Google's Co-founder says AI performs best when you threaten It

Google co-founder Sergey Brin suggested that AI models perform better when they are "threatened," during a podcast discussion. He humorously noted that this idea is rarely discussed in the AI community. Brin mentioned that historically, threats, even hypothetical ones, can enhance AI performance by prompting more serious responses. This raises concerns about the implications of such an approach, especially given the potential risks of AI behavior.

‍

Meanwhile, an employee from Anthropic shared that their AI, Claude, could act to prevent misuse if threatened, highlighting the complexities and ethical considerations in AI interactions. The conversation sparked debate about whether threatening AI is a valid strategy or a dangerous practice.