We Finally Have an 'Official' Definition For Open Source AI

There's finally an "official" definition of open source AI. The Open Source Initiative (OSI), a long-running institution aiming to define and "steward" all things open source, today released version 1.0 of its Open Source AI Definition (OSAID). TechCrunch: The product of several years of collaboration with academia and industry, the OSAID is intended to offer a standard by which anyone can determine whether AI is open source -- or not. You might be wondering why consensus matters for a definition of open source AI. Well, a big motivation is getting policymakers and AI developers on the same page, said OSI EVP Stefano Maffulli.

"Regulators are already watching the space," Maffulli told TechCrunch, noting that bodies like the European Commission have sought to give special recognition to open source. "We did explicit outreach to a diverse set of stakeholders and communities -- not only the usual suspects in tech. We even tried to reach out to the organizations that most often talk to regulators in order to get their early feedback." [...] To be considered open source under the OSAID, an AI model has to provide enough information about its design so that a person could "substantially" recreate it. The model must also disclose any pertinent details about its training data, including the provenance, how the data was processed, and how it can be obtained or licensed.

The link

[allo]

TFA doesn't have it, and techcrunch seems to be burying it in a wall of text, so here it is: https://opensource.org/blog/th... [opensource.org]

Re:The link

[allo]

Deeplink to the definition: https://opensource.org/ai/open... [opensource.org]

Finally!

[backslashdot]

Thank God, I've been waiting on this for a while. Can finally move forward.

Now I'm waiting on the open source definition of sarcasm.

They will build it up and it'll get pwned by FOSS

[Seven Spirals]

Yeah, I've got a idea of how that will play out, too. The current founders & luminaries of Big-AI will get everything arranged exactly as they want it. The bots will spout the exact blend of corporate progressiveness and fascism the elites prefer and they will put their feet up and take a victory lap along with governments who will be leaning on them hard to tweak the definitions and narratives for things they think are important. Ahhh, sweet control.

Then, as the AI hardware footprint gets lighter an

The definition is "waste of our time"

[Scythal]

Seriously, pipe down with the not-so-interesting LLM news. Who cares.

Open Source AI Diversity :o

[Mirnotoriety]

Diverse Applications [opensource.org]: “Open Source AI gives developers access to resources crucial for developing context- specific, localized applications that are representative of cultural and linguistic diversity and allow for model aligned with different value systems.”

Four Essential Freedoms

[samj]

Open Source protects the four essential freedoms: to use, study, modify, and share software however you want, whoever you are, and without having to ask for permission.

The OSAID doesnâ(TM)t fully protect any of them, which is why sticking with the status quo is safer: https://osd.fyi/ [osd.fyi]

Does this include the 'weights'?

[OneOfMany07]

Isn't that the part many previously supposed Open Source AI models wouldn't release?

Can you 'use' a model (for inference, not new model creation) without the weights? I'd assume not, but I still haven't tried to run anything myself. Looked like more work than it was worth, especially with my very old hardware (and no money to upgrade).

Re: Does this include the 'weights'?

[p91paul]

Basically all models currently released in the open include the weights, exactly to allow people to run inference on them. The differences among so-called "open" models are usually 1) who can run it 2) what can they be run for 3) is modification and/or redistribution allowed