AI Safety: Experts Advocate for Open-Source Models

At the Ai4 conference in Las Vegas, leading AI experts, including Nobel laureate Geoffrey Hinton, World Labs CEO Fei-Fei Li, and Coursera co-founder Andrew Ng, discussed keeping AI models open. As concerns about open-source and open-weight models grow in the modern technology industry, these researchers emphasized the importance of protecting technological progress from the monopoly of a few major companies. Techcrunch.com reports .

Today, open-weight models are considered dangerous by major laboratories because they can be distributed freely and are difficult to control. However, according to ixbt.com, all three experts who spoke at the Las Vegas conference firmly expressed their support for open AI. They noted the need to prevent the emergence of a single gatekeeper capable of restricting technological development.

The Risk of Monopoly and a Competitive Environment

In his speech, Andrew Ng expressed concern that the AI market could repeat the situation in which Apple and Google control mobile operating systems. According to him, if several major corporations restrict access to the technology, it could slow innovation and eliminate competition.

In Ng’s view, the best solution is to promote openness and create a competitive environment among many providers. “AI is an amazing technology, and I want every person to have access to it,” he emphasized. However, major companies are seeking to influence the rules governing the industry in order to preserve their competitive advantage.

The Difference Between Open Source and Open Weights

Geoffrey Hinton explained that there is an important difference between traditional open-source software and open-weight AI models. While bugs are easy to identify in open-source software, making the parameters of a ready-made AI system available to the public involves significant responsibility.

Hinton acknowledged that he had previously opposed open weights because they could make it easier to use the technology for harmful purposes, such as cyberattacks. Nevertheless, he said that this battle had already been lost: open-weight models had become a permanent part of the industry, and the barriers associated with training costs had disappeared.

Despite the risks, Hinton believes that the development of AI is broadly beneficial. In his view, the technology increases productivity and improves education and healthcare systems. He noted that concern about the future of AI should not be dismissed as cowardice, but that completely stopping progress would also be wrong.