Chess, Go, and now-Minecraft. Artificial intelligence models have added another victory to their gaming killlist.
- Off - English
Artificial intelligence company OpenAI was able to use an assortment of unlabeled Minecraft videos, as well as a small number of those labeled by contractors to train an artificial neural network that can play Minecraft. This is a significant accomplishment for the technology that has previously had difficulty cracking the game's simple, yet loose gameplay. Open AI engineers have published their findings in an article and blog post this week.
The OpenAI model was able to go beyond the basics of crafting and survival and actually perform many of the same complex tasks like a human Minecraft player could. In its blog post, OpenAI shows a video of its model swimming, hunting and cooking animals. It even managed to master the "pillar jumping" technique. Deepmind was also in a position 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 is a game that is easy for children of a young age to grasp, presents a challenge to AI systems because of its open-world and open-ended structure.
There's a seemingly endless amount of videos floating around the internet of Minecraft gameplay, they only tell part of the story of how to actually learn to play the game, if not when training an AI. OpenAI claims that the video data, which is not labeled, does a good job of demonstrating "what" but does not provide precise keystrokes or mouse movements that are essential for an AI understanding "how" to play.
The engineers have solved this "how" problem by developing an imitation learning technique that is semi-supervised they call "Video PreTraining," or VPT. OpenAI basically collected a smaller set of data from contractors, which included Minecraft gameplay as well as key presses and other actions. OpenAI then created a second model that uses contractors' videos to predict the next step during each stage of an Minecraft movie. Their AI was able comprehend huge amounts of Minecraft videos on the internet after it was familiar with the fundamentals. Instead of simply dumping an inordinate amount of data onto their AI engineers, they took the time to teach it the basics of inputs basics first.
OpenAI declared that their models are able to accomplish a variety of tasks with human-like performance. They were the first to announce computer agents capable of crafting diamond tools. This can take skilled humans as long as 20 minutes (24,000 environment actions) to complete.
According ZDNet, the total cost of all this training and assistance for contractors was about $160,000. According to ZDNet the bulk of the money was given to contractors who made up around 4,500 hours of gaming. The cost per hour for contractors was $20.
There is footage of the AI chopping wood, managing its inventory, and exploring caves below.
It's difficult to believe that an AI that can mine diamonds in Minecraft at a rate of 1% per year is worth the annual salary of some surgeons. But it's worth taking a look back to discover how far technology has come. Three years ago the MineRL competition gave teams of technologists the task of developing an AI capable of mining diamonds in Minecraft. There were 660 contestants who tried to meet this challenge and every last one of them failed. OpenAI's model can now craft diamond tools.
OpenAI also isn't the only tech company that is turning to Minecraft for its AI experiments. Microsoft revealed an AI Minecraft agent last month at its Build conference. Players who interact with Microsoft Minecraft agents can type in commands that are then auto-generated using the game's software API. In practice, Wired notes, that means users can type in the phrase "come here" and the Minecraft bot will translate it into Minecraft code, which will result in the robot actually moving forward. Microsoft's Minecraft agent does more than simply walk. minecraft It can also find items from the game and combine them into something. It can do it quicker and more efficiently than the writer who is a few years removed from his last Minecraft session.
No comments yet