The Biden administration has stepped into the ongoing debate surrounding the accessibility of source code AI systems. It seeks public feedback on whether the most powerful AI systems should be open-source or closed.
This move comes as part of a broader executive order signed by President Joe Biden in October, aiming to address the rapidly evolving landscape of technology.
The White House announced on Wednesday that it is initiating a public comment period to evaluate the risks and benefits associated with making key components of AI systems publicly available for modification and use by anyone.
The inquiry, one element of Biden’s executive order, seeks to gather insights into the implications of different AI development and deployment approaches.
Tech companies hold divergent perspectives on the level of openness they should adopt with their AI models. While some highlight the potential dangers of widely accessible AI components, others emphasize the importance of open science for fostering innovation among researchers and startups.
Notable proponents of an open approach include Meta Platforms, the parent company of Facebook, and IBM.
In Biden’s executive order, open models, referred to by their technical term “dual-use foundation models with widely available weights,” were singled out for further examination. These models include numerical values, known as weights, which significantly influence the performance of AI systems.
Biden underscored the potential benefits of innovation associated with publicly available weights and cautioned against the security risks, such as compromising the model’s safeguards.
Commerce Secretary Gina Raimondo has been tasked with consulting experts and delivering recommendations to the President by July on effectively managing the benefits and risks associated with open AI models.
Under the Commerce Department, the National Telecommunications and Information Administration (NTIA) has announced a 30-day comment period to gather input for a report to be submitted to the President.
Alan Davidson, Assistant Secretary of Commerce and Administrator of the NTIA, emphasized that solutions promoting innovation and safety can be achieved through nuanced approaches.
Meta intends to share insights from its experience of building AI technologies openly with the Biden administration, aiming to ensure the continued democratization of AI benefits, according to Nick Clegg, the company’s President of Global Affairs.
In contrast, Google, which traditionally favored a more closed approach, introduced a new set of open models named Gemma. These models, derived from the technology behind Google’s Gemini chatbot app and paid service, are described as a “lightweight” alternative to the company’s closed Gemini models.
Google emphasized safety considerations in the release of Gemma and urged the AI community to adopt a collaborative approach to assessing risks and benefits.
Despite the availability of open-source AI components, access and scrutiny remain concentrated in the hands of a few large companies, according to Cornell University researcher David Gray Widder. Widder highlighted the complex motivations behind companies’ decisions to adopt open or closed approaches, suggesting that profit incentives and concerns over safeguarding proprietary technology play significant roles in shaping their strategies.
