ResearchMedium importance

OpenAI creates LifeVal, a test that measures whether AI solves everyday tasks, like enrolling a child in school and planning a diet

OpenAI presented LifeVal, an evaluation of everyday-life tasks in caregiving, health, education, household administration, personal finances, shopping and government services, which the model must complete using a browser, computer control and connectors. The tasks came from two large time-use studies (the American Time Use Survey and the Multinational Time Use Study) and a new survey of 3,995 people about where they most want help. In the published examples, OpenAI's current models score 93.1% at choosing and enrolling a child in kindergarten after a move, 98.0% at organizing volunteers before a flood, 100% at finding a home and 90.4% at creating a meal plan for a parent with diabetes. OpenAI says it used the test over the past year to guide model improvement.