Ask HN: Do you think differently about working on open source these days?

32 points by gillyb ↗ HN
Wondering if anyone here has changed their opinion on open sourcing their projects or contributing to open source now that LLM's are a thing...

I don't know yet what I think, but my latest side project I decided to create privately on github.

39 comments

[ 3.0 ms ] story [ 44.7 ms ] thread
Aaahahah people tend to forget that GPL exists. If they use your software to build another, then you can sue and make all their software opensource. Also, you dont have to use github where your source is easily pray of LLM. You can host your own.
LLMs have not changed my opinion on contributing to open source. Slop PRs are quite easy to deal with, and I'm not too concerned about LLMs being trained on the fairly niche things I work on. If my projects were more popular I might have a different opinion.
For me, I think the value of sharing with other people still outweighs the 'leeching' by proprietary for-profit vendors. Ideally, I'd like to be able to set an option on a repo: no LLM training, only training material for open weight models, and training material for all models. Though sadly, I do not think we could trust all vendors to respect such a flag, so now it is all or nothing.

Personally, I would select only training material for open models. Pandora's box is already open, but if we are going to have LLMs, I want them to be available to everyone and not gated by a small number of companies. Since open models generally have less of a benefit of scale in terms of GPUs, I want them to have the benefit of more available/higher quality training data.

In the grand scheme of things, even GPLing your code doesn't matter as the AI is still trained on it. As AI increases in capabilities, it can just read and rewrite your code. And all code is useful in some way.

As mentioned in the comment, private on GH has no bearing, it is still in full sight of the AI.

From what I can tell, OSS submissions are on the rise as people embrace AI to work on things they could not previously.

It hasn't changed anything for me. I don't really care if AI trains on my code or not. I write open source to share the code for other developers and give back as a way of pay it forward, to all the people's libraries I've relied on before.

Are people open sourcing their works in hopes to make money and that's their concern? I've never heard of that from people involved in open source.

I consider it an honor to have my work be a part of large language models.
Yes, differently: I contribute more now that I don't have to spend time getting ramped up on the code base. I'm fixing issues left and right.
Sure, all the slop code projects I produce get MIT licensed on public repos. It wasn't mine to begin with, so I wouldn't prevent anyone from using it.
I've not released some stuff because I don't want the hassle of dealing with slop in PRs. If someone can't be bothered to write their code, I'm not wasting time reviewing it.

I don't love the idea of my work training an ai without compensation either.

It's impossible to develop open source software without having a credit/debit card and cell phone present. Feels more like a scam to me than not but I understand it from an operator's perspective; you have to prevent fraud. From a contributor's perspective who had their life derailed for more than a year, it's a pain to get everything back up and running as it was before.
Opinion has not changed. I still think information wants to be free, which is why I'm still publishing open source code and encouraging ~~piracy~~ distillations of closed models.
I owe my entire 20 year career to open source, so if anything, I plan to increase the number of contributions I make as time goes on; it's one of the few areas in life where I feel indebted.
I've got mixed feelings.

"A little copying is better than a little dependency" becomes much easier to fulfil with AI, in a literal sense, and I support a lot of "little dependencies"

Me I open source less and 100% keep off github all new code unless it's for pay for a client who wants to use github.
Open Source is one of the most humanistic, progress driving cultures people have ever come up with.

I mean I am fucking shocked that people don't get this, our whole fucking modern world, all the parts that make stuff work, every last bit of it is built on top of or dependent on OSS.

There isn't a single lab, company, person or country that doesn't use and benefit from open source. Whether they know it or not.

It is what has supported widespread fractal improvement starting at the individual level. It's the greatest grassroots story movement ever and it's still driven by grassroot adoption.

It's like programatic peer review writ large with no gate keeping journals and its changed humanity forever, if we ever deflect that asteroid headed towards earth or if we make it to the stars, or if we figure out how to avoid the heat death, it's because open source got us there.

Long before LLMs, I had code on github as a portfolio, none of it was licensed to be copied or modified. People still forked my code.
Without my open source projects, my career would not have taken off. Simple as that.
An oft overlooked part of being involved in open source is the networking and mentorship, both technical and personal. Being a part of one or more well run projects lets you interact with some incredibly smart people discussing technical topics well outside your day job. Not all of that will pan out to something directly useful, but over time it adds up to both a wider skillset at actual development and dealing with a distributed team. The evolution of your written communication skills alone are probably worth it.
Specifically I'm looking for licenses that bar loading the code into LLMs. Github is feeding open source code into Copilot and selling it to people and they need to be stopped.
Not due to llms.

Put simply… Too much drama, too little technical joy.

I had my best technical achievement writing proprietary software at one faang, if i have to be honest.

I kinda want to quit making my code publicly available. It is my believe that the model may output license contaminated code and any license I put into my code will not be accurate, so my policy is I don't use AI coding on my publicly available open source code. However, this also means that I'm working for free fueling the AI industry.

However, it is quite fun to remove the boring part in programming with AI, so any hobby code I write I won't be making them public.

Currently I'm working on a way to use models trained from MIT-licensed code (eg. Comma) by using normal commercial model to supervise it. I believe this make the output code only be tainted with permissive code, and so I can now slowly use AI to write open source code again.

Open Source was always a collection of anarchist experiments in different ways of working. It turns out that it's quite effective and popular, and people will organize themselves naturally around projects they find interesting.

That said, capital has always been squeezing open source. Whether it was the Embrace; Extend; Extinguish mantra of Microsoft, Amazon's hosting of Open Source in AWS to control the market for it, or Oracle's litigiousness about trademarks and patents. To say nothing of all the companies who profit from it and give nothing back in return.

LLMs being trained on Open Source software is nothing new with respect to capital attempting to consume it and profit from it but not giving anything back in exchange.

So no, I'm not worried, I'm not going to change anything. I expect maybe we see a license that says you cannot use it as AI training material at some point in the future, and the lawyers will fight over that for a decade or two.

No, it hasn't changed my approach. I open source things precisely because I don't care if other people profit from it or not. In fact, I *hope* the users of my code gain a better life because of it, just like how I hope the readers of my blog learn something (however small, sometimes).

Whether that process is intermediated by a LLM or not is not really relevant.

I really don't care about LLM training. If I don't want code to be used, I won't share it.