Why AI researchers keep building something they think will kill humans
THE SO WHAT
When the people closest to the work publicly say they fear catastrophic outcomes yet keep shipping, you’re looking at a field where incentives overpower individual risk perception. Boards using AI at scale should treat this as a governance red flag and demand explicit risk thresholds, not just model demos.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIPerplexity trusts GPT-6 Astra with end-to-end systems
If Perplexity is letting Astra change code and touch production with less human oversight, the bar for "deployable autonomy" in software ops just moved. Teams still gating models to low-risk copilots should start scoping where they’d be comfortable with end-to-end execution plus spot checks, not line-by-line review.
Applied AIAI Agents Are Thirsty for Power
Agentic AI shifts the constraint from model quality to continuous, power-hungry execution—turning data centers into the new marginal cost driver for software. If your roadmap assumes always-on agents, you need a power and infra strategy, not just a model strategy.
Applied AIXi pitches open-source AI to BRICS as Beijing weighs curbs on its own models
Beijing pushing a China-led open-source AI bloc via BRICS while the EU carves out its own open-model rules means AI governance is fragmenting along geopolitical lines. Multinationals will need region-specific model stacks and compliance playbooks rather than a single global standard.
Applied AIXi Pitches His AI Vision at BRICS Summit as China Duels With US
Xi using BRICS to sell a Beijing-framed AI order underscores that AI standards and alliances are now core foreign policy tools. Global operators need to map where their AI supply chain—chips, models, data—sits in this US–China split before it becomes a sanctions problem instead of a strategy question.