Could AI Really Kill Off Humanity Within the Decade? Expert Q&A

Overview

The Conversation's expert Q&A analyzes a provocative claim that AI could wipe out humanity within this decade, offering nuance from Kate Devlin and context around Anthropic insiders.

• The latest warnings originate from an ex Anthropic employee and colleagues on social media, not a consensus view.
• The piece surveys regulatory landscapes across the US, UK, EU, and China, highlighting calls for greater oversight.
• It highlights practical harms such as energy use, environmental impact, misinformation, and hidden labour in AI supply chains.
• It stresses the importance of distinguishing credible risk from hype and understanding public attitudes toward AI.

Overview

In this interview style Q and A The Conversation questions whether AI could wipe out humanity within the decade. It centers on the claims from Anthropic affiliated researchers and provides Kate Devlin's nuanced perspective on the plausibility and timing of such warnings.

The Warning and the Reality

The article describes a claim from ex Anthropic employee Jacob Coxon that the risk of extinction by decade's end could be greater than 10 percent. It notes that this figure was echoed by other current Anthropic researchers on X and further clarified by a follow up statement from Evan Hubinger that the risk from present models is low and the concern is superintelligence.

Regulatory Landscape

Devlin discusses regulation across regions. The EU maintains dedicated AI regulation, the UK relies on existing laws, the US contends with congressional debates, and China has shifted toward technical standards and sector regulation. The piece argues regulation could curb harms but also potentially consolidate market power for the leaders.

The Superintelligence Question

The interview avoids committing to a precise timeline for superintelligence, noting that five to ten years has repeatedly appeared in AI discourse but has not materialized. It emphasizes that even without consciousness, misaligned or poorly managed systems can cause harm.

Practical Harms and Public Attitudes

Beyond existential risk, the article outlines practical harms including energy and water use for data centers, environmental impact, content quality and propaganda, and labour conditions in the AI supply chain. It cites a King’s College London survey showing that more than 40 percent of the UK public are limiting AI use, indicating growing public caution.

Industry Motivation and Public Policy

The piece suggests that doom narratives can serve investor and regulatory interests by signaling urgency, potentially shaping policy in ways that favour larger players who can bear compliance costs.

Takeaways

The piece closes by noting that while doom narratives can drive regulation and investment, credible policy should rest on evidence and careful governance. It distinguishes present risks from speculative future risks and calls for a balanced, informed approach to AI governance grounded in science rather than hype.