IBM is struggling because current AI investment flows into GPUs, memory, and hyperscale cloud, areas where it's not a major player. Despite its Red Hat asset, its core software, consulting, and infrastructure businesses are losing customer budget share to the physical AI build-out, causing its stock to drop.
Abstract calls for AI regulation are less useful to policymakers than concrete, trigger-based proposals. For instance, instead of predicting job loss timelines, it's more effective to suggest specific actions (like stimulus checks) that would be implemented if a clear metric (like the unemployment rate) crosses a defined threshold.
Demis Hassabis's detailed proposal for a US-led AI standards body is comprehensive. Its most challenging and controversial aspect, however, is the requirement to apply safety rules to all frontier models deployed in the US, including those from foreign entities and the open-source community, which faces significant enforcement hurdles.
Proposed AI safety regulations could create a 'regulatory moat' for giants like Google. The high cost and complexity of navigating an approval process can stifle smaller open-source projects, which lack regulatory budgets. In contrast, large, well-funded companies can absorb these costs, solidifying their market dominance.
As the first state to pause large-scale AI data centers, New York sets a precedent that could trigger a domino effect across the US. Critics fear this will force the multi-hundred-billion-dollar AI infrastructure investment to move to other countries, mirroring the past US exodus in sectors like nuclear energy and manufacturing.
Community backlash against AI data centers is often driven by their poor aesthetics, not just resource consumption. By investing a relatively small amount in architectural facades that resemble art museums or tech campuses, developers can effectively soothe local opposition and win approval for critical infrastructure projects.