jaded
Beercules
B.UTLER
BostonGuy
To all my homies
this is interesting
he’s a hippy
surfer
And kind
I recently posted this on his facebook page:
Dear Mustafa Suleyman,
I just finished listening to your podcast conversation with Sam Harris. There was a lot of talk about "intelligence," and to be frank, I think the subject would have been more understandable if the discussion had been somewhat less abstract, and you would have given some carefully explained examples of how exactly AI has been a breakthrough from earlier programming strategies. Regardless, I am pretty sure that however the word intelligence was employed, a fundamental question was not addressed, and that is: Can AI actually do something "creative?” -- which would seem a key aspect of human intelligence, especially the "peak" intelligence of the world's great historical mathematicians and scientists (as well as musicians, artists, etc.). I think that this is an important question and that a test should be designed for the experts in the field to see if they can solve it with AI.
Would you agree that if such a test can in fact be designed, then it would greatly help define what intelligence really is, as well as if machines are capable of duplicating it?
Just a little background on myself: My name is Chris Boys, and I graduated from UCSB in 1973 with a degree in physics with honors. Mathematics, however, was my best subject, and I was accepted into graduate school in it, though I did not pursue that career path. My background in mathematics gave me the idea for the test, in great part because mathematics is so precise and rigorous without any of the “cloudy,” subjective criteria that would complicate the test in the fields of art, music, philosophy, sociology etc. (or even to a degree in the sciences such as chemistry and physics).
Surprisingly, the actual test then followed quite quickly. It basically involves freezing the knowledge of mathematics at about the year 1850. I see this as training AI to master the high school subjects of algebra, Euclidean geometry, and trigonometry, and then mastering the first year university course in calculus. When I took this course the text was by Thomas from MIT. You would know better than I, but it seems that AI could be trained to master these subjects fairly quickly, including not just solving all of the problems (thousands and thousands if necessary) in the subjects, but also mastering mathematical proofs – how to do a proof.
Mastery of the math courses I have just listed was all that was necessary for learning the next course I took in math, which was one of the most beautiful: functions of a complex variable. Just so, the test would simply involve giving AI the complex variable x+iy, and then asking it to develop the calculus of a complex variable. In principle the test presents a carefully defined task and AI can either do it or not. It also involves a creative challenge (basically the same one that confronted Riemann, Cauchey and others), which follows from the calculus of a real variable but in a significantly non-obvious way. What do you think? I welcome your response.
Beercules
B.UTLER
BostonGuy
To all my homies
this is interesting
he’s a hippy
surfer
And kind
I recently posted this on his facebook page:
Dear Mustafa Suleyman,
I just finished listening to your podcast conversation with Sam Harris. There was a lot of talk about "intelligence," and to be frank, I think the subject would have been more understandable if the discussion had been somewhat less abstract, and you would have given some carefully explained examples of how exactly AI has been a breakthrough from earlier programming strategies. Regardless, I am pretty sure that however the word intelligence was employed, a fundamental question was not addressed, and that is: Can AI actually do something "creative?” -- which would seem a key aspect of human intelligence, especially the "peak" intelligence of the world's great historical mathematicians and scientists (as well as musicians, artists, etc.). I think that this is an important question and that a test should be designed for the experts in the field to see if they can solve it with AI.
Would you agree that if such a test can in fact be designed, then it would greatly help define what intelligence really is, as well as if machines are capable of duplicating it?
Just a little background on myself: My name is Chris Boys, and I graduated from UCSB in 1973 with a degree in physics with honors. Mathematics, however, was my best subject, and I was accepted into graduate school in it, though I did not pursue that career path. My background in mathematics gave me the idea for the test, in great part because mathematics is so precise and rigorous without any of the “cloudy,” subjective criteria that would complicate the test in the fields of art, music, philosophy, sociology etc. (or even to a degree in the sciences such as chemistry and physics).
Surprisingly, the actual test then followed quite quickly. It basically involves freezing the knowledge of mathematics at about the year 1850. I see this as training AI to master the high school subjects of algebra, Euclidean geometry, and trigonometry, and then mastering the first year university course in calculus. When I took this course the text was by Thomas from MIT. You would know better than I, but it seems that AI could be trained to master these subjects fairly quickly, including not just solving all of the problems (thousands and thousands if necessary) in the subjects, but also mastering mathematical proofs – how to do a proof.
Mastery of the math courses I have just listed was all that was necessary for learning the next course I took in math, which was one of the most beautiful: functions of a complex variable. Just so, the test would simply involve giving AI the complex variable x+iy, and then asking it to develop the calculus of a complex variable. In principle the test presents a carefully defined task and AI can either do it or not. It also involves a creative challenge (basically the same one that confronted Riemann, Cauchey and others), which follows from the calculus of a real variable but in a significantly non-obvious way. What do you think? I welcome your response.