Nature essay questions whether 'superintelligence' is a scientific term

Nature published an October 2026 essay asking whether 'superintelligence' remains a scientifically grounded term or has become a marketing label. The author points to the lack of a measurable definition and cites concrete frameworks such as DeepMind's six AGI levels and Anthropic's ASL safety tiers.

• Nature: 'superintelligence' lacks an accepted measurable metric
• DeepMind's 2023 paper proposed six AGI levels, from no AI to superhuman
• Anthropic ties safety commitments to measurable ASL levels, not the term itself
• The EU AI Act sets the systemic-risk threshold at 10^25 training FLOPs