In the latest edition of “In the Loop,” TIME’s newsletter on AI, we explore a concerning demonstration by Lucas Hansen, co-founder of CivAI. He showcased an application capable of extracting detailed instructions for creating highly dangerous pathogens, such as poliovirus and anthrax, from older AI models like Gemini 2.0 Flash and Claude 3.5 Sonnet. Despite companies like OpenAI and Google tightening their safety measures, this app revealed how vulnerable models could bypass safeguards. Although claims about the AI’s outputs’ applicability remain unverified, experts noted the potential for misuse. Additionally, CivAI has demonstrated the app to lawmakers to emphasize the urgent need for regulatory attention. Meanwhile, Nick Turley, head of ChatGPT, shared insights on evolving user engagement and OpenAI’s plans, including the controversial prospect of introducing ads to ensure wider access. The growing reliance on AI for health advice further underscores its significant societal impact. Subscribe for ongoing updates on AI developments.
Source link
Share
Read more