zebomon

Seconding this. Maniac is a beautifully written book about von Neumann. It also led me to Benjamin Labatut's When We Cease to Understand the World, another great read.

I was glad to see the direction this essay took. I agree with you: it's the old challenges that are gone and there's a whole new class of challenges (the former impossibilities) in front of us.

Posting for visibility a reminder that this type of watermarking is comically easy to beat. See Scott Aaronson's 2023 research in which he coined the term "pineapple attack": one must simply prompt the model to add the word "pineapple" after every other word and then do a search + replace to remove the word, to obliterate the watermark altogether.

https://www.youtube.com/live/2Kx9jbSMZqA?si=0QgCPBX2_KPZ0QTU...

I built and open sourced something similar between 2024 and 2025.

https://github.com/humthentic/itypedmypaper-v1

As others have pointed out, it's relatively a lot of effort to create an artifact that realistically current systems can pretty well forge.

I don't know that there is a scalable and comfortable solution to this problem (or at least one that is scalable and comfortable proportional to the demand for it).

I think you're thinking of Down and Out in the Magic Kingdom, by Cory Doctorow.

This exactly. I don't understand the argument that seems to be, if it were real intelligence, it would never have to learn anything. It's machine learning, not machine magic.

I saw him perform 26 times in my life, and still those were rookie numbers. Still I thought there would be so many more too. Thanks for all the music, Bobby Weir.

I've been doing pretty much the same thing since 2019. The only big change I made was in early 2023, when I started saving a new version of the long txt file each day. It works very well for me but I recognize it isn't the right system for everyone!

I won't be providing my full name and email without more information, like at least a peek inside. Interesting legal argument regarding the brand though.

The author's argument seems to be a practical one and two-part: 1) without death, there's nothing to motivate us to live life well and 2) unless we live life well, there's no point in living.

I just disagree with both postulates, and that's fine. The author can go on thinking that life needs to be something specific in order for it to be desirable. I myself like being productive. I also like eating fast food every once in a while. I think I'd be able to go on living (with some happiness to boot) if I never had another productive day or another McD's burger ever again.

Life can be its own end. If we manage to end death by aging, someday there will be children who have never known another world, and they'll marvel at all the death-centric thinking that permeated the societies of their past.

It honestly seemed like pretty sharp marketing to me already when I read about it on AV Club.

I first used Duolingo back in 2018. That was how I started learning French. I majored in Classics in college and had taken Spanish all eight years of middle school and high school, so my vocab progress was very fast. Within that year, I felt like Duolingo had become too slow, and decided to switch my learning over to reading books and watching movies in French.

Earlier this year, I got back on Duolingo because my partner and her brothers were trying it out, so it was more a social thing than anything. I was on it for about a month before we all agreed that the quality was too poor and the pace too slow for it to be worthwhile.

Duolingo is a case study in a good-enough-to-ship product that needed improvements and instead got dark-patterned into something much, much worse than it had been previously. I'm sure there are many superior platforms for language learning online today. I've gone back to books and movies. I'm currently enjoying watching Blaise le blasé (a Quebecois cartoon) and reading Chair de poule (Goosebumps in translation).

I can see that your approach would be effective at shutting down conversation and stopping people from telling you their wild conspiracy theories.

I think that with my experience, I've had to recognize how fragile some of the most important incentives are. Like the safety that underpins trust. To have trust, it needs to be safe for people to be wrong. That means I often have to listen respectfully to views that I find abhorrent, in order to get to the point that I can share my own thoughts fully.

My experience has been very different from the one you're describing.

The person I was talking to is someone who cares deeply for me (and whom I care for deeply too), someone I've known for almost my whole life. He wasn't having fun contradicting me. In fact, it was making him visibly uncomfortable to do so. He was engaging in the conversation in good faith. He just doesn't have the foundation to understand what he doesn't understand. I'm optimistic that even though he came away still disagreeing with me irrationally, there is a chance that by exposing him to a fuller explanation, he'll seek out more information for himself at some point in the future.

It's a combination of never having learned the basics of science and now seeing the falsehoods they've been fed as equivalent to science.

Take the Tylenol thing. You can explain to one of them the scientific method, what a survey of studies is, why correlation often appears when there is no causation, etc. I experienced this last week: at the end of my explanation, the person (a 45-year-old) replied that he "simply disagreed."

The coal, the climate, etc. are all the same. There is a broad sense that because they've been convinced of the value of expanded oil drilling through lines like "Drill baby drill," their current perspective on it is of the same merit as actual scientific research.

I think providers of so-called "answer engines" will read the writing on the wall and find new ways to support content creators in order to keep their databases fresh, relevant, and centered on human perspectives. I also think that to see how the economics will play out, you only need to look at similarly centralized systems like Apple's App Store and Google's AdSense.

Because the providers will act as a single tunnel that all content passes through before reaching the end user, the tolls they collect will be large. So, I don't doubt that there will still be opportunities for content creators to earn money as answer engines siphon off more and more of the web's traffic, but expect those opportunities to be broadly low-paying, falling decidedly in the "side hustle" category.

AI providers will want to incentivize content creation. There will still be a glut of ready providers, and little reason for providers to make anything but small, nominal payments.

I've been using the voice chat in ChatGPT more and more frequently. I'm curious now to see how the costs associated with this would work through the API on some user-facing features. It's a cool update at a glance.

Intriguing analysis. I'll be following along with interest!

This is very interesting. I think a lot of people may be quick to overlook the value of such simulators when thinking about AI agents at the extremes. (Either they're not good enough to trust or they're so good they'll leapfrog over any economic value here.)

My own experience makes me lean toward thinking that the truth is somewhere in the middle in this situation, and that simulators like these will be valuable. I've been experimenting a lot with computer use on my website Bingeclock, passing through different prompts along the lines of "make a movie marathon based on X." The newest agents are consistently impressive, while also being consistently imperfect in surprising and interesting ways.

Whether or not all the labs are already running this kind of thing internally for themselves, you would know better than I. But it's an idea that seems very useful nonetheless. Congratulations on the launch!

I think this attitude is here to stay: people don't like reading something only to realize that it's been written by an LLM. That's only partially because of what the author describes here (low value to word count ratio). More fundamentally, if a human couldn't be bothered to write it, then there better be a very good reason that I as a human am being bothered to read it.

This attitude provides a clue as to 1) the ways we're using LLMs now that will soon seem absurd (an LLM should never make writing longer, only shorter) and 2) the ways LLMs will be used after the novelty wears off, like interpreting loosely specified requests into computable programs and distilling overly long writing to maximize relevance.

Creator here, just seeing this. Thank you very much for the share! I'm reading all the comments now and will make updates to our database accordingly.

Very engaging look at a very difficult topic to approach analytically.

I'm reminded of something I learned about the founder of Stormfront, the internet's first white supremacist forum. His child went on to attend college away from home, her first time away from her family, and over a period of roughly two years, she attended dinners with a group of Jewish students who challenged each of her beliefs one at a time. Each time, as she accepted the evidence her friends presented to her about a particular belief, she nonetheless would integrate the new information with her racist worldview. This continued piece by piece until there was nothing left of her racist worldview at all.

It's both heartening and disheartening at the same time, because if this person can change her mind after almost two decades of constant indoctrination during her formative years, then surely anyone can change their mind. That's the heartening part: the disheartening part is, of course, that the effort it took is far from scalable at present and much more difficult to apply to someone who remains plugged into whatever information sources they are getting their current fix of nonsense from.