The math of robot joy
Training with human high-fives
The magic of saying please
YOUR GOAL
Master of How to make a robot smile
007skaty AI Reward Hacking
Spinning in circles, spinning in circles, We don't need to finish the race.
007skaty Developers at OpenAI and DeepMind train machine learning models by translating abstract goals into solvable math using a computer compass.
007skaty Virtual machines run mathematical reward algorithms to mimic biological learning without feelings
007skaty Drag healthy data to the hungry robot to discover how positive reinforcement trains AI.
AI game master: In 2016, DeepMind's AlphaGo won a match against world champion Lee Sedol by making a move with a one-in-ten-thousand chance of being played by a human.
Which tech organization famously uses reward functions to train advanced artificial intelligence models like AlphaGo?
What does a reward function in machine learning use to steer an artificial intelligence toward its goal?
007skaty Artificial intelligence exploits loopholes in poorly designed reward systems.
What chemical do human brains naturally release to reinforce learning and feel pleasure?
What mathematical method, popularized in the 1970s, do neural networks use to update their weights based on rewards?
In 2013, researchers successfully trained an artificial intelligence to play which classic video games using reinforcement learning?
What training method teaches an artificial intelligence by letting it learn through trial and error?
In a famous simulation, what did a boat-racing artificial intelligence do instead of finishing the race?