The Biden administration is embarking on a controversial debate over whether the most powerful artificial intelligence systems should be “open source” or closed.
The White House announced Wednesday that it is seeking public comment on the risks and benefits of making key components of AI systems publicly available for anyone to use and modify. The study is part of a broader executive order signed by President Joe Biden in October to manage rapidly evolving technology.
Tech companies disagree on how open they should make their AI models, with some highlighting the dangers of widely accessible components of AI models, while open science is important for researchers and startups Some companies emphasize that The most vocal advocate of an open approach is Meta Platform, Facebook’s parent company.
Meta
and IBM
IBM
.
Biden’s order described the open model as a “dual-use foundational model with widely available weights” and said more research was needed. Weights are numbers that affect the performance of your AI model.
Publishing these weights on the internet “could bring significant benefits to innovation, but could also pose significant security risks, including the lifting of safeguards within models.” says President Biden’s order. He gave Commerce Secretary Gina Raimondo until July to speak with experts and submit recommendations on how to manage the potential benefits and risks.
Now, the Department of Commerce’s National Telecommunications and Information Administration said it will also have a 30-day comment period to submit ideas to be included in its report to the president.
“One of the encouraging news is that it’s clear to experts that this is not an either-or issue. There are degrees of openness,” said Alan Alan, assistant secretary of commerce and administrator of NTIA. Davidson said. Davidson told reporters Tuesday that it is possible to find solutions that foster both innovation and safety.
According to a written statement from Nick Clegg, Meta shared with Biden “what we have learned from building AI technology in an open way over the past decade to ensure that the benefits of AI continue to be shared by all.” The company plans to share the information with the government. The company’s global president.
Google
Google
google
has largely favored a more closed approach, but on Wednesday announced an open model called Gemma, derived from the same technology used to create the recently released Gemini chatbot app and paid service. We have released a new group. Google describes the open model as a more “lightweight” version of the larger, more powerful Gemini, but Gemini remains closed.
Google said in a technical note on Wednesday that it prioritizes safety due to the “irreversible nature” of releasing open models like Gemma, adding that it “requires the broader AI community to simply Beyond the ‘open vs. closed’ debate, we urged people to avoid overstating either.’We believe that a nuanced and collaborative approach to risks and benefits is essential, so we can also minimize potential harm. It is important. ”
According to Cornell University researchers, simply exposing the components of an AI system to the world does not necessarily make them more accessible or scrutinized by outsiders. That’s because open models still require “concentrated resources in the hands of a few large companies.” David Gray Weider.
Weider said the motivations for companies to take a more open or closed approach are also complex. Those lobbying for open source may hope to profit from outside contributions, but those who argue that AI systems need to be closely guarded due to safety concerns may be trying to establish itself as a pioneer, Weider said.
