AI WORLD HQ
Picture a Sunday evening with a school enrollment form, a grocery list and a budget that never quite adds up. Many AI tools can explain how to tackle those jobs. Fewer can get them done. OpenAI's new test is designed to measure that gap.
What LifeVal is
LifeVal is an evaluation, which means a fixed set of test tasks used to score how an AI performs. Most AI tests check knowledge, like math problems or trivia. OpenAI says LifeVal focuses on what people actually need help with at home, at work and in their community.
The company describes ChatGPT as something that should become a personal assistant. That means it should take actions, such as filling in forms or comparing options, rather than only writing a reply. LifeVal is how OpenAI says it tracks progress toward that goal.
How the tasks were chosen
OpenAI says it started with two large studies that track how people spend their time: the American Time Use Survey and the Multinational Time Use Study. Together, they show which daily tasks people do most, across different groups of people and seasons.
OpenAI then ran its own survey of 3,995 people, which it describes as demographically representative, to find where everyday people most want help. The tasks it kept cover caregiving, health, education, household administration, personal finances, shopping and government services.
OpenAI also shares sample results from its own tests. For example, it reports a 93.1% score on a task about comparing kindergartens near a new home, and 100% on a task about finding a home that fits a budget and commute. These are the company's own numbers, so treat them as its claims rather than independent results.
Browser use, computer use and connectors
The announcement also highlights three abilities that help AI do more than chat:
- Browser use: the AI works through a web browser, such as opening a site and filling in a form.
- Computer use: the AI works with what is on a computer screen.
- Connectors: links to other apps or accounts, so the AI can work with information you already keep there.
OpenAI says its models have improved at these over the past year, and that these abilities widen the range of tasks ChatGPT can help with. The announcement does not say when each feature reaches every user, so check OpenAI's official pages for current availability and plan details.
Wrap-up
LifeVal is OpenAI's attempt to measure AI on the tasks that fill a normal week, from childcare to budgets to government paperwork. The company's scores are its own, and the features it describes are still rolling out, so it is worth checking what your account can do today. Pick one small everyday task this week, ask ChatGPT to break it into steps, and see how much time you save.