AMA | November 2025

/ Podcast / 6 Comments

Welcome to the November 2025 Ask Me Anything episode of Mindscape! These monthly excursions are funded by Patreon supporters (who are also the ones asking the questions). We take questions asked by Patreons, whittle them down to a more manageable number -- based primarily on whether I have anything interesting to say about them, not whether the questions themselves are good -- and sometimes group them together if they are about a similar topic. Enjoy!

In the intro I mentioned a couple of my favorite TV shows of this year. Here is a more thought-out list (no particular order):

Pluribus and Down Cemetery Road also look promising, but too early to tell. (There are a huge number of shows I've never seen, so feel free to add recommendations.)

AMA

Support Mindscape on Patreon.

6 thoughts on “AMA | November 2025”

  1. Ophir Averbuch

    I wanted to clarify my question (about the entangled coordination device).

    Yes, the sending of signals would have to be preceded by a classical signal that communicates the bank. But the idea is this: The bank of signals contains 1,000 plans, and it doesn’t tell anybody who intercepts it or receives it which one of the 1,000 will be executed. Each one of the 1,000 plans would work if all the pockets of resistance were to execute the same one, so now after sending the bank (classically), you only want to assure they are coordinated.

    The entangled particles can presumably achieve exactly that. The only problem arises if the empire can intercept one of the particles in such a way that the other recipients cannot know that it was intercepted (and de-cohered). If they can tell that it decohered, they just disregard THAT particle (the coordination can be facilitated by sending 10^6 sets of entangled particles such that you can fall back on ones that were not de-cohered.

    If the problem is just that particles would likely de-cohere naturally when going over large distances, but still these de-coherences can be detected at the destination, perhaps that can be solved again by sending a lot of these particles and just listen to the first that is known not to have decohered?

  2. moe

    I know this doesn’t add much to the conversation, but thank you. This is the stuff I have been waiting for since the beginning of the month.

  3. Perry Whan

    Does chirality affect the compression of information? DNA says the same thing twice; but differently. Does one side contain less information but is more robust? Is there a leader follower dynamic?

  4. Alex Houseman

    On Sean’s dismissal of Superintelligent AI, he has a couple of blindspots which I’d like to point out.

    1) Current Gen AI is just a “token predicting machine” very unlike humans. The point proponents of AGI and SIAI are making is that if you turn your analytical skills on yourself, what is it that you are doing when you are talking, thinking, acting, creating or just living that isn’t “predicting the next token”. I’m not sure I have an answer to that. Sean clearly thinks he does. The high end, real world AIs – very similar to the video generation models combined with the system 1, system 2 reasoning models seem to be creating between them them a world model similar to to a human-level “next token” prediction model.

    2) Sean doesn”t live in a world where smart people can’t just manipulate stupid people. No; He’s right in a sense. There’s many an anti-vaxxer that will be unswayed by our arguments. The examples you should look at are very different. Humans are vulnerable not to reason – but to people and things who know how to pish their buttons; Algorithms, populists, pickup artists, memes, conmen. The evidence is staring you in the face. That’s the source of Musk’s tweet, “ SIAI will one-shot the human limbic system”. Elon is now officially a madman in my book, but he’s closer to the truth than Sean is here.

  5. Steve

    @Ophir Averbuch

    Although I think SC didn’t quite understand your question, I do agree with his answer that entanglement won’t help if used the way you describe.

    To make sure I understand your setup:
    – If all the rebels are going to attack A and the empire knows it, then it will reinforce A.
    – If all the rebels are going to attach B and the empire knows it, then it will reinforce B.
    – If the empire only knows the list of N possible attack plans, it can only reinforce each one 1/N as much.
    – If the rebels don’t all attack the same point, then the diffuse attack is stil crushed by diffuse reinforcement.

    So your approach is to:
    – Classically distribute an ordered list of N attack points
    – Use entanglement to distribute the random number that says which point to attack

    Why I think it won’t work:
    – If the empire intercepts and destroys all the entangled bits, your approach fails because no other rebels receive it.
    – If the empire observes the entangled bits as they pass by, the entanglement will be destroyed, so the rebels cannot coordinate even though they think they can.
    – If the rebel HQ sends many entangled bits, the rebel recipients can detect the presence of an eavesdropper. But critically, they still can’t _do_ anything about it.
    – And worse, if the empire observes sufficiently many entangled bits, they’ll know where to reinforce.

    As far as I know, the only way that entangled bits let you communicate secretly is Quantum Key Distribution. First you use entanglement to get a shared key. Then you check that the key is is shared. If it is, you can guarantee that it is secret. If it is not shared, then either you used a noisy channel, and/or there was an eavesdropper, so you must discard the key and start over. Once you have a shared secret key, you use it to encrypt your message and send it classically. Once you have a shared secret key, it’s no better to send an open list of 1000 plans and an encrypted index into the list. You might as well just send 1 encrypted plan.

  6. Elias

    Hey Sean! If it helps with the podcast advertising business, what most people that I listen to say is “rate us wherever you get your podcasts” or maybe “on iTunes, or wherever you listen”. Though maybe as a loyal user of that app, maybe I should suggest “rate Mindscape on Podcast Addict” instead ;-).

    Also, you should have Nate Soares or Eliezer Yudkowsky on the show, that would be a fun discussion to listen to. They currently have a book to promote, too.

Comments are closed.

Scroll to Top