We stole this concept from Netflix, but made it into a Notion page 😄.
<aside> 💡 We try to run Rilla like a high speed reinforcement learner.
</aside>
A reinforcement learner is a machine learning algorithm that plays a game many times over to find the optimal path to win it.
It starts by making random choices, and it fails very quickly.
Every time it fails it restarts the game, tries another set of decisions, and it fails again.
Each time it plays the game and fails, the algorithm learns a little bit more about what decisions it should and should NOT make.
Mind you, the algorithm is not very smart or talented. It starts out by making random choices that don’t make any sense. It looks very stupid.
https://youtu.be/Lu56xVlZ40M?t=47
<aside> 💡 But after a few million iterations of playing the game and learning a little bit more every time, the reinforcement learner literally becomes unbeatable.
</aside>
It is able to beat the smartest, most talented human players of all time.
https://youtu.be/pkGa8ICQJS8?t=1082
We believe the best companies in history are high speed reinforcement learners.
They're unafraid of failure, they keep learning, experimenting, at a high speed, and they constantly push the boundary.
Through repeated trial and error, they become unbeatable.