evnp
How does the Elixir community feel about the ethical complexities of AI, OSS, and learning/teaching?
How do you all feel about the use of AI tools for building Elixir/Phoenix projects, with respect to these topics? Longform context below. I get a strong sense of enthusiasm for these tools and the creativity they can bolster from this community, which is great ~ what’s harder to come by are discussions that range beyond the practical applications of these tools, from folks writing from a personal viewpoint (not a longform blog, which have become hard to take at face value on both sides of the discussion these days).
Thank you for anything you’d like to share.
-
LLMs behind these tools likely trained on data taken from OSS authors without their permission, and without attribution ~ do the benefits of these tools for OSS outweigh this?
-
LLM-driven-development may make it harder for the less-experienced among us (myself included, when it comes to Elixir/Phoenix) to learn our craft and build valuable experience ~ this is highly individualistic, but industry forces and corporate influences cannot be ignored when it comes to the pressures on the budding software dev in 2025.
I know there are other areas of ethical complexity with these tools, but these are the two I’d like to focus on for now. I’m less interested in debate here; mostly just looking for a solid set of data points from members of the community. All thoughts and opinions are welcome.
~
Hi all, I’m passionate coder, tinkerer, and creator who’s spent a couple decades of their career focused on web dev through JS (now TS) and professionally, Python on the backend ~ it just keeps happening.
I have a strong interest in and alignment with OSS ideals, even having had the opportunity and satisfaction of making a few notable contributions myself over the years.[1] I’ve experienced the frustration and joy of pouring hours into OSS project maintenance, fixing bugs reported by eager (occasionally annoyed) folks using your project, for no real reason you can explain to yourself other than “it’s kind of fun.”
Over the last two years, I’ve nurtured a voyeuristic passion for Elixir and Phoenix ~ probably a result of all too much time spent working with JS-fueled hacks- I mean frameworks- intended to paper over limitations of the web platform. I guess one could describe Phoenix Liveview this way, but as someone who’s lived and breathed most of what JS has to offer, I’ll say this websocket-centric approach feels like the right way to hack the limited capabilities of the browser into something new. It’s truly incredible, and I dream of being so lucky as to work professionally with these tools someday.
Last year I had the pleasure of working through George Arrowsmith’s excellent Phoenix course[2] with the twist of interpreting the HEEx into Temple[2] for fun. This might be controversial, but it felt to me like the best medium for expression of UI I’ve ever witnessed ~ sometimes you just gotta try something because it seems cool. The results of that experiment can be found below[4] if anyone is interested. I even had some fun and built a couple of Hex packages in the process[5][6] though progress has slowed since then (becoming a parent; its own set of challenges/adventures).
I’ve sensed a lot of enthusiasm for AI-driven software tools in the Elixir community of late, which mirrors the strong enthusiasm that seems to be taking hold across our field. I have no commentary here ~ to each their own and like most of us (I assume) I’m just trying to adapt to the increasing levels of change around us with as healthy an approach as I can figure out how. My natural tendencies have led to a natural skepticism of AI, but I also put high value on an open mind and supporting the creativity of others’ through whatever methods they personally find to be the most enriching.
I haven’t personally used LLMs much in my professional work, but pressure to do so is definitely increasing by the month (and even week). My small company has an upcoming “AI Hackathon” where I had an idea – combine my interests with those of leadership, and take the opportunity to go all-in on AI tools to rebuild part of our product using LiveView. I’m sure I won’t learn as much about Elixir & Phoenix as I would building the old-fashioned way ~ or maybe I’ll learn more, who knows! In any case, I only have a week.
Repeating my intro paragraph from the start, I’d like to know ~ how do you all feel about the use of AI tools for building Elixir/Phoenix projects, with respect to these topics?
-
LLMs behind these tools likely trained on data taken from OSS authors without their permission, and without attribution ~ do the benefits of these tools for OSS outweigh this?
-
LLM-driven-development may make it harder for the less-experienced among us (myself included, when it comes to Elixir/Phoenix) to learn our craft and build valuable experience ~ this is highly individualistic, but industry forces and corporate influences cannot be ignored when it comes to the pressures on the budding software dev in 2025.
I know there are other areas of ethical complexity with these tools, but these are the two I’d like to focus on for now. I’m less interested in debate here; mostly just looking for a solid set of data points from members of the community. All thoughts and opinions are welcome.
Thank you for anything you’d like to share, and for all contributions you’ve made to this vibrant and welcoming community. Learning through lurking here has been a blast, and your effort and enthusiasm for these tools over the years has made this amazing ecosystem what it is today.
~
[1] GitHub - evnp/tmex: Lightweight tmux cmd/layout composer · 1 shell script · 0 dependencies exc. tmux
[2] https://learnphoenixliveview.com/
[3] GitHub - mhanberg/temple: An HTML DSL for Elixir and Phoenix
[4] cocktails.coffee/lib/cc_web/live/realms_live/world_map.ex at main · evnp/cocktails.coffee · GitHub
[5] regex_formatter | Hex
[6] unique_words_sigil | Hex
Most Liked
AndyL
IMO the best way to address AI-anxiety is to learn the tools really well, and to push for open-source solutions where possible (models, hardware, prompts, agents, etc.)
AI tools for Elixir have made great progress in recent months! Tidewave, UsageRules, AshAI, Hermes, etc. Everyone should learn these, and should run local models with Ollama and Nx. OpenCode.AI is a great model-agnostic coding TUI…
Way back in olden days, computing was Timeshare controlled by corporate. Then came PC and OpenSource that we love. With big AI models, we’re back to Timeshare. But already there are great open LLMs, and the cost/complexity of local GPUs will come down. I bet that local autonomy won’t die.
zachdaniel
As someone who’s written million lines of OSS code I have some perspective here. not all OSS is the same. My projects are made available under the MIT license. I have given my permission to use it for whatever purpose you or anyone else likes. Intentionally.
But as for the ethics here, the things that were wrong and illegal are still wrong and illegal and we should pursue companies to that end. I’m more than happy to see investigations and lawsuits for using copyrighted material, and I hope that any found infringements are punished substantially. The individuals, and the company.
For this one, honestly it’s just hard to say and I haven’t seen convincing evidence one way or the other on how it will impact things. There is compelling evidence in both directions that actually using an LLM to learn will or won’t help etc. But “LLMs existing while you learn” is not something I’ve seen measured. It could perhaps have a negative effect via encouraging bad habits in beginners, but also beginners have always had bad habits and reality is what trains it out of them. (i.e you’ll get fired or reprimanded for bad work or slop, and thus have to invest time learning things). Very up in the air for me though.
sodapopcan
Replace “better” with “easier” and I’d agree.
The ethics are long out the window. I also don’t see the problem with feeding it OSS code, but github have been incredibly vague as to whether or not they are scanning private reops. Maybe they’ve come out and said it at this point? I’m not sure, I can’t keep up.
dogweather
I have some intellectual property experience. I’ve never been persuaded by this point. Legally I don’t think it’s infringement. Morally, I don’t see how it’s against the spirit of the MIT or most OSS licenses.
AstonJ
You may want to have a look at some of these threads too:
- How useful do you think AI tools are/will be? (Poll)
- What are your feelings on AI in general?
- In an age of AI, what can programming languages do to survive/compete?
You’ve sort of answered this in your question (bit in bold). Most OSS contributions were made before AI existed, and if we look at what’s going on in the wider space we see there are numerous lawsuits against AI companies for essentially stealing other people’s work (lots of threads relating to this on Devtalk). It’s a bit different with OSS since most OS licences pretty much allow anyone to do whatever they want - tho it could be argued that the intent was significantly different back then - and that those licenses were more to offer ‘free to use’ type rights while offering guarantees in case the maintainer/s died or lost interest or changed their minds.
I expect the pushback from OSS will come a little later than say with artists, musicians and other industries which are being impacted more severely right now - because the impact in the dev industry may be felt a little later, through job losses (tho it’s already started) or AI companies essentially killing off languages once they’ve used them to get where/what they want.
Many would argue we should exercise caution - history tells us that big tech/capitalism on that level simply cannot be trusted:
There will probably be pros and cos, though some early indications are not good:
Does ChatGPT harm critical thinking abilities? A new study from researchers at MIT’s Media Lab has returned some concerning results.
The study divided 54 subjects—18 to 39 year-olds from the Boston area—into three groups, and asked them to write several SAT essays using OpenAI’s ChatGPT, Google’s search engine, and nothing at all, respectively. Researchers used an EEG to record the writers’ brain activity across 32 regions, and found that of the three groups, ChatGPT users had the lowest brain engagement and “consistently underperformed at neural, linguistic, and behavioral levels.” Over the course of several months, ChatGPT users got lazier with each subsequent essay, often resorting to copy-and-paste by the end of the study.
It may be too early to tell - one thing may well be a given though, we probably don’t want to walk into the age of AI blindly or without caution…
Popular in Dev Env & Tools
Other popular topics
Latest Phoenix Threads
Latest on Elixir Forum
Sponsor Spotlight
Producing high quality Elixir screencasts since 2017.
Categories:
Sub Categories:
Forums
Our Sponsors
Build Elixir applications with speed and confidence.
Supporting innovation across the BEAM ecosystem.
We build reliable cloud platforms for business-critical systems.
Catch errors, track performance, monitor hosts and more.
Practical resources that improve the lives of professional developers.
Producing high quality Elixir screencasts since 2017.
The deployment platform built for Elixir: PaaS ease, VPS control.
Courses that'll move you from confusion to "Aha, now I get it!"