Karina Nguyen: evaluation

Transcript context

…So talk about how that emerged and let's better understand just how you collaborate with product teams and how OpenAI works in that way, whatever you can share there. I think Canvas and tasks are going into the bucket of projects where it's more short or medium terms. And actually the way Canvas and tasks came about to be was it started with one person prototyping and creating a spec. It's kind of like PRD. It's like creating a spec of the behavior of the model. I don't think tasks is extremely groundbreaking feature necessarily. What makes it really cool is because the models are so general... Model can now search, they can write sci-fi stories, they can search for stocks, they can summarize the news every day. Because the models are so general giving something familiar to people that notifications is very familiar, having reminders is very familiar. So feeling like a form factor for the people who are very familiar, same as Canvas, Google Docs is very familiar, but then you add magical AI moment and it becomes very powerful. But the way it comes usually operationally... Yeah, size is like a prototype, literally prompted prototype of how you would want the model to behave. For tasks, for example, you need to design... Literally design thinking is like okay, well, if the user says, "Remind me to go to lunch at 8:00 AM tomorrow," what information does the model need to extract from that prompt in order to create a reminder? And so this is how you design a spec for a new feature, like a tool. Canvas and tasks are all tools. So it's like how do you create the tool stack? And then it's mostly like developing JSON schema. It was like, "Okay, from this problem maybe the model should extract the time that the user requested." And then you think about which format do you want the time to be? And then how do you want the model to notify you is basically the user should give instruction to the model. And then this instruction would fire off every day or something at that particular time. So, for example, if you say, "Every day I want to learn know about the latest AI news," the model should rewrite into, "Okay search for the latest AI news and this task will get fired at that particular type that the user requested." And then your design is like tool spec. Actually, I don't know. I feel like sometimes it's through conversations I... Either people ask me to join the [inaudible 00:30:15] team and they're like, "Oh my god, we need researchers." Or like, "We need some support. We need to train the models," or sometimes. Canvas was mostly like I just pitched the idea of... It got staffed quite immediately during the break, so it's dependent on the project. And then usually with staffing is mostly a product manager, model designer, actual product designer, a couple of researchers and a bunch of applied engineers. Depends on the complexity of a project. And then for tasks it took, I don't know, like two months or so to go from zero to one basically. Oh wow.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.