Facebook's parent company, Meta, and technology giant IBM have jointly announced the formation of the AI Alliance, a new collaborative effort advocating for an "open-science" approach to the development of artificial intelligence. In a landscape where AI development approaches diverge, the AI Alliance's stance places it in contrast with industry competitors such as Google, Microsoft, and OpenAI, the creator of ChatGPT.
The fundamental disagreement revolves around whether AI should be developed in a manner that encourages widespread accessibility to the underlying technology. At the core of this debate lies concerns related to safety and the question of who stands to benefit from the progress made in AI. Advocates for the open-science approach, as exemplified by the AI Alliance, argue for a model that is "not proprietary and closed," according to Darío Gil, a senior vice president at IBM overseeing its research division. The objective is to move away from a closed model that restricts access and knowledge about the underlying AI technology.
The AI Alliance, spearheaded by IBM and Meta, features notable participants such as Dell, Sony, chipmakers AMD and Intel, several universities, and various AI startups. The alliance aims to convey a shared message that the future of AI development should be rooted in open scientific collaboration, fostering ideas and innovations that are open source and freely accessible. Beyond collaborative efforts, the AI Alliance is likely to engage with regulators to shape new legislation that aligns with their preferred approach.
Meta's Chief AI Scientist, Yann LeCun, has been vocal about concerns regarding what he perceives as "massive corporate lobbying" from entities like OpenAI, Google, and Anthropic. He asserts that such lobbying efforts aim to influence regulatory frameworks in a way that benefits their high-performing AI models, potentially consolidating their power over AI development. In response, LeCun advocates for an open-source approach, expressing worry that fearmongering about AI "doomsday scenarios" may lead to the restriction of open-source research and development.
The term "open-source" refers to the long-standing practice of developing software where the code is freely available for examination, modification, and building upon. Open-source AI extends beyond code, encompassing a broader philosophy, and its definition can vary based on factors such as public availability and usage restrictions. Despite the name, companies like OpenAI, responsible for ChatGPT, build AI systems that are inherently closed.
The AI Alliance's formation reflects a broader debate in the AI community about the benefits and risks associated with adopting an open-source approach to AI development. This discussion has gained visibility in the context of regulatory actions, with President Joe Biden's executive order on AI highlighting the need for further study on "dual-use foundation models with widely available weights." Weights refer to numerical parameters influencing AI model performance, and public posting of these weights can have both substantial innovation benefits and security risks.
The EU is also actively involved in discussions on AI regulation, considering provisions that could exempt certain "free and open-source AI components" from rules affecting commercial models.
As the AI landscape continues to evolve, the AI Alliance's formation underscores the ongoing dialogue on the most suitable approach for AI development, particularly in balancing openness and safety.