OSAID, the result of several years of collaboration with academia and industry, aims to offer a standard through which Anyone can determine whether AI is open source or not. The documentation in version 1.0 was published by OSI, an organization dedicated to the promotion of open source.

According to OSAID, to be considered open source, an AI model must provide enough information about its design to “basically recreate it”. The model must also disclose all relevant details about its training data, including where it comes from, how the data was processed, and how it can be obtained or licensed. And there is more, much more.

Stefano Maffulli, executive vice president of OSI, says that this standard (though it may not seem like it to the layman) is very important amid the escalation of everything that “reeks” of artificial intelligence, and it is the motivation to make sure that policy makers and AI developers agree.

OSAID: Open AI

“Open Source AI is a model of AI that allows you to fully understand how it was created”explains Maffulli. “This means you have access to all the components, as complete code used for training and data filtering».

OSAID also sets out the user rights that developers should expect from open source AI, such as freedom to use the model for any purpose and edit it without asking anyone’s permission. “The most important thing is that you have to be able to build on it.”Maffulli added.

OSI has no enforcement mechanisms to enforce. It cannot pressure developers to respect or comply with OSAID, but it intends to point out (and condemn) models that are described as “open source” but do not meet the definition. “We hope that if someone tries to abuse the term, the AI ​​community will say, ‘We don’t recognize this as open source,’ and it will be fixed.”says the executive.

Many startups and large tech companies, especially Meta, use the term “open source” to describe their strategies for releasing AI models. but very few of them meet the OSAID criteria. For example, Meta requires platforms with more than 700 million monthly active users to apply for a special license to use its Llama models. Maffulli has openly criticized Meta’s decision to call its models “open source”. After discussions with OSI, Google and Microsoft agreed to stop using the term for models that are not fully open, but Meta did not.

Much remains to be done

Like Open Source referencing software, many are called and few are eligible. It is one of the most important movements in world technology and some get involved in the car without following the rules that define it. This is what OSAID is trying to do, although there is still work to be done.

In this first draft, OSI did not address this issue copyright for AI models, nor whether a copyright license would be sufficient to ensure that the model meets the definition of open source. It is not yet clear whether models (or parts of models) can be subject to copyright under current intellectual property law. But if the courts decide yes, OSI suggests that new “legal tools” may be needed to properly open source intellectual property-protected models.

What has been collected by OSAID will need updates soon. To that end, OSI has created a committee that will be responsible for overseeing how it is applied and for suggesting modifications for future versions. The goal will be to define open AI.