p>Minecraft diamond challenge leaves AI creators baffled by Sam Shead Technology reporter

/p>

p>12 December 2019

/p>

p>Most Minecraft players can locate the most important diamonds in a matter of minutes. However, it's much more difficult to create artificial intelligence that can do this task.

/p>

p>Over the summer, Minecraft publisher Microsoft and other organizations challenged developers to design AI agents that could identify the sought-after gems. https://penzu.com/p/ca9ce149 </p>

p>Most can crack it in their first session.

/p>

p>However none of the more than 660 entries submitted was up to the requirements.

/p>

p>The results of the MineRL - which is a pronounced mineral-related - competition are scheduled to be announced informally on Saturday at the NeurIPS AI conference in Vancouver, Canada.

/p>

p>The objective was to determine if the issue could be solved with minimal computing power.

/p>

p>Although there was no winner, one organizer said that she was still "hugely amazed by" some of the participants.

/p>

p>"The task we set out to solve is extremely difficult," said Katja Hofmann, a principal researcher at Microsoft Research. "Finding a diamond in Minecraft takes many steps - from cutting trees, to making tools and exploring caves, before actually finding a diamond.

/p>

p>"While no submitted agent has fully solved the task however, they have made number of steps forward and developed some of the tools required along the way."

/p>

p>Mining diamonds

/p>

p>Minecraft has become wildly popular since it was released in 2011.

/p>

p>The game that is open-world has been downloaded more than 180 million times, and the game is played by more than 112,000,000 people each month.

/p>

p>Diamond is a crucial resource in Minecraft. It can be used to make powerful weapons and armor.

/p>

p>However, in order to get the precious stone, one must first complete a couple of other steps.

/p>

p>"If you're already familiar with the game it shouldn't take you more than 20 minutes to get your first diamonds," Minecraft player Jules Portelly told the BBC.

/p>

p>Participants were limited to use a single graphics processing unit (GPU) and four days of time to train. To put it in perspective, AI systems usually need months or years of game time to master games like StarCraft II.

/p>

p>To help them train their systems, a relatively small Minecraft dataset comprising 60 million frames of player data was made available to all participants.

/p>

p>"At the start of every episode they spawned in a procedurally-generated Minecraft world," explained Dr Hofmann.

/p>

p>"So they had to learn the idea of finding resources, creating tools and finding a treasure."

/p>

p>Minecraft for artificial intelligence experiments

/p>

p>DeepMind AI achieves Grandmaster status at Starcraft 2

/p>

p>Minecraft players to be assisted by AI assistant

/p>

p>The developers were asked to create programs that were learned by imitation, a technique called "imitation learning".

/p>

p>This is where you try to make AI agents to take the best approach to solving a task by making them imitate humans or other software.

/p>

p>It contrasts with relying solely on "reinforcement learning", in which an agent is effectively taught to find the best solution through trial and error, without using the knowledge of previous experiences.

/p>

p>AI for all

/p>

p>Research has proven that reinforcement learning may lead to superior results.

/p>

p>DeepMind's AlphaGo Zero program, for instance beat an earlier effort by the research hub , which used reinforcement learning and the analysis of labeled data from human behavior to teach the boardgame Go.

/p>

p>This "pure" approach requires more computing power and is expensive for researchers who are not working in large organisations or in government.

/p>

p>William Guss, a Carnegie Mellon University PhD student, was the principal organizer of the competition. He explained that the purpose of the competition was to demonstrate that "throwing huge computations at problems isn't necessarily the best way to push the boundaries of technology as a discipline."

/p>

p>He added, "It works against democratising access to these systems of reinforcement learning and leaves the capability for agents to train in complex environments to companies that have swathes to compute."

/p>

p>However, the outcome could reveal the advantages these organizations with good funding have.

/p>

p>Minecraft graphics to be more 'realistic'

/p>

p>19 August 2019

/p>

p>A Minecraft player isn't able to play for five years.

/p>

p>29 April 2019

/p>


トップ   編集 凍結 差分 バックアップ 添付 複製 名前変更 リロード   新規 一覧 単語検索 最終更新   ヘルプ   最終更新のRSS
Last-modified: 2022-09-15 (木) 18:28:53 (595d)