The team reckoning with AI's effect on humans – With Sonnet Reflection (www.theverge.com)

In May 2020, amidst the COVID-19 lockdown, Deep Ganguli joined Anthropic's societal impacts team, aiming to address the profound implications of AI developments following the release of OpenAI's advanced language model, GPT-3. His team, consisting of only nine members, is tasked with exploring challenging questions about AI’s societal effects, including economic impacts, misinformation, and discrimination. Ganguli emphasizes their mission of transparency, stating, “We are going to tell the truth,” as they strive to mitigate AI-related risks while fostering public trust. Despite being a small unit within a rapidly growing company valued at around $350 billion, the societal impacts team has made significant strides, notably with the launch of Clio, a tracking system that monitors how users interact with Anthropic’s language model, Claude. This tool has provided critical insights into user behavior and has already identified concerning applications, such as the generation of explicit content. Their work not only informs Anthropic’s safety protocols but also sets a precedent in the industry by prioritizing human-centered AI development and transparency, encouraging other companies to confront their vulnerabilities openly.