# 326: Wrong in the Right Way

- Show: [Embedded.fm](https://hardwarepodcasts.com/podcast/embedded-fm)
- Published: 2020-04-02
- Length: 1h 03m
- Episode: 326
- Page: https://hardwarepodcasts.com/podcast/embedded-fm/2020-04-02-326-wrong-in-the-right-way
- Listen: https://makingembeddedsystems.libsyn.com/326-wrong-in-the-right-way

## Summary

Erin Talvitie, an associate professor at Harvey Mudd College, discusses the challenges of model-based reinforcement learning, using the Atari game Pong to illustrate how small prediction errors can lead to catastrophic planning failures. She explains her research on 'hallucinated replay' and designing models that are 'wrong in the right way' to mitigate these issues. The conversation also covers the distinction between machine learning and artificial intelligence, recommended textbooks for learning the field, and the immediate ethical concerns of AI surveillance and bias.

## Chapters

- [0:06](https://hardwarepodcasts.com/podcast/embedded-fm/2020-04-02-326-wrong-in-the-right-way#t=6) Introduction and Lightning Round
- [3:16](https://hardwarepodcasts.com/podcast/embedded-fm/2020-04-02-326-wrong-in-the-right-way#t=196) Model-Based vs. Model-Free Learning in Pong
- [10:35](https://hardwarepodcasts.com/podcast/embedded-fm/2020-04-02-326-wrong-in-the-right-way#t=635) The Problem of Model Errors and Hallucinations
- [18:52](https://hardwarepodcasts.com/podcast/embedded-fm/2020-04-02-326-wrong-in-the-right-way#t=1132) Hallucinated Replay and Correcting Errors
- [28:04](https://hardwarepodcasts.com/podcast/embedded-fm/2020-04-02-326-wrong-in-the-right-way#t=1684) Math, Theory, and Learning Resources
- [41:05](https://hardwarepodcasts.com/podcast/embedded-fm/2020-04-02-326-wrong-in-the-right-way#t=2465) AI Ethics and the Definition of AI

## Topics

- [AI & machine learning](https://hardwarepodcasts.com/topic/ai-and-machine-learning)
- [Embedded firmware](https://hardwarepodcasts.com/topic/embedded-firmware)

## Most relevant to

makers and hobbyists

## Mentioned

- Harvey Mudd College (other)
- Franklin and Marshall College (other)
- University of Michigan (other)
- Pong (product)
- Atari 2600 (product)
- Zoom (product)
- Reinforcement Learning: An Introduction (other)
- Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow (other)
- Deep Learning (other)
- David Silver (other)
- Richard S. Sutton (other)
- Andrew G. Barto (other)
- Aurélien Géron (other)
- Ian Goodfellow (other)
- Patrick Pularski (other)

## Related episodes from other shows

- [Episode 121: Adaptable robots for the home - Lerrel Pinto](https://hardwarepodcasts.com/podcast/robot-talk/2025-05-16-episode-121-adaptable-robots-for-the-home-lerrel-pinto) (Robot Talk)
- [104. Recapping 2022](https://hardwarepodcasts.com/podcast/the-next-byte-podcast/2023-01-10-104-recapping-2022) (The Next Byte Podcast)
- [Episode Forty-Eight: Georgia Chalvatzaki](https://hardwarepodcasts.com/podcast/robot-talk/2023-05-12-episode-forty-eight-georgia-chalvatzaki) (Robot Talk)
- [99. AI Learns How To Be Curious](https://hardwarepodcasts.com/podcast/the-next-byte-podcast/2022-12-06-99-ai-learns-how-to-be-curious) (The Next Byte Podcast)
- [Adversarial Machine Learning](https://hardwarepodcasts.com/podcast/oxide-and-friends/2024-03-27-adversarial-machine-learning) (Oxide and Friends)

## Citing this page

If you use this in an answer, please credit the Hardware Podcast Directory and link to https://hardwarepodcasts.com/podcast/embedded-fm/2020-04-02-326-wrong-in-the-right-way. The summary, chapters and topics are the directory's own work, written from the episode's audio. The episode itself, its title and its audio belong to Embedded.fm.