iloveoof

I agree that AI companies shouldn’t be allowed to bar distillation under grounds of fairness and societal benefit.

But I don’t see how copyright has anything to do with anything. Copyright is a legal concept, not an ethical concept. Training an LLM on copyrighted material is legal. Ethically, whether a work that the LLM trained on is copyrighted or not has no relevance

Agents are the greatest rubber ducks ever created.

Another perspective is that this isn’t a new problem. Every dev manager or team lead with very strong direct reports has this problem. They don’t lose the ability to reason, they strengthen their ability to reason more quickly, and know when to trust the expert and when to poke holes.

I’m not an artist I’m a software developer, so the tools I use don’t matter, the outcomes do. If I am completing my tickets and the code is high quality and well designed then I am doing my job.

Yes, but it’s usually only hours to validate. I understand the codebase so I have the intuition to know almost immediately if the theory makes sense. Validating and reproducing the root cause is often an order of magnitude easier than identifying a production-reported issue.

I don’t see how that’s possible. When I am doing a comprehensive code review with rigorous functional testing of another developer’s work, it usually takes me around half as long to review as the developer took to write it. That includes the back and forth of MR issues and fixes. If AI writes it why would it take 100x longer to review than write it? At worst case it’s just a draft of something I can develop myself, so it shouldn’t take longer than 20 minutes. At best case it is a code review exercise so it takes me half as long.

Are you using the best models? I feel like my experience is completely different. I am an expert in a small part of a huge monolithic codebase that I’ve worked in for years. When customers report issues that would take me days or weeks to debug, AI can figure it out on the first try.

If landlords can already raise the price of rent, why haven’t they?

I hate how the framing is always that “an AI agent” did something. No, an OpenAI researcher breached Medicare by using an AI agent. These tools only do what they are directed to do. The human operator is wholly responsible for what the tool does.

[flagged]

That’s like saying private libraries are bad for freedom to read, or Blockbuster was bad for freedom to view movies

Distillation is a great thing for consumers. It improves competition and reduces the massive moats that OpenAI and Anthropic have in compute that would otherwise lead them to be duopolists. It’s also only fair that AIs trained on humanity’s wealth of knowledge for Pennie’s allow competition to train on humanity’s wealth of knowledge at market cost.

This is basically a compiler, but we’re moving up a layer of abstraction.

A “distillation attack” is like a concrete company calling a competitor building a factory with its concrete a “construction attack”

This article is terrible and I hate that people always bring it up whenever MUMPS is mentioned here.

MUMPS is easy to read, understand and maintain when written well.

These examples (and the Wikipedia ones) are from code written when hospitals needed to run their whole EMR with extreme storage constraints, so more terse code was necessary to fit more on the disk. Of course they aren’t readable by modern standards.

MUMPS’ bad reputation is totally unearned. Most examples online (the wiki page, the awful Case of the MUMPS article) are examples of MUMPS code from the 60s-70s when storage was a premium and enterprise-size programs had to fit in a handful of megabytes, so there was no room for readability.

MUMPS code is more readable than most languages, it’s very simple and procedural. And the ability to interact directly with durable data the same as you do local data is very nice. I don’t know why no modern languages try to do something similar.

Ahhh! Software engineering!

I could not agree more with EFF.

There’s a difference between training a model and using a model. Training involves copyrighted works but fair use is not just about use of copyrighted works, it’s about whether the use is transformative and substitutes the original market. I struggle to see how is not transformative under these criteria.

The use of the model (being able to output copies of GPL software) is a different question. This depends on the circumstances: if GPL code is exactly reproduced then it very well could be subject to the license of the original work.

I don’t understand the legal objections to the fair use of protected IP. Licenses are legal documents, not moral imperatives. GPL only exists because of copyright law, and you can’t write a license that supercedes copyright law if you don’t like the law.

The Claude Code example is completely different, hosting a repo with the leaked code is clearly not fair use.

This tutorial is very bad, and the time estimates are pretty absurd.

The explanations are extremely short and I imagine a new Rust dev would not understand what is going on.

The Brown tutorial is far better, compare its section on mutables and ownership to this.

And yes, this entirely thing is AI generated. Why was this created?

> Surely there has to be some number that pushes individual wealth from a net benefit for the country to a net detriment? Surely we are seeing lots of data and examples right now where we can start to come up with that line as a society? Surely it’s very reasonable that we might want to develop policy to make it hard for an individual to end up with more than, say, ten billion dollars1. That doesn’t seem very controversial to me.

This seems very controversial to me. Why does increasing wealth eventually cause negative outcomes? What do you mean by “make it hard”?

This article skips over justifying its argument. I don’t think, for example, that someone should have come along to Warren Buffett in the late 80s and said: “Your company is too successful, you can’t be the full owner of it anymore”

Individual, founder-led companies are a good thing, especially if the companies are very successful. You’re more likely to force these companies into being led by investors and private equity if you don’t let the founders lead them.

Also, credit can be frozen and credit cards can be canceled. But the medical record, once stolen, can’t be frozen or canceled, so it’s always valuable

Universal encryption of PHI at rest is going to be INCREDIBLY painful. Hospitals mostly have onprem, locked down mainframe IRIS systems that host data. If the IRIS data is encrypted at rest then it can’t be compressed which means hospitals will have to buy a bunch more hardware which is super expensive, especially these days.

This doesn’t get you much in terms of security. The IRIS system itself is an OLTP so it’s going to need to constantly pass around the encryption key and use it constantly, and if an attacker gets disk access they also will have access to the keys.

So this is a big waste of everyone’s time and money.

Software engineering has always worked this way, just not to ICs.

“The LLMs produce non-deterministic output and generate code much faster than we can read it, so we can’t seriously expect to effectively review, understand, and approve every diff anymore. But that doesn’t necessarily mean we stop being rigorous, it could mean we should move rigor elsewhere.“

Direct reports, when delegated tasks by managers, product non-deterministic outputs much faster than team leads/managers can review, understand or approve every diff. Being a manager of software developers has always been a non-deterministic form of software engineering.

I’m not saying this is aligned with FLOSS, FLOSS is a collaboration model. I’m saying the outcome of easier access to knowledge should be celebrated by supporters of FLOSS. Licenses and copyright aren’t good for their own sake, they’re tools for increasing people’s freedom to use, study, modify, and build on existing software. LLMs are another tool for increasing people’s freedom to make new software or improve existing software.

I don’t know if this author supports OSS but I’ll share this because HN generally is full of people with that mindset.

It’s deeply ironic that if you forget about LLMs and look only at the outcome—-we’ve found a way to legally circumvent copyright and the siloing of coding knowledge, making it so you can build on top of (almost) the whole of human coding knowledge without needing to pay a rent or ask for permission—-it sounds like the dream of open source software has been realized.

But this doesn’t feel like a win for the philosophy of OSS because a corporation broke down the gates. It turns out for a lot of people, OSS is an aesthetic and not an outcome, it’s a vibe against corporate use or control of software, not for democratized access to knowledge.

This is like saying that you should oppose electricity or factories if you don’t want them in your backyard.

This article would have been fine and a good send-off if the maintainers just said they were moving on to greener pastures. The discussion of the Anthropic job offer and the cult-like praise of them seems out of place, especially the unnecessary defensiveness in the tone.

It’s okay to make money and change up your career! But this communication is bizarre.

Try MUMPS, widely used but little training data online. Probably less than some esolangs

Moore’s Law for AI Progress: AI metrics will double every two years whether the AI gets smarter or not.