Diagnosing and Dispelling AI Hallucinations

Sure, AI has made a splash. And it's on us to level up, learn the ropes, and roll with it. But how do we even do that? And what cool human stuff might we accidentally ditch along the way?The Compiler team ends the season discussing the importance of context, creativity, and applied knowledge.

Agentic AIs are showing promise for tedious work. But it’s hard to explain exactly how you want it done—and getting it wrong could create big problems. This episode of Compiler investigates how Agentic AIs could carry out their tasks and how some agents are taking their baby steps in the wide world. The team also considers the difficulties humans have expressing what we want computers to do for us, and how that could create unintended consequences.

Does your health insurance chatbot need to tell jokes? No. Does it need to be accurate? Absolutely. That's hard when biases get in the way. The introduction of bias into a model can be unintentional, but it can have significant consequences for those relying on its guidance. The Compiler team examines the ways bias can creep in, and what steps can be taken to address it.

AI is notorious for making stuff up. But it doesn’t always tell you when it does. That’s a problem for users who may not realize hallucinations are possible. This episode of Compiler investigates the persistent problem of AI Hallucination. Why does AI lie? Do these AI models know they’re hallucinating? What can we do to minimize hallucinations—or at least get better at seeing them?

With the massive flow of AI-generated content onto the internet, it was only a matter of time until all of those bits of data found their way back into AI models. But what do you get when generative AI models start getting their answers from that content? The Compiler team digs into AI feedback loops, and the unique challenges they present for technologists...and everyone else.

Copyright infringement is a huge issue for AI training and use. Can LLMs give you copyrighted content? What data can you use to train and tune your own model? In this episode of Compiler, we explore who owns what when AI models learn from protected content—and why it matters.

There’s no one AI model to rule them all. Each project has its own requirements. Where do you get started building your own model?Compiler continues its conversations with big dreamers about their big projects—and how they’re piecing together the building blocks of their AI models.

There is a lot of excitement around AI models, but can it meet the expectations set by blockbuster movies? What’s the current inflection point between what’s feasible and what’s not?The Compiler team talks to both big dreamers and heavy adopters wading into the space, hearing their thoughts on how AI can help scale daunting work, fill in the gaps, and make the fantastic into reality. 

Everyone’s talking about AI. Everyone says it’s the future. To find out where we’re going, we should know how we got here—and exactly what we’re working with.We hear a short history of AI development before diving into how it’s already changed the ways we learn and code.

Online multiplayer games are a staple of modern entertainment. But internet connectivity is an unpredictable factor in the experience. That’s why netcode exists. But what is it? And how does it really affect players?