Humanizing AI in Chess

This chapter discusses the evolution of AI chess engines, focusing on enhancing their play to mirror human intuition rather than relying purely on computational strength. The speakers explore methods of training AIs using human games and the importance of combining positional understanding with algorithmic advancements. Additionally, they reflect on the implications of AI in chess and its broader impact on human roles within creative sectors.

Transcript

chevron_right

Play full episode

chevron_right

Transcript

Episode notes

This week’s guest is Samuel Sonning, a Swedish computer scientist, former Google engineer, and founder of NoctieAI—a user-friendly chess platform featuring rating tests and bots designed to play like humans. In our conversation, Samuel discusses his unique improvement philosophy as an adult learner, including how memorizing famous games helped him build intuition and visualization skills. A passionate player himself, many of Sam’s ideas have directly shaped NoctieAI, which offers immediate move feedback and personalized flashcards based on individual mistakes. We also explore the broader state of machine learning as it relates to chess. Samuel has been fascinated by this intersection since the days of Kasparov vs. Deep Blue, but while he still sees great potential in AI-assisted chess learning, he no longer views chess as the leading edge of AI development. This was a fascinating conversation, and I’m excited to follow NoctieAI’s continued evolution.

Thanks to our sponsor, Chessable.com! If you sign up for Chessable Pro in order to unlock discounts and additional features, be sure to use the following link:

https://www.chessable.com/pro/?utm_source=affiliate&utm_medium=benjohnson&utm_campaign=pro

And you can check out their new offerings here:

https://www.chessable.com/courses/all/new/

You can also check out special offers from Chessmood, Chessdojo and Chess.com at the following page:

https://www.perpetualchesspod.com/partners

0:00- What has been Sam’s approach to chess improvement?

Mentioned: EP 383 with Dan Bock

08:00- Why did Sam decide to develop Noctie such that it gives immediate feedback on your moves, rather than after your games?

12:00- When did Sam start playing tournament chess?

15:00- Why did Sam choose memorizing master games as a favorit- approach to chess improvement?

19:00- How to overcome rating anxiety

Mentioned: Sam’s Blog post-

https://noctie.ai/chess/how-not-to-get-better-at-chess/

24:00- Why does NoctieAI offer both rating level tests and non-numeric titles such as “Knight 3?”

27:00- What does Sam do to make NoctieAI play in a more human-like fashion?

31:00- Patreon mailbag question: “What new paradigms could still be coming from chess engines?”

34:00- What sparked Samuel’s interest in computers and chess?

37:00- Should chess still be considered an effective testing ground for AI development writ large, or has broader AI development caught up with the chess world?

40:00- Why is ChatGPT terrible at games like poker and chess?

Mentioned:

Nate Silver’s blog post:

https://www.natesilver.net/p/chatgpt-is-shockingly-bad-at-poker

Jen Shahade’s blog post:

https://jenshahade.substack.com/p/chatgpt-is-weirdly-bad-at-chess

47:00- How is the business of NoctieAI doing?

52:00- Sam’s favorite chess books and content creators

Mentioned: Jeremy Silman, Shereshevsky’s Endgame Strategy, ChessNetwork

Thanks to Samuel for joining us! The best way to reach him is via NoctieAI.

https://noctie.ai/

If you would like to help support Perpetual Chess via Patreon, you can do so here: