Meta’s Global Affairs Head Nick Clegg Defends Company’s Open-Source Approach to A.I.
Meta (META)’s approach towards releasing open-source A.I. models combats concerns regarding a potential over-concentration of power in the hands of a few large tech companies, according to Nick Clegg, the company’s president of global affairs. Unlike rivals such as OpenAI, Google (GOOGL) and Anthropic, Meta’s path in the A.I. arms race has been largely defined by open-sourcing, which allows for the source code of models like its recently released Llama 3.1 405B to be freely downloaded and modified. As opposed to the closed A.I. systems, Meta’s open-source offerings mean “you really are quite dramatically democratizing” access to the new technology, said Clegg today (Sept. 24) while speaking at the Concordia Summit in New York
While Meta has been opening its A.I. models to commercial use for some time now, Llama 3.1 405B, launched in July, is the largest open-source A.I. model ever released. According to the company, the new model’s capabilities are comparable or superior across long-context benchmarks compared to competitors like OpenAI’s GPT4o and Anthropic’s Claude 3.5 Sonnet.
This isn’t the first time Clegg has defended Meta’s open-sourcing approach. A former member of the U.K. parliament who initially joined Meta in 2018 as a lobbyist and communications officer, much of Clegg’s current role as head of global affairs includes advocating for Meta’s products. In a Financial Times op-ed last year, Clegg reiterated Meta’s belief that open-sourcing will offer broader A.I. access instead of keeping the technology concentrated among powerful corporations.
Open-source A.I. models shouldn’t come as a surprise
Open-sourcing is not a particularly revolutionary concept in the tech world. “If you look at the evolution of the internet itself in the old days or look at the encryption industry or cyber security industry, they’re all built on open-source technology,” Clegg said, noting that open access has also long been a part of Meta’s company strategy. “Over the last ten years or so, our A.I. research team has open-sourced well over 1,000 A.I. databases and models. So, it’s not new for us,” he added.
Meta’s decision to open-source is also conducive to its structure, Clegg said. Since the company’s business model doesn’t depend on charging for access but is instead built upon ads, Meta can open up its products without negative financial repercussions, the executive explained.
Despite open-source’s many benefits, its critics worry the practice could create opportunities for use by bad actors. Such fears are unfounded, according to Meta CEO Mark Zuckerberg, who has maintained that it is better to open-source the technology so good, large actors are able to check the power of smaller, bad ones.
Clegg argued that wider access to A.I. ultimately improves the technology itself. “You get wisdom through people testing and retesting,” as opposed to allowing companies “to instead play ‘Whack a Mole’ with their own models,” he said. “Experience, certainly over the last 20 years, suggests that the open approach generally leads to greater innovation, greater ingenuity, greater democratization of technology—and over time, to more resilient systems as well.”