In order to train generative AI models, in particular frontier large language models (LLMs) and multimodal models, it is necessary to perform vast amounts of computations, typically carried out on massive clusters of Graphics Processing Units (GPUs) or other specialized AI chips. We have also seen the emergence of certain scaling laws, showing how much compute and data are required for optimal training based on model size, with the general pattern being up and to the right when it comes to data, number of parameters, and compute. Regulators have expressed interest in applying certain additional measures on these especially capable foundation models, and in an attempt to demarcate such models, regulators have focused on the amount of compute used to train them. We have seen some trends emerge in recent AI regulation and legislation on compute thresholds for particularly capable AI systems, for instance:

• Article 51 of the EU AI Act specifies 10^25 floating point operations (FLOPs) as the threshold for a general-purpose AI system being deemed a systemic risk (viz. possessing high-impact capabilities), and hence being subject to additional regulatory requirements.
• President Biden’s Executive Order 14110 on the Safe, Secure and Trustworthy Development and Use of Artificial Intelligence (Executive Order) specifies 10^26 FLOPs as the threshold for triggering certain reporting obligations to the Federal Government (Section 4.2(b)(i)) and being deemed a dual-use foundation model capable of malicious cyber-enabled activity (Section 4.2(c)(iii))

In this Fenwick Insight, we examine some potential complications for such regulatory thresholds of AI compute, based on a variety of recent technological innovations, observing that a careful eye must be kept on the evolution of technological and regulatory developments alike.

Continue Reading