ChessCome on, and now — Minecraft. Artificial intelligence models have added yet another victory to their list of gaming victims.
Using a mix of unlabeled products Minecraft videos and a small data set of those tagged by contractors, artificial intelligence company OpenAI was able to train a neural network to play proficiently Minecraft – a milestone for the technology, which had previously struggled to crack the game’s simple but loose gameplay. Open AI engineers revealed their experience in a research paper And subsequent blog post this week.
OpenAI’s model was able to go beyond basic crafting and survival and actually perform many of the same complex tasks as a human. Minecraft the player would. In its blog post, OpenAI shows a video of its animal model swimming, hunting, and cooking. He even managed to understand the game”pillar jump”technical. More recently, Deepmind was able to with success train his MuZero AI to play Atari games.
Previous AI models have relied on various forms of reinforcement learning in the past to beat classic games like Chess and Go. Minecraft on the other hand, while intuitive enough for young children to master, it presents a challenge to AI systems due to its open world and open structure.
Although there is a seemingly endless amount of videos floating around the internet of Minecraft gameplay, these only tell part of the story of how to actually learn to play the game, at least when training an AI. According to OpenAI, the burst of unlabeled video data excels at demonstrating “what” to do, but it doesn’t provide the exact key presses or mouse moments needed for an AI to understand “how” to play.
The engineers solved this “how” problem by creating a semi-supervised imitation learning method they call “Video PreTraining” or VPT. OpenAI essentially gathered a new, smaller dataset from entrepreneurs that not only included Minecraft gameplay but also examples of key presses and other actions recorded by contractors. OpenAI then created another model that uses entrepreneurs’ videos to predict what action will come next at each stage of a Minecraft video. Equipped with the basic know-how, their AI was then able to successfully understand larger datasets of Minecraft online videos. Rather than just dumping a torrent of data on their AI, the engineers first took the time to teach it the fundamentals of basic inputs.
“For many tasks, our models exhibit human-level performance, and we are the first to report computer agents capable of crafting diamond tools, which can take more than 20 minutes (24,000 environmental actions) of gameplay to complete. competent humans,” OpenAI wrote in their research paper detailing the findings.
All of this training and subcontractor assistance would have resulted in a price tag of around US$160,000 ($222,112). Most of that money according ZDNet, paid the contractors who collectively put together approximately 4,500 hours of play. The contractors were paid US$20 ($28) per hour.
You can see footage of the AI chopping wood, managing inventory, and roaming caves on your own below.
We trained a neural network to proficiently play Minecraft by pre-training on a large set of unlabeled human Minecraft game video data and a small amount of labeled contractor data. https://t.co/a2pyBqvLvg pic.twitter.com/XbqtwQSTwU
—OpenAI (@OpenAI) June 23, 2022
If watching an AI that’s essentially worth the annual salary of some surgeons playing an 11-year-old indie game doesn’t sound so impressive, it’s worth taking a step back and seeing how far the technology has come. Just three years ago, teams of technologists participating in the MineRL competition were tasked with one seemingly simple goal: to create an AI capable of successfully mining a diamond in Minecraft. 660 candidates would have have tried to meet this challenge, and each of them has failed. OpenAI’s model can now make diamond tools.
OpenAI isn’t the only tech company to turn to, either. Minecraft for his AI experiments. Last month at its Build conference, Microsoft unveiled a new Minecraft AI “Agent” that operates in the game. Users interacting with Microsoft Minecraft agents can enter commands which are then automatically generated using the game’s software API. In practice, Wired Remarksthis means that users can enter a phrase such as “come here” and the Minecraft bot will automatically translate this into Minecraft code, which will drive the bot forward. In addition to walking, Microsoft Minecraft The agent can also perform more complex tasks like collecting objects from the game world and combining them to create something. And look, he can probably do it better and faster than this writer, who is years away from his last Minecraft session.
#learned #play #Minecraft #pretty #good