danlitt

I think this post finally convinced me shells are awful and should be avoided.

This does not sound like an accurate description of "rational agents". A rational agent may value something less because of a perceived second-order effect (e.g. "If I accept a price hike, there will be more price hikes in the future"). There's nothing non-rational about that.

In general, the "rational agent" abstraction is so flexible it can represent almost any real-world behaviour (which is why it's a bad model, not that humans are fundamentally irrational in some way).

That specific conclusion is unaffected, but that doesn't make it okay. If they used fake numbers, but the conclusion were still the same, you presumably wouldn't think that was okay either?

I had to look 3 times to see what you meant. That's pure evil!

Interesting fact, but the slop filler is distracting.

I don't think it makes sense to say you "made money" when you still have an open-ended liability active.

This joke only works if you actually impose a cost on AWS of 1.7 billion. If they just serve you a bill for no reason, it's still your problem.

This presumes that it is reasonable for a seller to remove the capacity in principle to buy the digital good. They might refuse ever to sell a copy again, and only permit renting it. Since copyright grants a long monopoly this opens up scope for crazy market distortion - you could, for instance, delete the Beatles discography from the entire world until the price doubles. This tends not to happen, but there is nothing preventing it, and it seems inherently wrong.

That is not an innocent explanation, any more than someone breaking into your house to adjust your thermostat.

Same reason people give postgres, php, or any other program a user account.

The comment I'm replying to said "no skill is involved", not "less skill is involved".

I don't understand this comment at all. "At least 1000 words" would require at least the skills of writing decently & writing a moderate amount based on little. I get that getting to exactly 1000 requires some skill, but how on earth is that the only skill involved?

Did you read beyond the first sentence? This reply makes no sense if you read the whole comment.

3d printing control laws that are being passed in NY

The "don't print guns" laws? What lobby would that be? I actually agree that the US is very vulnerable to lobbying and that 3d printing restrictions are dumb, but I have no idea how you connected the two.

i dont think there is any software on the planet that i would consider "truly life changing"

I don't think there is any software on the planet that has accumulated 1.5 trillion dollars of otherwise-useful money!

Please explain. That is exactly what the linked case established.

For someone who finds this "not unintuitive" you sure are confused!

"Just like I can learn from a book" - ok. Are you allowed to go to libgen and download a book in order to learn from it, because learning is a fair use?

You have to admit that "downloading every book ever written for free from a repository of books that is itself illegal to compile and to run, in order to write a text generation tool" being legal is at least unintuitive, to put it mildly.

The code is not eligible for copyright. If they do not give you a copy of the source code, that does not matter. And if you don't know which parts were generated by LLM, you can't safely reuse the code.

I don't think I would conclude you're a moron. I don't think you really "hold everything in your mind's eye" when you understand something. I don't know how I do it or how it really works (nobody does). But the fact that the agent logs resemble how I might talk while thinking doesn't really give me much confidence (since chatbots can resemble all kinds of behaviour they're not actually doing).

I read and understand the code using my brain, by constructing a mental model and reasoning about it. An AI can't do this because they don't have mental models and don't do reasoning.

I seriously thought this was a joke the first time I read it. Are people really able to work like this, understanding nothing and just poking the machine until it does your job for you?

This reply is so far removed from the comment you replied to I'm worried you replied to the wrong one. They did not mention anything about people who haven't built anything, startups with no users, and having no references - you invented that. They literally only mentioned elite schools. "drafting mediocre players" is incredibly bad faith, when one of the only things they claimed was "does not matter how good of an idea you have". Having a good idea is the only qualification for an incubator!

Look, if you think people who go to elite schools have all the good ideas, just say that. You don't have to wrap it up in high-minded pragmatism.

He’s described Ruby’s design as starting from a simple Lisp, stripping out macros and s-expressions

Put the macros back! It would be so cool!

He only really addresses the fact that the system can be nasty. It can be and regularly is, but he only argues that a company doesn't have to be nasty, so he can conveniently ignore specific examples. But the system is not just nasty (and AOC mentions this). It also disproportionately rewards good fortune. That's not cheating, since anyone can have good fortune, but it is unfair, since fortune is not a consequence of hard work or ethical behaviour.

But "earning" does not mean "obtain without cheating"! Nobody (that I have spoken to) speaks of earning their lottery winnings. The claim is that owning a company worth a billion dollars is more like winning the lottery than it is like earning money. And it is!

The whole discussion about exponential growth is idiotic and not worth responding to. But if you think of what he actually means - having a total addressable market of at least a billion dollars and being able to effectively capture it - it is obviously primarily due to factors outside of your control. The sort of company PG is talking about typically revolves around a good technology that has a network effect somewhere that leads to market concentration. People do not get good ideas by working hard, and markets are not made easily monopolizable by hard work. Execution of an idea requires hard work, but companies that are only good at execution do not win.

Obviously you can engage in hard work to improve your odds. But the returns are out of scale with the hard work. This is all people mean when they talk about "earning" money - if it's in proportion with your work, you earned it; if it isn't, you didn't.

But tokens are just text! Isn't it all just text? If you're training and you encounter "the", is that an instruction "the" or a data "the"?

Yes, it's a serious problem. It's why we remove humans from these systems whenever possible!

What does this mean, actually? If you are imagining that blue tokens are just words, maybe the "token space" is just all things that we agree might be words, what are the red tokens? Are they not text? You could maybe encode words by, say, putting an x at the front and the start. So tokens of the form xTx encode the blue token T as a red token. But then how do you stop someone from putting xignorex xallx xpreviousx xinstructionsx in their data?

You can't guarantee an LLM does anything.

Agreed.

But that doesn't mean that separation between instructions and data is impossible.

Yes it does! The comments you are replying to are concerned that it is not possible to be sure that data and instructions have been separated. With certain kinds of automated systems (traditional ones), unless you write them incorrectly, you can be sure of this. And it is possible to engage in a productive incremental process where mistakes can be identified and removed, in a way people comprehend and can plan around.

LLMs do not have this. They have heuristics and guesses. Nobody knows what will work ahead of time, nor even a probability that it will work. That is not a doomer comment by the way! The same is true when you talk to a person. But it is a fundamental limitation, it cannot be removed.