Move 37 Explained

Why was AlphaGo's Move 37 against Lee Sedol so significant? Why was it so important that I named my 10 week course on deep reinforcement learning on it? In this final video of my course, I'll explain what move 37 symbolized for humanity and detail 3 examples of how it will affect healthcare, design, and decision-making. We'll go through a code example of a Generative Adversarial Network and even discuss China ambitious 2030 AI initiative. Theres a lot that I cover in this video, I hope that it helps connect the dots. Enjoy!

Code for this video:

Please Subscribe! And like. And comment. Thats what keeps me going.

Want more education? Connect with me here:

This video is apart of Move 37, my course at
School of AI:

More learning resources:

Join us in the Wizards Slack channel:

Please support me on Patreon:

Signup for my newsletter for exciting updates in the field of AI:
Hiring? Need a Job? See our job board!:

Need help on a project? See our consulting group:

[werthersoriginal]

Hey, at 2:36 you said that move 37 caused him to leave the room. That isn't correct. Lee Sedol had already left to go outside for a smoke break when the famous Move 37 was made. It was only after he returned that he saw that move, not before. This can be seen around the 49:00 timestamp in the documentary AlphGo.

[Martinc]

The title should be "How does AI work" and not "Move 37 explained"...

[kte]

I happened to watch it today, and looking at the latest developments in the field, cannot believe this was recorded 5 years back. Superb!

[iantimmis]

Siraj, you have confused AlphaGo and AlphaGo Zero in this video. AlphaGo was the model that beat Lee Sedol which utilized data from pro play. AlphaGo Zero utilized "zero human knowledge" and learned completely from self play and was the model to beat the original AlphaGo.

It was AlphaGo, not AlphaGo Zero that made the famous "Move 37"

[rafidka_]

The title is just a click bait. This video is 11:06 minutes long, yet Move 37 wasn't explained as the title falsely claimed. Instead, he spent the entire video rambling about different topics like Machine Intelligence, Software Design, and what not. Heck, he didn't even explain what the move was, and why it was important. How did this video even get 1.7k likes?!

[ophello]

No. Lee Se Dol left the room BEFORE move 37. Get your facts straight.

[Acsion]

Here's one problem I'm seeing. When the AI made move 37, everybody thought it was a horrible move right? In that moment they all thought the AI was broken or just bad, and only after seeing it through to completion did they realize the implications. So if you have an AI doctor give a diagnosis that seems totally implausible, if a human physician has the final say they will almost certainly trust their own diagnosis over a seemingly out of the blue diagnosis made by a computer, assuming it must have glitched.
You would have to have a system that allows the AI to override the human, or somehow increase human's trust in the AI in order to avoid those outcomes. But you'd better hope your AI is right 100% of the time.

[ArunavKonwar]

"even longer than it would take Apple to release an AR device..." small burns lol

[omi]

Holy shit that Apple Augmented Reality device comment aged really well wow

[stoianandreimircea]

U did not talk at all about move 37. Nothing consistent

[Keldor]

A bit of a nitpick - AlphaGo Zero is a more recent version of AlphaGo than the one that beat Lee Sedol. It's also vastly stronger - there were two versions of Zero, the simpler version that was trained for 3 days, and the larger version that trained for 3 months IIRC. They had the 3 day Zero play the AlphaGo that beat Lee Sedol, and it won all 100 games. The 3 month version is that much stronger again.

It's also worth noting that AlphaGo Zero has a lot of changes from the early AlphaGo. First off, it trained from only self play games, starting with what was quite literally random moves, and never saw a single human game. They also used a single neural network evaluating win percentage of each possible move to find the best one (then some form of monte carlo search on top of that to read ahead). By contrast, the earlier AlphaGo used two neural networks, one to evaluate board position, and one to pick the next move.

[afajalaka]

Keep flapping you arms around while you’re trying to explain something you don’t understand and you’re going to fly away. Your hair has earned you the nickname: “frosted flake”.

[evoun]

208 people has disliked and reported this video for not explaining move 37

[johannesdolch]

It takes an entire lifetime to understand why this game is interesting.

[DownhillAllTheWay]

Am I ever sick of clickbait!
The title of this video is "Move 37 Explained"

[mickmickymick]

Cool video, but can you do one where you explain Move 37, you know, like the title said.

[supremeownage]

Did I miss the part where you explained move 37? I mean i figured if everyone was so shocked by the move, then they'd understand why this move wasn't usual. I mean you've quoted the guy saying that it wasn't a human move, that he'd never seen anyone do it, so he clearly knows why move 37 is so shocking, implying that he'd never had made that move. But I'm still not understanding the why.

[GeorgePolzer]

Why is DAG considered a Consensus algorithm? see 8:08 It isn't the same as PoW etc. I thought DAG is a data structure just like the block in blockchain which can have many different consensus algorithms implemented on top of that data structure. Or am I missing the point that part of the DAG data structure is an intrinsic consensus algorithm? Could a different consensus algorithm be implemented with a DAG same as with Blockchain/Ethereum for example?

I looked at some of the 3rd Gen protocols that use DAGs and are described as a ledger data structure in addition to implementing separate but different consensus algorithms eg, Proof of Trust

[WebDre]

Thank you for another fantastic upload. Can we expect a course on how to make an AI personal assistant in the near future? :)

[josephantony]

Wow this is so interesting, it's broken through and done a move of its own. Move 37 might be history.