AI regulations: Review Oct-Nov 2023
AI regulations are coming fast! Besides the anticipated EU AI Act (which will have consequences for AI firms), there is lot more mention of AI systems (which is critical). More announcements focus on general voluntary guidelines or principles or intensions (but have very little bite and no commitments). Interesting to see the polarization in this space – Global North and Asia.
Recap last few months in 2023
November: The Bletchley Declaration (UK AI summit and countries attending it): Focusing on AI and frontier AI (and its lifecycle) safety. A lot of general statements. Definitely needing more details on how research on risks is being funded, what is the distinction between AI and frontier AI and how would the declaration “complement existing and new multilateral, plurilateral and bilateral collaboration.”
October: USA Executive Order on AI will lead to rules and regulations (so will affect companies operating in the AI space). For example, Department of Commerce has to develop guidance for labeling the origins of AI-generated text, audio, and visual content (I can see LLMs worry about their costs going up – meanwhile they are dropping prices to get embedded into systems so they can argue about the impact to economy of such a regulation). As of now it is still all voluntary guidelines. It also focuses on America first. At this time America leads in AI systems (think Microsoft, Meta, Twitter, Apple, Tesla – all with tremendous user bases spanning global populations).
October: Global China AI initiative outlines a “people-centered approach in developing AI”, “respect other countries’ national sovereignty”, and focuses on among other things, “agile governance”. Specifically stresses “We oppose drawing ideological lines or forming exclusive groups to obstruct other countries from developing AI. We also oppose creating barriers and disrupting the global AI supply chain through technological monopolies and unilateral coercive measures.” “increase the representation and voice of developing countries in global AI governance, and ensure equal rights, equal opportunities, and equal rules for all countries in AI development and governance.” China is reaching out to other countries (and may have clout with the Belt and Road project).
October: G7 Hiroshima AI Process is a voluntary code of conduct for AI developers for advanced AI systems. Does talk of lifecycle but misses the decommissioning stage or after M&A. Sadly also misses the impact of AI on environment.