programjames

I think this is downstream of The Debtor's Revolt[1] and antinormative conspiracies expropriating object-level gains[2] from creators. Those closer to the central bank have lower rates to buy out value-creating startups which are assimilated into the antinormative conspiracy. This ancestral mirroring is a fundamental issue with central planning, anathema to the federation necessary for capitalism to succeed.

[1]: https://benjaminrosshoffman.com/the-debtors-revolt/

[2]: https://www.lesswrong.com/posts/tihhx7iy8C6yyHaC2/the-corner...

put a number on your probabilities. there's a chance you have an aneurysm in the next few minutes and fail to see my comment. if so, oh well, i tried.

> competing ~~and~~ on advertising

FTFY. Do not conflate all competition as good for the consumer.

I feel like people don't understand you're being ironic.

Alternative title:

What I learned spending a lifetime making the world a little bit worse for everyone else for fun and profit.

-----

Their entire mindset seems to be, "how can I take advantage of other people making mistakes to make a few dollars here and there? and how can I induce them to make even more mistakes in my realm of influence so it is even easier to profit from them?"

This is why he is always chasing the latest fad—"dropshipping massage guns, rug tufting, rainbow-loom bracelets, you name it"—people who follow fads are less likely to do their research and more likely to click buy on the overpriced TikTok ad. AI just empowers him to do more of this.

The only skill he seems to work on is his "taking advantage of others" skill. It's astounding that he got to this age without knowing even the basics of programming... and he only learned those basics once it became relevant to taking advantage of others. Rather than trying to create something new, or create a better version of something old, he just wants to disrupt efficient commerce so he can skim off the top.

I'm sure he makes a ton of money doing this. I wish that weren't possible.

In an abstract sense, I agree that I value human output because I value humans, even if I never see those humans or their output. But I only can process 10 bits/second, while there is probably a million times that much being posted to the internet. I simply don't have the capacity to care about it except in the abstract sense.

Will we? I will value output from people I personally care about, because I want to know more about them. But what about the other 8 billion people? Do you value their output because they are human or for some other reason?

Say there was a Chinese artist in a room...

When people communicate to me I want them to maximize my comprehension bitrate. AI output is usually dense in incomprehensibility and sparse in bitrate. But the prompt they fed into the AI? Much higher bitrate. And if it's hard to understand, well I can always ask my own AI to help me interpret.

I understand it is a little selfish to want people to maximize my comprehension (what about the thousands of other readers?). I also understand there can purposes other than communication in writing (fiction -> pleasure, poetry -> art). I agree with you that we shouldn't care if it's AI or human writing for these other purposes. But when the goal is one human transferring bits from their brain to mine, playing telephone with an AI cannot be more efficient (it necessarily decreases channel capacity).

There are two exceptions.

1. When AI is good enough to "read people's minds" it can extract extra bits than what they explicitly meant to include in their prompt... but at that point why am I even bothering to listen to them, and not have the more powerful, smarter AI tell me what they think?

2. AI can run experiments or prove things, and I shouldn't have my AI duplicate that work. In this case I would want them to explain what tasks they set the AI out to do, and then provide the results. I would not complain if the results are an AI writeup, though usually I would want to know what in the results the human author felt was important.

You are referencing this from the article:

> There were some technical findings along the way—for example, Kavish found that DDPM, while theoretically ideal for denoising in diffusion models, diverged, and that flow matching performed much better.

When they say "theoretically ideal" they mean a specific thing: the DDPM loss is information theoretically the number of error correction bits needed to recover an image. At very low noises, this quantity goes singular. This is not actually a problem—you should bound the log-SNR of your schedule by the image quantization level (usually 9 bits) anyway, and a trained bound lands within 10% of this.

Flow matching does not diverge because it removes the log-SNR rate from the loss. Since the rate can span several orders of magnitude over your schedule, you should importance sample during training. My guess is Kavish forgot to do this, leading to divergence.

Maybe don't take words out of my mouth. I studied math when I was younger, and got to the top 0.001% in mathematics competitions. Why did I study so hard? In order:

1. I recognized that understanding math had incredibly profitable applications.

2. It was a sport.

3. It was a hobby.

Do I think the world will look the same when math is "solved"? Absolutely not! And we're seeing it transform right before our eyes. I routinely ask Claude, "here's the general idea for a math/cs/ai problem I'm working on, can you help me figure out the maths? oh and then implement it too."

Could I figure this out myself? Of course. Having the idea is halfway to formalizing it in math. But I remember even last year working on GPU kernels by hand and having a bunch of little arithmetic or indexing errors burn me, because I'm just slower and less perfect at doing the math correctly. Now I basically just tell Claude, "go optimize a GPU kernel for this problem". I still have to tell it what to look out for (architectural, algorithmic, memory improvements), but I'm sure the next model won't even need that.

Chess has always been "just a game". Sure, it teaches valuable life lessons, and it is fun to defeat your opponents through superior training and strategy, but it is "just a game".

Mathematics was historically "just a hobby". Much theoretical work is still "just a hobby". But, for the past few hundred years, humans that better understood the theory could find very profitable applications.

That is no longer true. Computers have solved mathematics, like they did chess two decades ago. A human will never find a better application than a computer, just like they will never find a better chess move. Yet people still play chess, and people still make money tutoring chess.

That is the future of humans in mathematics. Mathematics will become mathematics competitions. It will be just another intellectual sport. Sure, a sport that teaches valuable life lessons, but still "just a game". If you are going into mathematics today, your future career is as a competition mathematics coach.

Yes, people can always claim that wasn't their intent, but what you see quite often nowadays is people openly saying their intent is to get people to take action. "I wouldn't personally... but I have massive respect for those who do. We need more people with the balls to do something about it."

This is probably less restrictive than you are advocating for in your previous comment: as long as there is plausible deniability, the speech goes free.

The standard can be, "don't utter words with the intent of spurring others onto antisocial actions."

Where is the "tab" autocomplete?

I skimmed through the book, and it's lacking the information theory foundations. For example, "trust region methods" come from maximizing the policy's relative entropy (to a reference policy) under a tournament system where high-scoring agents are exponentially likely to survive. In general, a reward is the negative bits it costs an environment to propagate an agent (multiplied by some temperature).

Standard Intelligence released one two years ago:

https://si.inc/posts/hertz-dev/

It's only 8.5B and doesn't sound like it's quite conversational.

Didn't Standard Intelligence release a duplex model two years ago? Sounds disingenuous to market this as a new generation of voice models, when it is really OpenAI finally catching up to the current generation after two years.

https://si.inc/posts/hertz-dev/

Logicians' training is so different from philosophers' that it should be considered a separate discipline, or under the branch of computer science.

The appearance of a logical argument is easier to achieve and often good enough for their purposes (publishing papers, winning lawsuits).

Optimizing for the metric involves:

1. Optimizing for generally applicable skills that the metric is trying to measure.

2. Optimizing adversarially to hill-climb the metric.

You want candidates to do (1) and not (2). You can make them agnostic to the second by setting

    d(expected gain)/d(opportunity cost) = 0
      ==>
    expected gain \propto opportunity cost

It is the case that most metrics are logarithmic: it takes just as much effort to decrease one bit of error as the next bit. So

    log(score) \propto (opportunity cost) \propto expected gain

Thus, for them to be agnostic, you should filter candidates proportional to their log-score on the metric (where 0 is a perfect score). Because generally applicable skills are generally applicable, they will still benefit from improving those, they just no longer benefit from adversarial optimization, unless your score function looks very similar to others who have not adopted this filtering process.

The issue with a hard cutoff is that people near the boundary are extremely incentivized to adversarially optimize, as it is usually cheaper than working on generally applicable skills and actually pays off for them. You see this phenomenon on AoPS where (esp. Californian) students talk about grinding for MATHCOUNTS instead of learning calculus.

Nondeterminism is also a feature, not a bug. If you don't want people to optimize against your filtering process, you have to make it somewhat nondeterministic. For example, better candidates are exponentially more likely to pass the filter, instead of a hard cut-off at the top-100. Then it becomes no longer worthwhile to Goodhart the filtering process, because it barely increases your chances and there are so many more places you can use your time better.

1.0 is "natural units". If your energy corresponds to nats, you should be using temperature 1.0. If your energy corresponds to bits, you should be using temperature ln(2) ~= 0.7. The optimization pressure is

     max nats = max entropy + energy / temperature

Why might energy correspond to bits or nats? Imagine your goal is to play as many interesting games of chess as possible in a tournament. This implies you have to keep winning. If you look at the RL environment from the right perspective, you can turn it into optimizing bits or nats.

You should include the error correction code length in the description length. This means Newtonian mechanics was a much longer theory to describe Mercury's orbit than general relativity. It was only the shorter theory before they had the data showing a discrepancy. Which is the correct approach to describing your reality, because until you see a discrepancy, the extensional properties all follow the shorter rules.

I guess the argument from OP would look like: "Yes, now imagine we poke and extend our universe as far as we can. How much bigger do you think our final 'shortest description' would be? I imagine it may be orders of magnitude more complex."

Well, I can imagine a squared circle... doesn't mean the math checks out. I would reply that you do not have to imagine, you can go about looking at different mathematically possible universes in Tegmark IV and find the expected number of bits for the one you actually exist in. Which is ~0 bits more complex than the shortest description based on the data you currently have.

Also, note that Newtonian mechanics is not actually a very short theory for building a universe, because you have to instantiate every object in the universe. You actually get a lot more of the structure for free with general relativity (re: Wigner's classification of the particles). An observer in a presumed-Newtonian universe calling it a simple theory would be like saying, "I compressed Wikipedia to one byte, just by putting it all in the decompiler!"

That is very unlikely due to Solomonoff induction...

You should look into Solomonoff induction. Nature and physics, absolutely, tautologically, have to follow the "shortest explanation is more likely principle".

https://en.wikipedia.org/wiki/Solomonoff%27s_theory_of_induc...

They're likely talking about style transfer from a decade ago: https://arxiv.org/abs/1611.07004

My perspective is it is impossible to cut through the three layers of bullshit between you and anyone who knows what they are talking about. The only way to do this is with brand-name qualifications, like "MIT graduate", not things that are actually impressive. This is also why you see senior developers saying, "the offers I'm getting are bigger and bigger," meanwhile skilled younger developers need to become a marketing professional just to get an interview.

Recruiters have utterly given up on being efficient in the market. I do not know why, but there is something very wrong given "spamming the same brand-name fish all the other recruiters are spamming" is their only strategy. My guess is there is a combination of bad (or an entire lack of) hygienic data filtering and a disconnect between compensation and terminal goals (hiring the best candidates).

"A handful of years"? This is like the trend in K–12 education of blaming all issues on the COVID-19 pandemic. No, education was in a visible decline for five years before COVID, which means it had probably been declining for another decade or two before that.

If you click the website's name to the right of the title, it pulls up all the submissions from the same site:

https://news.ycombinator.com/from?site=github.com/philipl

My experience was pretty contrary to points (1) and (4). My best teachers/professors directly conveyed information or skills. I found most students did the bare minimum to pass their classes (where "pass" = "not get their parents mad"). I tried to get a CS club started at my highschool and basically no one was interested, not even my friends.

Now, I did have a great coach in middle school who "created the conditions where willing students will learn", but I don't think she would have been a good teacher. She was great at organizing club meetings, finding the right materials to study, utilizing intraclub competition to motivate everyone, and getting her former students to come back and teach in highschool. I'm sure there was a lot more going on behind the scenes that she just knew how to do right, which made the club a whole lot better. But she wasn't a teacher. Closer to an administrator, but I think "coach" in the (m)athletic sense makes the most sense.

And, this is probably why my computer science club was not the success I envisioned. Yes, people are generally underachievers, but I also did not have the coaching skills to create the conditions where people wanted to overachieve.