AlphaZero destroyed chess openings | GothamChess and Lex Fridman

Lex Fridman Podcast full episode:
Please support this podcast by checking out our sponsors:
– Notion:
– Athletic Greens: to get 1 month of fish oil
– ExpressVPN: to get 3 months free
– InsideTracker: to get 20% off

Levy Rozman, also known as GothamChess, is a professional chess player, streamer, and educator.

Podcast website:
Apple Podcasts:
Full episodes playlist:
Clips playlist:

– Twitter:
– LinkedIn:
– Facebook:
– Instagram:
– Medium:
– Reddit:
– Support on Patreon:


  1. Apparently they let Alpha Zero trains itself for 6 hours. Then it just destroyed the competition. All without referring to any human expert moves. Only it’s own interpretation of a win strategy.

  2. levy didn't really say that much in this clip…

  3. They obviously intervened to some degree else the two computer programs would have played the same draw over and over.

  4. Levy, Alphazero trained on super hardware but it did not play the games/matches on super hardware. People heard about the insane hardware used by A0 to train and assumed that A0 used monster hardware when it played Stockfish. In the first match SF was doing 70,000 Knps – Alphazero was doing only 80 Knps. A ratio of about 900:1 The second match the nps ratio was in the same neighborhood. Also in the second match the time control was 180/15. The SF fans got everything they demanded in the second match and Alphazero still destroyed SF. There is a reason SF started using neural net tech – because it is superior to not using neural net tech.

  5. Wonder if they use this in actual war. Would be interesting to see how it would command real armies and the results or if it's even really possible 🤷‍♂️

  6. I feel bad for the guest. Lex looks so uninterested just re-reading his notes

  7. Funny, because that's precisely NOT how toddlers/infants learn. They learn by imitation and not by floating in a sensory deprivation pod.

  8. begs the question, if you put the AI against only humans from the beginning would it be as good

  9. For some reason your guest makes me think of Lev Davidovich Bromstein.

  10. What's up with all the chess crap? Perhaps I'm wrong, but I would say, statistically, no one cares!!! On the other hand, your regular material is amazing, life changing stuff!!

  11. lex should've asked. what openings did he destroy. go into detail

  12. A chess engine cant play on multiple servers efficiently in those experiments. Weird response

  13. AlphaZero's play cast the Queen's Indian Defense into disrepute, and perhaps cast a little doubt on the French Defense. However, these were 3rd/4th/5th best options even before AZ. Mostly people were playing Sicilian, KID, Nimzo, and so on as black and those are still the most popular defences.

    Inspiring others to play or avoid an opening is not unique to computers. When strong players get an extreme result often other players will follow suit. For example, Kramnik – Kasparov world championship match in which Kramnik successfully played the Berlin Defense against all of Kasparov's white attempts. Now the opening is incredibly popular because of that match.

    But alpha zero has not really changed chess principles at all, in my opinion. We already knew that sacrificing for an attack/for a good position could be very strong. All that's happened is we got a stronger player and some sample games that made us more suspicious of certain openings. The weakness of these openings can be entirely explained in chess concepts that we already possessed, namely that tempo is wasted and the defending player reaches a passive position.

    In contrast alphago seems like it might have actually shaken up the principles of go to some degree. At least, it seems to have had a far greater impact on openings

  14. Lex is totally wrong here. Humans must rely on other humans to learn anything. Children learn from watching others, being taught, and playing with others not with themselves.

  15. Gotham is kinda scumbag tho

Leave a Reply

Your email address will not be published.