with.

This kind of conception seems to inform a commonly repeated view of intelligence as being a function of an agent's capacity to perform tasks that are assumed (or simply asserted) to be significant to the determination intelligence. This is what I was hinting at when I suggested that this kind of definition runs the risk of becoming circular, particularly in the context of machine learning systems and large language models.

Generally agree but I dispute the "anthropocentric" characterization of my point. For instance dogs, chimpanzees, corvids, parrots, octopuses, and plenty of other animals display some variety of intelligence--it's fairly obvious that they can understand things. Evidence indicates that AI at this point does not. Remains to be seen whether it ever does.

[JimC]

To me, human levels of intelligence require a conscious understanding of self, the motivations and agency that arise from selfhood, the ability to reflect upon an internal stream of thought, and the capacity for complex language to discuss those thoughts with others. At present, none of those attributes are possessed by either animals or AI.

[macdoc]

Hubris. Why should humans limited intelligence be the touchstone.?
Humans have little or no idea what goes on in the intelligent minds of other species.
Humans are woefully deficient in awareness of sounds, smells, tastes, colours of their surroundings or why Lake Ayr is filling this year.

[Brian Peacock]

AI is only 'intelligent' if you ignore the meaning of the word. Check any source -- the concept that will be used to ground the word is 'understanding.' There is no evidence that AI understands things in any genuine sense. See the article and paper I posted upthread.

Aye. The history of a definition of "intelligence" is unavoidably bound to the projects of late 19th and early 20th century social Dawinists and Eugenicists - with all the pitfalls one might expect. Many pioneers in psychology proposed definitions of intelligence that correlated certain abilities and traits ostensibly associated with a 'good' education (such as an understanding of mathematics and music, verbal comprehension and reasoning, and general knowledge etc) to create hierarchies of intelligence, so that someone scoring highly would necessarily be considered more intelligent than someone who scored less well. Unsurprisingly, by these descriptors the well-educated were always deemed more intelligent that the poorly-educated for that reason.

The IQ-test we all know and love (!) is an example of a hierarchical description of intelligence that persists to this day. And while the IQ-test may give some relative measure of a person's intellectual abilities and traits (at the very least their ability to do the IQ-test!) it is, as mac notes, entirely anthropocentric and says very little about what intelligence actually is or might be in a general sense. Nonetheless, IQ-test data is still being trotted out to 'prove' that black, brown, yellow, and poor people are dumber than educated white folks - if only for those who are already invested in that spurious idea to begin with.

This kind of conception seems to inform a commonly repeated view of intelligence as being a function of an agent's capacity to perform tasks that are assumed (or simply asserted) to be significant to the determination intelligence. This is what I was hinting at when I suggested that this kind of definition runs the risk of becoming circular, particularly in the context of machine learning systems and large language models.

Generally agree but I dispute the "anthropocentric" characterization of my point. For instance dogs, chimpanzees, corvids, parrots, octopuses, and plenty of other animals display some variety of intelligence--it's fairly obvious that they can understand things. Evidence indicates that AI at this point does not. Remains to be seen whether it ever does.

I agree. Your point was far from anthropocentric, as you were/are clearly talking about intelligence in the general sense.

[Brian Peacock]

To me, human levels of intelligence require a conscious understanding of self, the motivations and agency that arise from selfhood, the ability to reflect upon an internal stream of thought, and the capacity for complex language to discuss those thoughts with others. At present, none of those attributes are possessed by either animals or AI.

That describes human level intelligence, but anyone who's lived with animals knows they have environmental- and self-awareness, motivations and agency, and complex languages for intra- and extra-species communication. This is only to say that general intelligence is a spectrum rather than a hierarchy, and that making human comprehensible representational language a condition of the definition tends "intelligence" towards the particular and the circular.

[Brian Peacock]

Hubris. Why should humans limited intelligence be the touchstone.?
Humans have little or no idea what goes on in the intelligent minds of other species.
Humans are woefully deficient in awareness of sounds, smells, tastes, colours of their surroundings or why Lake Ayr is filling this year.

To support the assertion that started this strand of the discussion you have to make the case that machine learning systems and large language models are intelligent minds.

[JimC]

To me, human levels of intelligence require a conscious understanding of self, the motivations and agency that arise from selfhood, the ability to reflect upon an internal stream of thought, and the capacity for complex language to discuss those thoughts with others. At present, none of those attributes are possessed by either animals or AI.

That describes human level intelligence, but anyone who's lived with animals knows they have environmental- and self-awareness, motivations and agency, and complex languages for intra- and extra-species communication. This is only to say that general intelligence is a spectrum rather than a hierarchy, and that making human comprehensible representational language a condition of the definition tends "intelligence" towards the particular and the circular.

I'm certainly not denigrating the interesting complexities of animal cognition, but I still think there is a significant gulf between their mental abilities and human cognition.

[pErvinalia]

To me, human levels of intelligence require a conscious understanding of self, the motivations and agency that arise from selfhood, the ability to reflect upon an internal stream of thought, and the capacity for complex language to discuss those thoughts with others. At present, none of those attributes are possessed by either animals or AI.

Sapience.