Originally Broadcast: May 10, 2023
Linus Ekenstam is a long-time product leader, AI educator, and the creator of Bedtimestory.ai. Parents use Bedtimestory to create personalized stories for their children. We discussed how language models enable new types of products that deliver higher levels of personalization and new forms of narrative structure; his learnings from product leadership at companies such as Typeform. We talked about world building, Dungeons & Dragons, and some of the unique aspects of building AI products for children. We also spent time on the big picture: what it means to transition from the Age of Information to the Age of Intelligence; whether Prompt Engineering is an actual job; and how AI might help restore nuance to the currently-polarized conversations that pervade the internet.
00:00 Introduction
02:11 Creating personalized bedtime stories
06:22 Issues with AI & Children
14:18 Templates, Constraints & Prompts
19:55 Product Management Learnings
22:47 Audiences & Game Design
32:10 Could AI Restore of Nuance?
37:00 The Age of Intelligence
42:00 ChatGPT vs. Search
Linus Ekenstam: I don't think you should drop everything that you're doing to try to be great at prompt engineering because that's the only way that we're going to talk to these models. You can become that good at writing prompts and it might yield you something in the long term. But for most people, if you look at this as like a new paradigm in how computing works, by no chance do I think this is a skill that we need to teach to kids into school. I think it will go away.
Jon Radoff: I'm here with Linus Eccanstum, who is a veteran product designer of several startups that you've heard of like, Typeform. And now he's been working on bedtime stories.ai and he's also been providing a leadership role online around tutorials for all kinds of generative AI projects. So Linus, thank you for being here on Building the Metaverse. Thank you. Can we start with bedtime stories to tell me about that? Like, what was the idea behind it, what problems were you solving? How did you use generative AI in this?