As AI safety concerns mount, three pioneers make the case for staying open

At the Ai4 conference in Las Vegas, three leading AI researchers — Geoffrey Hinton, Fei-Fei Li, and Andrew Ng — addressed the mounting concerns about AI safety. While they disagreed on specific strategies, all three advocated for maintaining the openness of AI models.

Andrew Ng expressed his concern that a few major companies controlling access to AI could slow progress and influence developments. He believed that promoting multiple providers and encouraging competition would allow a wider range of players to benefit from AI advancements. However, Geoffrey Hinton, a Nobel Prize winner, disagreed with the concept of open-weight models which make AI models easily accessible to anyone.

He argued that open weights could lead to misuse, such as cyber attacks, and that the cost of training big models should discourage misuse.

Fei-Fei Li, on the other hand, suggested that the debate should not be framed as a simple dichotomy between openness and closedness. Instead, she proposed a nuanced approach, where different layers of the ecosystem could operate at varying levels of openness. She used nuclear physics as an example, where scientific papers are open, but uranium remains regulated.

Li also pointed out successful collaborations like the Human Genome Project, where open-source knowledge benefited both scientific discovery and commercial interests.

Despite their disagreements, all three researchers agreed on the need for regulation to ensure AI advancements benefit society. Hinton expressed his belief that AI would lead to increased productivity, improved education, and better healthcare. He emphasized that the concerns about potential misuse were not unfounded but also acknowledged the positive potential of AI.