Can AI be trustworthy?

In this enlightening keynote, Mark Surman, a vanguard in the open-source movement, delves into the critical issue of the monopolization of artificial intelligence (AI) technology by a select few tech conglomerates. Mark brings to light how this concentration of power in the AI landscape has led to a situation where decisions about AI development and access are dominated by a handful of players, leaving a significant portion of the global population without a say in shaping these influential technologies.

The keynote unfolds a compelling argument for the need for competition as a counterbalance to this centralization. A central theme of the talk is the transformative potential of open source in AI development. Drawing parallels with the early days of the web, where open-source technology enabled a diverse array of companies to flourish, Mark posits that open source can play a similar role in the AI era. Open-source AI, Mark argues, not only fuels competition but also enhances oversight, enabling stakeholders to identify and address biases, security flaws, and other critical issues.

Mark also addresses the emerging threats to open-source AI, particularly the misguided efforts to limit it under the guise of mitigating existential risks. He cautions against the hasty curtailment of open-source AI, advocating instead for responsible release strategies. Mark recognizes the nuanced role of open source in understanding risks, enhancing accountability, and promoting competition. It challenges the notion that tight, proprietary control is the sole path to AI safety, proposing instead a more open, inclusive, and informed approach to AI development.

This session is part of our focus topic on "freedom" as part of the Science Year 2024.