4 ms·

This is really an excellent essay if you have any interest in AI and the view that it is a philosophic problem (more specifically, epistemology) and not a techn

by dmfdmf 11y ago

This is really an excellent essay if you have any interest in AI and the view that it is a philosophic problem (more specifically, epistemology) and not a technology or engineering problem. My only comment is that Deutsch assumes that Popper's theories are correct but they are derivatives of Kant's false theory of consciousness and false theory of concepts and thus a deadend.

lisper 11y ago
I dunno, I find Deutsch's rendition of Popper to be pretty compelling. What don't you like about it? I haven't read Kant. What do you think he got wrong that Popper followed?
- dmfdmf 11y ago
  Popper just codified the current state of logic in his "paradigm" of how science works and his principle of falsifiability. We can prove something unequivocally false because of Aristotle's principle of non-contradiction but we can't prove anything unequivocally true (i.e. certain). To do the latter we need the principles of valid induction. Popper's model of how science works by proving that old theories are "false" equivocates on the meaning of proof. In his view Einstein "proved" Newton was wrong but this is not the case. Newton's theory is true and still true in the context in which it was defined. Einstein proved a different theory that applied to a wider context of knowledge so he built on the work of Newton. Einstein himself admitted that his theory was implicit in Newton's theory. Scientific progress is the accumulation of that "wider context of knowledge" known (or assumed) to be true, not by proving things false (though this is part of the process). Deutsch is throwing the baby out with the bath water by dismissing the problem of induction as non-sense and a waste of time. The principles of induction would tell us when a there is enough evidence and in what form to validly say a statement or theory X is true. Deutsch is right that Hume's question about how do we know the sun will rise tomorrow is rank skepticism if asked in today's context (we'd have to ignore everything we know about planets, orbits, gravity, etc). But somewhere between primitive man's observation that the sun will rise tomorrow (which is unproved and uncertain knowledge) and modern man's understanding of astronomy/physics we had enough evidence and crossed an identifiable line to certainty. Induction defines that line and after crossing the line the onus would be on the skeptics to explain WHY they think the sun won't rise tomorrow and their explanation would have to be consistent with everything we now know, not just some arbitrary dropping of the context. Everyone is a Kantian today and so I don't think I was going out on a limb to assume that Popper's theory is an application of Kant's theory. The essence of Kant's view is that true knowledge is unattainable because we are dependent on our senses and the world we perceive is "mere" appearances and not reality itself. This is a denial of consciousness as such and also denies that the base of knowledge is sensory data. One trap that modern physics has fallen into is the same Platonic view that the essence of reality is mathematical. This leads to a whole raft of errors and the solution is a clear understanding of epistemology and how concepts work. In any case, Deutsch is groping in the right direction that AI is an epistemology problem not a technology problem which why I liked his article.
  - lisper 11y ago
    > Induction defines that line Really? How? > the onus would be on the skeptics to explain WHY they think the sun won't rise tomorrow You've just contradicted yourself, and sided with Deutsch (and hence with Popper) here by admitting that if a skeptic can explain why the sun won't rise tomorrow that the skeptic might be right. That's Deutch's (and Poppers) whole point: it's not about induction, it's about explanations. BTW, it's a question of when, not if, there will come a day when the sun will no longer rise. The best estimate at the moment is about 5 billion years from now. A better example of the failure of induction is this: Barack Obama has been the president of the United States for about 2500 days now. And he will almost certainly be the president tomorrow, and the next day, and the day after that. But I can nonetheless predict with 100% certainty that, one way or another, he will not be the president one year from today.
    - dmfdmf 11y ago
      > Induction defines that line Really? How? You are asking me to solve the problem of induction. I appreciate your vote of confidence. ;-)
      - lisper 11y ago
        Perhaps I misunderstood you. When you said "Induction defines that line" I thought that meant you believed that the problem of induction had been solved. Not necessarily by you, but if you were claiming it had been solved I was just asking what you thought the solution was. But I'm still confused then, if you concede that the problem of induction hasn't been solved, then why do you find the Popper/Deutsch solution to it (or, if you prefer, end-run around it) so unappealing? Do you prefer a world where this problem remains unsolved?
        dmfdmf 11y ago
        P1: That was my fault, I didn't state that clearly. I should have said "The valid principles of induction define that line..." We have the principles of deduction that tell us when something is false but we are missing the principles of induction that tell use when something is true (in a given context). This is a general need so it spans all theories in all fields; how do we know and are certain the sun will rise, evolution is true, there ARE atoms, etc. There must be a criteria by which to say, in the current context of knowledge we know enough to say something is true and certain. P2: You anticipated my answer, Kant/Popper is an end-run evading the issue, i.e., we can only prove thing false never true and certain. The truth is unattainable. I think it is a valid question and needs to be solved. But to solve it we need a full, valid epistemology explaining how reason works in toto to achieve certain knowledge. I think this is the same problem as in physics and all the contradictions and paradoxes in QM and GR. Physics is hung up on all sorts of confusing issues that seem to revolve around the nature of the observer and time. This makes sense to me because deduction is based on a metaphysical fact that contradictions do not exist which we (try) to abide. But this is half the story, what is missing is an epistemological principle that tells us what our brains do to create true knowledge. Whoever solves this problem will revolutionize physics and allow us to create AI (in my judgement).

by dmfdmf 11y ago

lisper 11y ago

I dunno, I find Deutsch's rendition of Popper to be pretty compelling. What don't you like about it? I haven't read Kant. What do you think he got wrong that Popper followed?

dmfdmf 11y ago
Popper just codified the current state of logic in his "paradigm" of how science works and his principle of falsifiability. We can prove something unequivocally false because of Aristotle's principle of non-contradiction but we can't prove anything unequivocally true (i.e. certain). To do the latter we need the principles of valid induction. Popper's model of how science works by proving that old theories are "false" equivocates on the meaning of proof. In his view Einstein "proved" Newton was wrong but this is not the case. Newton's theory is true and still true in the context in which it was defined. Einstein proved a different theory that applied to a wider context of knowledge so he built on the work of Newton. Einstein himself admitted that his theory was implicit in Newton's theory. Scientific progress is the accumulation of that "wider context of knowledge" known (or assumed) to be true, not by proving things false (though this is part of the process). Deutsch is throwing the baby out with the bath water by dismissing the problem of induction as non-sense and a waste of time. The principles of induction would tell us when a there is enough evidence and in what form to validly say a statement or theory X is true. Deutsch is right that Hume's question about how do we know the sun will rise tomorrow is rank skepticism if asked in today's context (we'd have to ignore everything we know about planets, orbits, gravity, etc). But somewhere between primitive man's observation that the sun will rise tomorrow (which is unproved and uncertain knowledge) and modern man's understanding of astronomy/physics we had enough evidence and crossed an identifiable line to certainty. Induction defines that line and after crossing the line the onus would be on the skeptics to explain WHY they think the sun won't rise tomorrow and their explanation would have to be consistent with everything we now know, not just some arbitrary dropping of the context. Everyone is a Kantian today and so I don't think I was going out on a limb to assume that Popper's theory is an application of Kant's theory. The essence of Kant's view is that true knowledge is unattainable because we are dependent on our senses and the world we perceive is "mere" appearances and not reality itself. This is a denial of consciousness as such and also denies that the base of knowledge is sensory data. One trap that modern physics has fallen into is the same Platonic view that the essence of reality is mathematical. This leads to a whole raft of errors and the solution is a clear understanding of epistemology and how concepts work. In any case, Deutsch is groping in the right direction that AI is an epistemology problem not a technology problem which why I liked his article.
- lisper 11y ago
  > Induction defines that line Really? How? > the onus would be on the skeptics to explain WHY they think the sun won't rise tomorrow You've just contradicted yourself, and sided with Deutsch (and hence with Popper) here by admitting that if a skeptic can explain why the sun won't rise tomorrow that the skeptic might be right. That's Deutch's (and Poppers) whole point: it's not about induction, it's about explanations. BTW, it's a question of when, not if, there will come a day when the sun will no longer rise. The best estimate at the moment is about 5 billion years from now. A better example of the failure of induction is this: Barack Obama has been the president of the United States for about 2500 days now. And he will almost certainly be the president tomorrow, and the next day, and the day after that. But I can nonetheless predict with 100% certainty that, one way or another, he will not be the president one year from today.
  - dmfdmf 11y ago
    > Induction defines that line Really? How? You are asking me to solve the problem of induction. I appreciate your vote of confidence. ;-)
    - lisper 11y ago
      Perhaps I misunderstood you. When you said "Induction defines that line" I thought that meant you believed that the problem of induction had been solved. Not necessarily by you, but if you were claiming it had been solved I was just asking what you thought the solution was. But I'm still confused then, if you concede that the problem of induction hasn't been solved, then why do you find the Popper/Deutsch solution to it (or, if you prefer, end-run around it) so unappealing? Do you prefer a world where this problem remains unsolved?
      - dmfdmf 11y ago
        P1: That was my fault, I didn't state that clearly. I should have said "The valid principles of induction define that line..." We have the principles of deduction that tell us when something is false but we are missing the principles of induction that tell use when something is true (in a given context). This is a general need so it spans all theories in all fields; how do we know and are certain the sun will rise, evolution is true, there ARE atoms, etc. There must be a criteria by which to say, in the current context of knowledge we know enough to say something is true and certain. P2: You anticipated my answer, Kant/Popper is an end-run evading the issue, i.e., we can only prove thing false never true and certain. The truth is unattainable. I think it is a valid question and needs to be solved. But to solve it we need a full, valid epistemology explaining how reason works in toto to achieve certain knowledge. I think this is the same problem as in physics and all the contradictions and paradoxes in QM and GR. Physics is hung up on all sorts of confusing issues that seem to revolve around the nature of the observer and time. This makes sense to me because deduction is based on a metaphysical fact that contradictions do not exist which we (try) to abide. But this is half the story, what is missing is an epistemological principle that tells us what our brains do to create true knowledge. Whoever solves this problem will revolutionize physics and allow us to create AI (in my judgement).