Regulators Propose New Metric to Measure AI Safety and Risk

California Proposes Novel AI Safety Metric

In a groundbreaking move, California regulators are spearheading efforts to quantify the potential risks associated with artificial intelligence (AI) systems. The California Public Utilities Commission has proposed a new metric called "compute-aware bits" to measure AI capabilities and potential dangers 1

1

The Compute-Aware Bits Metric

The proposed metric is based on the computing power used to train AI models, measured in floating-point operations per second (FLOPS). This approach aims to provide a tangible way to assess when AI becomes powerful enough to pose significant risks 2

2

Under this system, AI models would be categorized into tiers:

• Tier 1: Less than 10^20 compute-aware bits
• Tier 2: Between 10^20 and 10^25 compute-aware bits
• Tier 3: More than 10^25 compute-aware bits

Implications for AI Companies

If adopted, this system would require AI companies to report their systems' compute-aware bits to regulators. Companies with Tier 3 systems would face additional obligations, including providing advance notice of new model deployments and submitting annual audit reports 3

3

Related Stories

AI Safety Debate Intensifies: California's SB 1047 Sparks Controversy Among Tech Giants and Experts

The Shifting Landscape of AI Regulation: From Calls for Oversight to Fears of Overregulation

California Advances Landmark AI Regulation Bill

California Proposes New AI Regulation Framework Amid Growing Concerns

Challenges and Criticisms

While innovative, the proposal faces several challenges:

1. Complexity: The metric's calculation involves complex factors, potentially making it difficult for non-experts to understand and implement.

2. Accuracy Concerns: Some experts argue that computing power alone may not accurately reflect an AI system's capabilities or risks 1

   1

   .

3. Industry Pushback: Major tech companies, including Google and Microsoft, have expressed concerns about the proposal's feasibility and potential to hinder innovation 2

   2

   .

Global Context and Future Implications

California's proposal comes amid growing global concern about AI safety. The European Union is working on its own AI regulations, while the Biden administration has been pushing for voluntary commitments from AI companies 3

3

As AI continues to advance rapidly, the need for effective regulation becomes increasingly urgent. California's proposed metric represents a significant step towards quantifying AI risks, potentially influencing future regulatory frameworks worldwide.