More podcasts, conversations, talks

Here’s my talk “Are We Cooked?” about the AI Mathocalypse—the one where, every time I give it, I need to update it with all the big theorems AI has proved since the last time I gave the talk. This version is from the Institute for Foundations of Machine Learning (IFML) here at UT Austin, back on October 2 (so, ancient times).


Here I am with Rabbi Neil Blumofe of Agudas Achim synagogue in Austin. Rabbi Blumofe asks me about quantum mechanics, AI, consciousness, free will, why I say “please” and “thank you” to GPT, and why I blog about Israel. The questions from the audience, unexpectedly, were more like “will cryptocurrencies that use RSA or Diffie-Hellman take longer to break with quantum computers than those that use elliptic curve cryptography?” Thanks to Rabbi Blumofe for doing this.


And here’s a podcast with Yu Shen, with Chinese subtitles. I haven’t watched this one and don’t remember what we talked about. (Oh — looking at the title graphic, apparently we talked about when AI will solve the P vs NP problem?)

11 Responses to “More podcasts, conversations, talks”

  1. The Selph Says:

    On whether AI had anything to do with the air space fiasco

    https://youtu.be/UZ5KIqaglxc

  2. Doug S. Says:

    It would be *hilarious* if it turned out that simulating quantum mechanics was in P. It hasn’t been proven that it’s not! 😈

  3. Prasanna Says:

    Scott,
    What are the energy consequences of saying please etc when billion users use it frequently. I recollect Sam Altman did raise this concern. Also it’s not a big leap from here to voting rights.
    Models may be trained to ignore these words if energy use is a concern?

  4. Adam Treat Says:

    Congratulations! Big fan of your new career as celebrity podcast guest 😉 Seriously, I very much enjoy them and also am heartened this media gives you an opportunity to address the greater community with hopefully a lot less toxicity than you get from this blogs comments at times.

  5. Scott Says:

    Prasanna #3: You mean, the energy consequences of the one or two extra tokens? Probably comparable to the energy consequences of people leaving blog comments like the one you just did. 🙂

  6. Scott Says:

    Adam Treat #4: Thanks so much! Yes, old career was proving theorems, new career is going on podcasts to talk about AI proving the theorems…

  7. Andrei Says:

    I miss when I could read your opinions instead of listening to them, which I don’t. I dislike podcasts since they force me to go linearly and at their pace, not mine. (This is of course not a criticism of you.)

  8. Adam Treat Says:

    Scott,

    In the last of the linked podcasts you discuss the four color theorem as an example of a problem we’ve had the answer to for a very long time yet know one truly *understands* why the plane can be colored by four colors. You described how this was an early harbinger of the controversy and upsides/downsides to how we might very well be inundated with an ever growing mathematical castle built of proofs we don’t understand.

    But this makes me wonder… has anyone gone back and checked with an LLM and asked the LLM to help us *understand* why four colors suffice? I wonder if a sufficiently modern LLM can do better than the brute force proof that we found before. In general, many problems in math are open to many different proofs of varying length complexity. I wonder if a new avenue is going to be trying to find the shortest and least complex proof … ie, the ones with highest information density (this reminds me of kolmogorov complexity which is undecidable :P) for a given problem.

    Anyway, it might be interesting to see if an LLM can aid our understanding of this proof that we’ve had in hand for awhile before LLM’s were brought to the table.

    Love the last podcast btw!

  9. Scott Says:

    Doug S. #2:

      It would be *hilarious* if it turned out that simulating quantum mechanics was in P. It hasn’t been proven that it’s not!

    “Hilarious” suggests that this is a question that people somehow never considered or took seriously, when of course, carefully asking this question was the entire starting point of quantum computing theory…

  10. Scott Says:

    Adam Treat #8: The proof of the four-color theorem has been substantially improved since Appel and Haken in the 1970s—I believe you can now just download code that formally verifies the entire thing—but every known proof still requires checking hundreds of cases.

    I agree, it would be interesting to set the latest AIs on the challenge of finding a four-color proof that “doesn’t require massive case analysis” (however we define that), if that hasn’t already been tried (or maybe even if it has).

  11. Scott Says:

    Andrei #7: You could try YouTube’s generate-a-transcript feature? Or maybe a volunteer could have their AI generate some cleaned-up transcripts of these talks and podcasts?

    The massive advantage of podcasting over blogging is that yes, there’s someone there to ask you questions and push back on your answers, but they’ll essentially always do so with an attitude of curiosity and respect rather than contempt and hostility (otherwise, why did they invite you in the first place, and why did you agree?).

Leave a Reply

You can use rich HTML in comments! You can also use basic TeX, by enclosing it within $$ $$ for displayed equations or \( \) for inline equations.

Comment Policies:

After two decades of mostly-open comments, in July 2024 Shtetl-Optimized transitioned to the following policy:

All comments are treated, by default, as personal missives to me, Scott Aaronson---with no expectation either that they'll appear on the blog or that I'll reply to them.

At my leisure and discretion, and in consultation with the Shtetl-Optimized Committee of Guardians, I'll put on the blog a curated selection of comments that I judge to be particularly interesting or to move the topic forward, and I'll do my best to answer those. But it will be more like Letters to the Editor. Anyone who feels unjustly censored is welcome to the rest of the Internet.

To the many who've asked me for this over the years, you're welcome!