Hacker Newsnew | past | comments | ask | show | jobs | submit | spullara's commentslogin

If a comment is useful I don't really care if it was written by a human or not unless the speaker somehow matters more than the content.

Now define useful, specifically in the context of a comment on hackernews.

An LLM summarizing the contents of a blog post might be useful to you, but is a comment here the right place for something you could geneate on your own?

I would guess for most people here, real insight or opinions from others is the "useful" aspect of reading hackernews comments.

Using LLMs to generate or refine comments only moves things further away from that goal (in my opinion).


It seems like the law was poorly written. If it is civil, automated speeding tickets and red light tickets should be just added to the registration cost. If it is criminal, you definitely need to identify the person in order to prove they are guilty.

They are not required to know who is driving their registered vehicle at all times, just that anyone that is allowed to drive it has a license.

What are some scenarios where a vehicle owner knows that the vehicle is being driven by someone with a license, but not who that person is?

Happens all the time. You and your spouse do the same or similar route (e.g. bring child to school) and a month later you get a ticket. Who was driving that day?

You have one spouse and several children, and all of them have a license and are allowed to use the car?

Your court date will be weeks after receiving the citation in the mail. Most families talk once in a while.

The state is constitutionally unable to force you to testify against yourself because the judge ruled these tickets are quasi criminal

if the actual text of the code isn't the same or obviously derivative, copyright doesn't apply at all.

What does derivative mean here? Because IMO it means that the existing work was used as input. So if you used a LLM and it was trained on the existing work, that's a derivative work. If you rot13 encode something as input, so you can't personally read it, and then a device decides to rot13 on it again and output it, that's a derivative work.

In order for it to be creatively derivative you would need to copy the structure, logic, organization, and sequence of operations not just reimplement the functionality. It is pretty clear in this case that wasn't done.

It's not clear at all.

As a cynical person I assume all the frontier LLMs were trained on datasets that include every open source project, but as a thought experiment, if an LLM was trained on a dataset that included every open source project _execept_ chardet, do you think said LLM would still be able to easily implement something very similar?

There is no doubt in my mind that it could still do it.

Of course, the problem with this interpretation is that all modern LLMs are derivatives from huge amounts of text under completely different licenses, including "All rights reserved", and therefore can not be used for any purpose.

I'm not sure how you square the circle of "it's alright to use the LLM to write code, unless the code is a rewrite of an open source project to change its license".


> Of course, the problem with this interpretation is that all modern LLMs are derivatives from huge amounts of text under completely different licenses, including "All rights reserved", and therefore can not be used for any purpose.

> I'm not sure how you square the circle of "it's alright to use the LLM to write code

You seem like you're on the cusp of stating the obvious correct conclusion: it isn't.


> Because IMO it means that the existing work was used as input

That's your opinion (since you said "IMO"), not the actual legal definition.


LLMs do not encode nor encrypt their training data. The fact they can recite training data is a defect not a default. You can understand this more simply by calculating the model size as an inverse of a fantasy compression algorithm that is 50% better than SOTA. You'll find you'd still be missing 80-90% of the training data even if it were as much of a stochastic parrot as you may be implying. The outputs of AI are not derivative just because they saw training data including the original library.

Then onto prompting: 'He fed only the API and (his) test suite to Claude'

This is Google v Oracle all over again - are APIs copyrightable?


I find the "compression" argument not very strong, both because copyright still applies to (very) lossy codecs (e.g. your 16kbps Opus file of Thriller infringes, even if the original 192khz/32bit wav file was 12,000kbps), and because copyright still applies to transformed derivative works (a tiny midi file of Thriller might still be enough for the Jackson's label to get you)

> This is Google v Oracle all over again - are APIs copyrightable?

Yes this is the best way to ask the question. If I take a public facing API and reimplement everything, whether it's by human or machine, it should be sufficient. After all, that's what Google did, and it's not like their engineers never read a single line of the Java source code. Even in "clean room" implementations, a human might still have remembered or recalled a previous implementation of some function they had encountered before.


> LLMs do not encode nor encrypt their training data. The fact they can recite training data is a defect not a default.

About this specific point, it is unclear how much of a defect memorization actually is - there are also reasons to see it as necessary for effective learning. This link explains it well:

https://infinitefaculty.substack.com/p/memorization-vs-gener...


> This is Google v Oracle all over again - are APIs copyrightable?

No, it is completely different.

Claude was trained on chardet, anything built by Claude would fail the clean-room reimplementation test.


"The clean-room reimplementation test" isn't a legal standard, it's a particular strategy used by would-be defendants to clearly meet the standard of "is the new work free of copyrightable expression from the original work".

See also: https://monolith.sourceforge.net/, which seeks to ask the question:

> But how far away from direct and explicit representations do we have to go before copyright no longer applies?


Copyright protects even very abstract aspects of human creative expression, not just the specific form in which it is originally expressed. If you translate a book into another language, or turn it into a silent movie, none of the actual text may survive, but the story itself remains covered by the original copyright.

So when you clone the behavior of a program like chardet without referencing the original source code except by executing it to make sure your clone produces exactly the same output, you may still be infringing its copyright if that output reflects creative choices made in the design of chardet that aren't fully determined by the functional purpose of the program.


If you pirate a movie and reencode it, does that apply as well? You can still watch the movie and it is “obviously” the same movie, even though the bytes are completely different. Here you can use the program and it is, to the user, also the same.

human developers work best when the user defines their acceptance criteria first.

I have always said that if we had done for developers what we are doing for agents the whole world would have been a much better place.

Perhaps we will finally emerge from this decades-long dark age of bloated, do-everything GUI development tools being the fashionable happy path.

The AWS cli tool wants to have a talk… hard to find a more bloated mess of strung together scripts held together by tape.

I think all the SDKs and the cli tool itself are mostly code generated.

Even if the AI bubble bursts hard, we'll still have all of the better tooling for us actual humans.

But saying "it's for AI" is a corporate life hack for you to get permission to build said better tooling... =)


why wouldn't they just make it so the SDK can't use claude subs? like what are they doing here?


When your company happens upon a cash cow, you can either become a milk company or a meat company.


Have your LLM write a simulation of the deck rather so it can play 10,000 games in a second. I think that is a lot better for gold fishing and not nearly as expensive :)

https://github.com/spullara/mtg-reanimator

I have also tried evaluating LLMs for playing the game and have found them to be really terrible at it, even the SoTA ones. They would probably be a lot better inside an environment where the rules are enforced strictly like MTG Arena rather than them having to understand the rules and play correctly on their own. The 3rd LLM acting as judge helps but even it is wrong a lot of the time.

https://github.com/spullara/mtgeval


Yeah, that's why I'm using XMage for my project - it has real rules enforcement.


I was really hoping they could play the game like a human does. Sadly they aren't that close :)


i'm sure claude has been trained on every open source compiler


in actual datacenters you often don't even bother swapping parts and just let things die in place until you replace whole racks


Not my experience at a hyperscaler, at least a while back. It definitely made financial sense to swap a small part to get a ~50-100k$ server's capacity back online.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: