Hacker News Reader

Comment by leonidasrup | Hacker News Reader

In case of programming.

How much do the current LLMs invent solutions for user tasks, how much they just copy and adopt existing open-source solutions from from Github and other code repositories?

This not a problem for open-source code under permissive software license, but works derived from open-source code with copyleft software license should be also under copyleft license.

Could the biggest commercial benefit of LLMs be just working around limitations of copyleft licenses?

What is the monetary value of human work put into copyleft software and later used to train LLMs? It's hard to estimate, but the study "Estimating the Total Development Cost of a Linux Distribution", estimated that it would cost $1.4 billion to develop the Linux kernel alone.

https://consortiuminfo.org/metalibrary/estimating-the-total-...

View on HN

They do invent code solution for the problem that exists in your codebase. Latter implies that the code solution LLM synthesizes is usually unique of a kind, so, it's not a copy-paste neither it is a simple extract from "another codebase" and adopted.

IMO they operate pretty similarly to humans - we synthesize our solutions, and therefore build-up our knowledge, by collecting knowledge from multiple other sources, including technical books and blogs, open-source code repositories, and our past experiences.

LLMs can output near-exact segments of copyrighted code used for training.

https://arxiv.org/html/2408.02487v3

I wonder how would Microsoft react if someone would synthesize a code solution based on Windows source code.

https://en.wikipedia.org/wiki/Shared_Source_Initiative

I guess you're not writing code much or haven't done much so as your professional career?

My argument is if AI companies are ignoring copyright law and looking at all training data as commons, then we should look at LLM output as something that is not protected by copyright law.

Of course I'm a bit naive here, because we are talking about the richest companies in the world with lot of money to spend on lobbying (or bribes).

https://www.theguardian.com/technology/2026/may/23/trump-ai-...

https://www.bbc.com/news/articles/c98r8r7dz5no

https://en.wikipedia.org/wiki/Commons

I wanted to understand your background first because what you initially said is a very oversimplified view of LLM mechanics, and generally not quite the way how software is in practice written. Since you didn't answer that question, I will assume that you're not a SWE by a call. To give you an example of what I am trying to convey is: imagine a data-intensive workload hitting your storage/database/kernel implementation, and it's painfully slow, your customers are not happy. Then you as engineer sit down, spend days profiling and understanding the code, researching about existing algorithmic solutions to the same or similar issues found in the wild, you read some open-source implementations of viable approaches, you ditch some, some you take, you also read books, articles, other peoples experiences etc. And finally you end up, let's put it bluntly, with some sharded data structure by which you solve the bottleneck. It's not novel, the technique is so common and is already implemented across many many different products in slightly different flavors so I am wondering why do you think this is not a copyright breach but the LLM, which does more or less the same thing, is?

Reply on news.ycombinator.com