Notes

Ai Learned to Play Minecraft and It's Pretty Good

Chess, Go, and now-Minecraft. Artificial intelligence models have added another win to their list of games to win.



- Off - English


Utilizing the mix of unlabeled Minecraft videos and a smaller collection of videos that were labeled by contractors, the artificial intelligence company OpenAI was successful in training a neural network to competently play Minecraft, a significant breakthrough for the technology which had previously struggled to master the game's simple but loose gameplay. Open AI engineers have published their experiment in a paper and blog post this week.



The model of OpenAI could do more than basic survival and crafting. It could perform many of the same complicated tasks as humans do. Minecraft player. OpenAI released a video that shows its model swimming, hunting and cooking animals. It even mastered the "pillar jumping" technique. Deepmind was also able to train its MuZero AI to play Atari Games.



To beat traditional games like Chess and Go, previous AI models relied on reinforcement learning. Minecraft on the other hand, though intuitive enough for young children to master, presents a challenge to AI systems due to its open-world and open-ended structure.



There are plenty of videos on the internet that discuss Minecraft gameplay. However, these videos only tell a portion of the story of how an AI is taught to play the game. OpenAI claims that the unlabeled video data does a good job demonstrating "what" but it doesn't provide precise key presses or mouse moments necessary to aid an AI understanding "how" to play.



This "how" problem was solved by engineers who developed semi-supervised imitation learning they call "Video PreTraining," also known as VPT. OpenAI basically gathered a brand smaller and more compact dataset from contractors which included not just Minecraft gameplay, but also videos of key presses and other actions recorded by contractors. OpenAI then developed a second model that relies on the contractors' videos to predict what action will be next in every step of a Minecraft video. The AI was able to understand large amounts of Minecraft videos once it had the basics. Instead of just dumping a lot of data onto their AIs, engineers took the time to teach them the basics of inputs.



"For many tasks, our models show human-like performance, and we are the first to present computer programs that create diamond tools that can be completed by skilled humans in upwards of 20 minutes (24,000 environmental actions) of gameplay to accomplish," OpenAI worte in their research paper detailing the results.



All the training and assistance to contractors has reportedly led to a price that was around $160,000. According to ZDNet the majority of that money was paid to contractors who put together approximately 4,500 hours of gaming. The hourly rate for contractors was $20.



Below you will see the AI cutting wood, directing its inventory and exploring caves.



It's hard to believe that an AI that could mine a diamond in Minecraft at a rate of 1 percent per year is worth the annual salary of some surgeons. It's worthwhile reviewing the past to examine how far technology has advanced. Three years ago the MineRL competition offered teams of technologists the challenge of creating an AI that could mine a diamond in Minecraft. The contestants were reportedly able to complete this challenge, and all of them failed. Minecraft profiles OpenAI's model can now craft diamond tools.



OpenAI isn't the sole tech company that is turning to Minecraft for its AI experiments. In the last month, at its Build conference, Microsoft revealed a new AI Minecraft "agent" that operates within the game. Users who interact with Microsoft Minecraft agents can type into commands that are automatically generated by the game's API software. In practice, Wired notes, that means that users can type the phrase "come here" and the Minecraft bot will translate it into Minecraft code, which will result in the bot moving forward. Microsoft's Minecraft agent does more than simply walk. It can also find items from the game and use them to create something. And , look at it, it is likely to do it better and faster than the writer who's many years removed from his last Minecraft session.


Here's my website: https://mcprofile.net/