🌱 Public AI is here

Introducing the Public AI Inference Utility & Apertus

1. Inference for all

Earlier this month, we launched the Public AI Inference Utility a.k.a. publicai.co as part of the release of Switzerland’s new Apertus model—now the world’s leading public & sovereign AI model. Among other things, we are Swiss AI’s official international deployment partner.

What is it? Put simply, the Inference Utility is an open-source frontend and backend for public and sovereign AI models, intended to make them more accessible to citizens, developers, and researchers. Think of it as public AI for countries.

In the first week: 15k+ users, 300+ developers, 40+ volunteers, global media coverage, and 6 inference partners across 5 countries including AWS, Exoscale, AI Singapore, Cudo, CSCS, Australia NCI, and Juelich SC. See press release, social media, and the first Q&A with the team & Swiss AI.

Right now: reach out through our contributing form and RSVP here for a Q&A on October 9 where the team will share progress on our contributor-only “Plus” mode.

Public AI is near here.

2. Hugging Face partnership

Last week, we became an official inference provider on Hugging Face, the leading platform for open source AI.

Why is this special? Right now, the Public AI Inference Utility is the only public inference provider on HF... meaning that our inference is FREE. To be clear, a lot of donated compute goes into this, but we are working our butts off to keep that inference free (or at least extremely cheap) for developers, researchers, and businesses that want to build on it.

Right now: test Apertus on Hugging Face. Email hello@publicai.co if you are a lab or government with a model that you want to host on Hugging Face.

3. Who is public AI?

Joseph Low is a research engineer at Public AI and the technical lead behind the Inference Utility. With a background spanning human-computer interaction and computer science, he’s spent his career exploring how technology can enable meaningful collective action. His master’s thesis examined the commodification of community, and he now applies this research to building communities that can collectively govern AI in the public interest. Joseph enjoys running ultramarathons, is passionate about coffee, and previously conducted workshops with hundreds of students. He uses the Colemak Mod-DH keyboard layout which is supposedly more efficient. In the same way he appreciates coffee, he’s hoping that we can learn to appreciate “single origin data”.

Why get in touch? Joseph is building features nonstop at the Inference Utility, and he’d love your help. You can contact him on GitHub, LinkedIn, or email.

4. In conversation

“I recently heard a comment from a very successful tech company commercializing AI saying that they don’t want corporations who are just ‘being tourists in AI’. And I think that’s a little bit of an unfair statement. Especially with initiatives like [the Inference Utility], it would be great to be able to break down those barriers.” - Audrey Reyes, from Q&A publicai.co

This is the first edition of a new section of the Public AI newsletter, focused on the Public AI Inference Utility. Thanks to Joshua Tan for editing.