Developing cutting-edge AI systems like ChatGPT requires massive technical resources, partially because they’re costly to develop and run. While several open source efforts have attempted to reverse-engineer proprietary, closed source systems created by business labs similar to Alphabet’s DeepMind and OpenAI, they’ve often run into roadblocks — mainly as a result of a scarcity of capital and domain expertise.

Hoping to avoid this fate, one community research group, EleutherAI, is forming a nonprofit foundation. The organization today announced it’ll found a not-for-profit research institute, the EleutherAI Institute, funded by donations and grants from backers, including AI startups Hugging Face and Stability AI, former GitHub CEO Nat Friedman, Lambda Labs and Canva.

“Formalizing as a corporation allows us to construct a full time staff and interact in longer and more involved projects than could be feasible as a volunteer group,” Stella Biderman, an AI researcher at Booz Allen Hamilton who will co-run the EleutherAI Institute, told TechCrunch in an email interview. “When it comes to a nonprofit specifically, I believe it’s a no brainer given our concentrate on research and the open source space.”

EleutherAI began several years ago as a grassroots collection of developers working to open source AI research. Its founding members — Connor Leahy, Leo Gao and Sid Black — wrote the code and picked up the info needed to create a machine learning model near OpenAI’s text-generating GPT-3, which on the time was getting quite a lot of press.

The corporate curated and open sourced The Pile, a set of datasets designed for use to coach GPT-3-like models to finish text, write code and more. And it released several models under the Apache 2.0 license, including GPT-J and GPT-NeoX, language models that for some time fueled a completely recent wave of startups.

To coach its models, EleutherAI relied totally on the TPU Research Cloud, a Google Cloud program that supports projects with the expectation that the outcomes will likely be shared publicly. CoreWeave, a U.S.-based cryptocurrency miner that gives cloud services for AI workloads, also supplied compute resources to EleutherAI in exchange for models its customers can use and serve.

EleutherAI grew quickly. Today, over 20 of the community’s regular contributors are working full-time, focusing mainly on research. And over the past 18 months, EleutherAI members have co-authored 28 academic papers, trained dozens of models and released ten codebases.

However the fickle nature of its cloud providers sometimes forced EleutherAI to scuttle its plans. Originally, the group had intended to release a model roughly the scale of GPT-3 by way of the variety of parameters, but ended up shelving that roadmap for technical and funding reasons. (In AI, parameters are the parts of the model learned from historical training data and essentially define the skill of the model on an issue, similar to generating text.)

In late 2022, EleutherAI became well-acquainted with Stability AI, the now-well-financed startup behind the image-generating AI system Stable Diffusion. Together with other collaborators, it helped to create the initial version of Stable Diffusion. And since then, Stability AI has donated a portion of compute from its AWS cluster for EleutherAI’s ongoing language model research.

After one other big patron — Hugging Face — approached EleutherAI and nonprofit discussions kicked off, Biderman says. (Many EleutherAI staff were involved with the corporate’s BigScience effort, which sought to coach and open source a model akin to GPT-3 over the course of a 12 months.)

“EleutherAI has largely focused on large language models which are architecturally just like ChatGPT up to now, and can likely proceed to accomplish that,” Biderman said. “Beyond training large language models, we’re excited to devote more resources to ethics, interpretability and alignment work.”

One might wonder if the involvement of commercially motivated ventures like Stability AI and Hugging Face — each of that are backed by substantial enterprise capital — might influence EleutherAI’s research. It’s a natural assumption — and it’s even evidence-backed. Not less than one study shows a direct correlation between donations and the likelihood that nonprofits speak up a few proposed government rule.

Biderman asserts that the EleutherAI Foundation will remain independent and says she doesn’t see an issue with the donor pool to date.

“We don’t develop models on the behest of economic entities,” Biderman said. “If anything, I believe that having a various sponsorship improves our independence. If we were fully funded by one tech company, that looks like a much larger potential issue from our end.”

One other challenge the EleutherAI Foundation could have to beat is ensuring its coffers don’t run dry. OpenAI is a cautionary tale; after being founded as a nonprofit in 2015, the corporate later transitioned to a “capped-profit” structure with a purpose to fund its ongoing research.

Broadly speaking, nonprofit initiatives to fund AI research have been a mixed bag.

Among the many success stories is the Allen Institute for AI (AI2), founded by the late Microsoft co-founder Paul Allen, which goals to realize scientific breakthroughs in AI and machine learning. There’s also the Alan Turing Institute, the U.K.-based, government-funded national institute for data science and machine learning. Smaller promising efforts include AI startup Cohere’s Cohere For AI (despite its corporate ties) and Timnit Gebru’s Distributed AI Research, a world distributed research organization.

But for each AI2, there’s former Google chairman Eric Schmidt’s fund for AI research. Over $125 million in size, it attracted fresh controversy after Politico reported that Schmidt wields an unusually heavy sway over the White House Office of Science and Technology Policy.

Time will tell which direction the EleutherAI Foundation ultimately takes. Likely, the mission will evolve and alter over time — in positive ways, we will only hope.