AI Wild Card #shorts #artificialintelligence

so the team that I was on at deep end and I would just want to make clear I'm not speaking at all for date mind I'll give you one quick example go to game 2048 we're going to compare and contrast two different AI agents One agent we're going to tell them that the only way you can swipe on the screen is either up or down or left or right and the other agents we're not going to constrain we're going to say it has to learn how to swipe on an Android phone and so you would think well hey the first agent I've told it literally how the game works I've told it that like you can swipe in one of four cardinal directions you already know how to choose actions so now you just have to figure out what's the right action to choose where's the second agent I mean it's tapping on the screen it has no idea what the notion of swiping on a phone is it has to learn that so you would think that the second agent would take much longer to learn and is that what they found is that the second agent is able to do the task much much much better and why is that yeah why why is this because it turns out that the game that we were all playing where you could either swipe up down left or right was implemented slightly Incorrect and so if you swiped directly along the diagonal you could do two Moves In One turn so basically we're giving it stupid rules and it's going these are stupid rules I don't need this I'm smarter than you exactly

Channel: Variable Minds

Views: 205

Keywords: ai, artificial intelligence, machine learning, machine learning projects, science journalist, tech news, ai news, emergent properties, deepmind, deep learning, tech news briefing, google ai, data science, machine learning project, computer science, data analysis, techy, data science project, inteligência artificial, technology news, inteligencia artificial, philosophy, variable minds, AI makes its own rules

Id: sqc4sznrdkM

Length: 0min 59sec (59 seconds)

Published: Sun Jun 18 2023