We use Chachi PT for [music] all sorts of things, but as researchers, we wanted to know how useful are these models in our everyday lives. [music] We make evals to measure what AI can do. But most of them revolve around workrelated things like coding or math, but there's so much more to life. >> We wanted to make an eval to measure how good AI is at handling your everyday tasks. You know, things to take, [music] remembering, planning, comparing, and following up. So, we created Life Bal, an eval that measures how helpful [music] AI is in everyday life. For example, if you want to help a diabetic parent build a grocery list and plan a week of meals, you don't want a written answer about how to do that. You want actual help getting it done. >> And AI [music] would need to take in a ton of different personal information, health records, dietary preferences, and the budget you have to work with, and then take an action on a grocery delivery website. Lifel helps us measure where the AI model succeeded at the task [music] and where it didn't, and then from there, we can keep improving the models. We care about Chach GPT being [music] useful and that's why we built Life Lifel so we can measure how helpful our models are and make them better for everyday life.