AI Summary
This video critiques Louis Vervoort's book 'Problem Solving in Philosophy: How to Do Philosophy in the Age of Ultra-Intelligent AI,' which argues that philosophy should adopt the methods of theoretical physics and that AI can settle philosophical debates. The host, Sabine, disagrees with both the author and a responding philosopher, offering her own perspective on the limits of AI in philosophy.
Chapters
Louis Vervoort has a PhD in physics and a second PhD in philosophy, lending credibility to his argument.
Vervoort argues philosophy should be done like natural science, using a 'maximally coherent theory' that explains the most cases and contradicts the least accepted knowledge.
Vervoort asked ChatGPT (GPT-4) which notion of causality is best; it answered 'counterfactual causality,' which he agrees with, but Sabine notes this is circular reasoning.
Vervoort believes AI will settle free will (which he thinks doesn't exist) and induction (which he calls a pseudoproblem).
A philosopher's comment argues that if philosophy is about generating theories, AI can do it, but if it's about thinking through problems yourself, it cannot.
Sabine argues that quantifying the 'best' explanation is itself a philosophical problem, and philosophy addresses questions science cannot answer.
Vervoort treats induction as an axiom, but Sabine argues the question of why it works is exactly a philosophical question.
Sabine suggests philosophers will increasingly use AI due to laziness, potentially improving discourse quality.
Sabine rates the book 6/10, acknowledging the method is worthwhile but not likely to make philosophy obsolete.
Sabine concludes that while Vervoort's method of quantifying explanations has merit, it won't make philosophy redundant, and if AI does, philosophers will debate it for decades.
Mentioned in this Video
💡 Key Takeaways
Philosophy as Science
Presents a provocative thesis that philosophy should adopt scientific methods, sparking debate.
00:38AI Experiment Flaw
Illustrates a logical flaw in using AI to validate one's own beliefs, highlighting confirmation bias.
01:46Philosophy's Role
Clarifies the distinction between scientific and philosophical questions, a key conceptual point.
03:20Induction as Axiom
Explains why treating induction as an axiom doesn't resolve the philosophical question of its validity.
04:00Full Transcript
[00:00] how artificial intelligence will make much of philosophy obsolete, soon. A philosopher thinks otherwise. And I think, well, they’re both wrong. Let me explain.
[00:14] The new book comes from Louis Vervoort who started out with a PhD in physics but then did a second PhD in philosophy. So this is not your stereotypical ignorant physicist with a
[00:26] YouTube channel, you have me for that. The book is titled “Problem Solving in Philosophy: How to Do Philosophy in the Age of Ultra-Intelligent AI.” To make a long story short,
[00:38] the author argues that philosophy should be done like a natural science in general, and like theoretical physics in particular. As he writes in the book, quote: “philosophy is now mature enough to come to reasonably solid answers to key
[00:53] problems” and “there is one overarching method to do so, in essence the same method as the natural sciences use to obtain answers, in particular theoretical physics.” He argues that philosophers should quantify how good an explanation is by using that explanation,
[01:12] which gives the largest number of coherent answers to the well-posed problems. He calls this the maximally coherent theory. The best theory is the one that explains the most cases, makes sense of the most counterexamples, and contradicts the least other accepted knowledge.
[01:28] And AI can do it, so his argument, because all it really comes down to is counting the questions answered and contradictions. He has even done an “experiment” in which he used AI to allegedly solve a thorny problem of philosophy, that is how to define causality.
[01:46] The experiment works like this: ask ChatGPT --- at the time that was GPT 4 --- which notion of causality is the best. ChatGPT said “counterfactual causality” which Vervoort thinks
[02:00] is correct because it’s the one he also likes best. Maybe doing philosophy like theoretical physics is not the flex he thinks it is. Other philosophy topics that Vervoort thinks AI will settle is free will, which clearly doesn’t exist, and induction
[02:17] which he thinks is a pseudoproblem. Though I have to add that he also says that he does not think AI will solve all philosophy problems, in particular he says ‘big questions’ like the
[02:29] meaning of life or the existence of god are still up for grabs, presumably until ChatGPT 7.8 max pro ultra agrees with him. The author published a paper with his main argument before the book, so we already have a comment from a philosopher titled “Can we
[02:46] automate philosophy through AI? And should we want to?” The philosopher basically says that if you think philosophy is about generating theories and arguing for and against them, then, yes, an AI can in principle do it. If you, on the other hand, think philosophy is
[03:03] about thinking through these problems yourself then an AI cannot do it for you. The remarkable thing about this argument is that he managed to blow it up to 15 pages which is basically my problem with all philosophy literature, loads of words to say the pathetically obvious.
[03:20] That said, I don’t have much sympathy for the physicists’ argument in this case either. Because the question of how to quantify the “best” explanation for a philosophical theory is itself
[03:32] a philosophical problem. And the entire idea that you can “sciencify” philosophy is nonsense because philosophy is exactly what we use to discuss the questions that science can NOT answer,
[03:45] or not yet. There are of course questions that eventually move from philosophy to science, consciousness is just doing this and free will is likely going to come next. The problem with the author's argument is nicely illustrated in his take on the
[04:00] problem of induction. That’s the question of why induction works, why do the laws of nature continue to work, why does the sun rise every morning, why do atoms remain stable,
[04:13] why does sabine keep making stupid jokes. You might say, well, it’s because we have seen in the past that that’s the case. Yes, but that would be proving induction by induction which is a circular argument. Vervoort basically says it’s a meaningless question because
[04:29] induction should be treated as an axiom and as such it doesn’t require a justification. But induction is an axiom in science. We need it for science to work. The question is why
[04:42] does it work? That’s a question science can’t answer which is exactly why it’s a question for philosophers. If you say a non-scientific question isn’t a question then yeah, I guess you can close the philosophy departments, but even Claude could tell that this is a pseudo-argument.
[04:57] That said, I think both of these gentlemen miss the obvious point, which is that humans are by default lazy. Most philosophy papers are already collections of useless words that no one reads, and of course increasingly more philosophers
[05:13] will let AI do this job. Indeed I am sure this is already happening. And I hope that this will ultimately encourage philosophers to improve the quality of their discourse. I give the book a 6 out of 10 on the bullshit meter. Why is it so low after all I’ve said?
[05:30] It’s because I think that the method that the author suggests, to quantify how “good” different explanations are, is worthwhile. I just don’t think it’s going to make philosophy redundant.
[05:42] And if AI really does make philosophy obsolete, philosophers will spend the next fifty years debating whether that actually happened. If you have a trip coming up, I have a recommendation for how you can make this so much more convenient with the
[05:57] electronic SIM card or eSIM from Saily. What's it good for? If you're traveling, you can avoid roaming fees just by switching to a local eSIM. That can save you a ton of money.
[06:11] Something else that I found it useful for is to keep using an old phone which no longer has a SIM card. You can also use it to set a virtual location. How does it work? It's super easy.
[06:24] Download the app, select a country under plan, and don't forget to enter my code SABINE before checkout because that'll get you 15% off. Download the eSIM and switch it on. And that's it. Yes,
[06:38] it's really that simple. Saily has 24/7 support and if the eSIM doesn't work on your phone, a money back guarantee. So use my code SABINE for an exclusive 15% discount
[06:50] on sale data plans and I wish you a good trip. Thanks for watching. See you tomorrow.