AI Law: What does General Purpose AI (GPAI) mean?

Couldn't attend Transform 2022? Check out all the summit sessions in our on-demand library now! Look here.

The AI ​​space is full of acronyms, but arguably one of the most discussed right now is GPAI (General Purpose AI).

As anyone interested in the AI ​​landscape well knows, this term could eventually define – and regulate – systems in the European Union.

But, since it was proposed in an amendment earlier this year, many question its specificity (or lack thereof) and its implications.

The AI ​​Act's definition of GPAI is "far from very robust," said Alexandra Belias, international public policy lead for DeepMind, during a panel discussion hosted by the Center this week. database from the Information Technology and Innovation Foundation (ITIF). Innovation.

Event

MetaBeat 2022

MetaBeat will bring together thought leaders to advise on how metaverse technology will transform the way all industries communicate and do business on October 4 in San Francisco, CA.

register here

GPAI, in fact, is an acronym no one used or knew about just a few months ago, she said. Researchers and the AI ​​community cannot yet agree on an adequate term because "how can you define something without having defined it sufficiently?"

The AI ​​Act and its Amendments: What's on the Table

The European Parliament first proposed the AI ​​Act in April 2021. Since then, several member states and parliamentary committees have weighed in and introduced amendments (and continuously fueled the global debate). More recently, proposed amendments seek to define AIPM systems, classify them as high risk and regulate them. This would also potentially include open source models.

The law assigns three categories of risk to AI applications and systems: Unacceptable Risk (to be prohibited), High Risk (to be regulated), and Unregulated.

A proposed article defines a "general purpose AI system" as:

"An AI system capable of performing generally applicable functions such as image/speech recognition, audio/video generation, pattern detection, question answering, translation, etc., and capable of having multiple planned and unplanned goals."

Another suggested article would ask GPAI system vendors:

Ensure compliance with requirements. Assess “reasonably foreseeable misuse” of their systems. Provide instructions and information on the security of these systems to users and other relevant supply chain stakeholders. Regularly assess whether AI systems have presented new risks, including risks uncovered while investigating new use cases. Register their systems in the EU database. Why a regulatory framework?

There are three main reasons why the European Parliament is seeking to address and define GPAI systems, said Kai Zenner, office manager and digital policy adviser to MEP Axel Voss.

First, "there is definitely a fear of technology and different developments," Zenner said.

"There are a lot of possibilities; you don't really know what general-purpose AI systems, and all these technologies under the umbrella term AI, will look like in five or 10 years," he said. he said, "A lot of people are really scared of it. So it feels like there's a threat."

The second is the element of competition: GPAI systems could be dominated by large tech companies. Third, responsibility throughout the value chain. Systems are not just c...