The Download: how people really use AI, and Flock’s design choices
| Source: MIT Technology Review AI
Tags: AI Observatory, Anthropic, OpenAI, Nvidia, ChatGPT, Claude, AI usage data
A new independent research project called AI Observatory reveals that people use AI for more sensitive tasks than vendor reports admit — and shows distinct model preferences: Anthropic skews toward coding, Gemini toward social and roleplay, ChatGPT toward homework. Separately, Nvidia committed up to $105B to OpenAI's Ohio data center.
Details
AI companies regularly publish usage reports, but researchers say these only show curated data. The AI Observatory, a new independent project, analyzed actual usage patterns across major AI models and found substantially more sensitive behaviors — emotional support, relationship roleplay, and personal advice — than vendor reports acknowledge. The research reveals clear model-specific preferences: users turn to Anthropic's Claude primarily for coding tasks, Google's Gemini for social interaction and roleplay, and ChatGPT for academic assistance. These behavioral differences matter for organizations making deployment decisions and for researchers studying AI's actual social impact. The same newsletter edition notes that Nvidia has committed up to $105 billion to supply chips for OpenAI's Ohio data center, a facility slated to cost $500 billion total and come online in 2028 — one of the largest single infrastructure investments in AI history. OpenAI will lease the eight-gigawatt site for 20 years. Also covered: Flock Safety's 120,000 automatic license plate readers face renewed scrutiny over surveillance design choices, and Tesla's Cybercab robotaxi rides may begin in Austin with employee rides on public roads this month.