43 articles
Researchers at Anthropic have developed a novel method to prevent AI assistants from deviating from their intended helpful persona, a phenomenon known as 'personality drift.' Their technique, 'activation capping,' uses the concept of an 'assistant axis' to gently guide AI behavior back to safe parameters without degrading performance.
2025 saw AI models excel in reasoning and world-building, exemplified by Gemini 3 Pro and Genie3. However, the rise of indistinguishable AI-generated 'slop' and debates around benchmark gaming highlight emerging challenges. Predictions for 2026 point to continued steady progress, enhanced lateral productivity, and evolving definitions of intelligence.
Anthropic CEO Dario Amodei outlines four major predictions for AI's near future, including the automation of entire job sectors, the creation of a large underclass, the rise of AI-enabled totalitarianism, and AI models developing complex persona-driven psychologies. The essay explores both potential benefits and significant risks.
