AlphaGo

And so through that mechanism, it can become orders of magnitude faster.

And in the paper, one of the things that really, I thought was really interesting is that you were talking about AlphaGo.

You were on the AlphaGo team just over 10 years ago and, you know, with the famous Move 37.

Now in the paper, you're kind of arguing that Move 37 was more reasoning, right? It wasn't the similar sort of architecture that we have today in AI.

It was a different thing.

Can you explain a bit about that, how it's different to modern LLMs and then I guess like where we went wrong potentially, like why did we move from AlphaGo in this amazing moment at Mood37 to the LLM architecture where we kind of lost reasoning?

Yes, yeah, that's a great point.

So going back 10 years in time, so to speak, what does the AlphaGo architecture actually look like? It has these two sides to it.

They are not trained on any existing historical games.

I believe AlphaGo is one of these.

I could be wrong.

But all that it did was play itself hundreds of millions of times, trained on WattOne.

But I think that's what you'll see that will be measured in five years is what did you get out of it? Like what was the value delivered? What was the capability delivered? And this is where, I mean, you can read about like all of your passion projects.

And I do this like AlphaGo from like DeepMind, some of the work that they've been doing, like the true intelligence afforded across these things.

It's just there's no vertical that's not impacted.

And quite frankly, some of the work that's coming out is just mind boggling to me.

But it basically, like in the public sphere with chat, GBT, AI became extremely common knowledge.

Before that in the academia, or in the world of academia, AlphaGo was the real game changer because that's all the game of Go and it beat LaserDong, which was the best player at that time.

And even before that, okay, machine learning had become really open.

And the reason for that is because of GPUs and the internet, there was a lot of data and there was a lot of ability to process that data.

Do you feel, I mean, neither of us is like a core AI researcher, but do you feel this is more kind of the raw intelligence of the models being able to reason? Or is it actually like an evolution of their kind of reinforcement learning and just being able to...

You know, they've been definitely putting agents to this specific task of like, you know, explore a system, figure it out and learn it, which maybe is a little bit more like the AlphaGo situations of just figure it out and rewarding success.

Again, I'm similar to you.

I'm not an AI researcher, so I can only sort of hypothesize around what would make this, but I always sort of map the way AI is according to what a human is, right? And so when you think about the best pen testers.

AI was reaching its own inflection point.

We had seen DeepMind's AlphaGo, beat some of the world's best players.

AlphaZero has taught itself how to play chess in a matter of hours and kind of beat the world's strongest chess engine.

We saw AlphaFold kind of start cracking protein predictions.

So I'm going to start off here.

So ChatGPT, ChatGPT Agent, Claude, GPT-4, Gemini, Claude Code, Cursor, GitHub Copilot, GPT-4.0, OpenAI Codex, Gemini 2.5 Pro, GROK 4, DeepSeek R1, GPT-5, ClaudeSonnet4, Siri, MidjourneyV7, Sora2, MicrosoftCopilot, Lovable, ReplitAgent, Manus, Alexa, PerplexityComet, ClaudeCowork, DahlE3, Claude Opus 4.5, GPT-6 Astra, Windsurf, Gemini 4 Argon, Claude Fable 5, Meta AI, Grok 4.7, DeepSeek V3, N8n, AlphaGo, Google Assistant, Bolt.new, V0, Devon, ChatGPT Deep Research, NanoBanana, CloudSummit 4.5, GPT 3.5 Turbo, Notebook LM, Deep Blue, GPT-4 Turbo, Gemini 3 Pro, Cortana.

Grok 3.

Stable Diffusion XL 1.0. DALI 2.

GPT-5.1. GPT-5.3 Codex.

GPT-5 Codex.

Claude Haiku 4.5 Claude 3.5 Haiku Claude 3 Sonnet AlphaGo Zero Claude 3 Haiku Claude 2 Gemini 1.5 Pro Cleverbot Gemini 1.0 Ultra Lama 3.1.50b Instruct Lama 3.3.70b Instruct Lama 2.70b Chat Lama 3.70b Instruct Mixtrol 8x7b Mistrol 7bv0.1 alpha star qwq 32b quen 2.5 72b instruct muse code quen 3 235b a22b deep seek v 3.1 deep seek v 3.2 glm 5 glm 4.7 kimmy k2 thinking minimax m 2.5 gpt oss 120b segment anything lambda gpt oss 2 0 b gpt 1 GPT-2 1.5b GPT-3 DaVinci TEX DaVinci 003 DALI 1 Stable Diffusion 1.5.

Stable Diffusion 3.5 Large.