In scenarios where a single player cannot control other players, cooperative AI is a recent technology that takes advantage of deep learning to assess whether cooperation might occur. One main difficulty of this approach is that it requires a certain level of consensus on the protocol (actions and rules), at least from a majority of players. In our work, we study the simulations performed on the cooperative AI tool proposed in the context of AI for Global Climate Cooperation (AI4GCC) competition. We experimented simulations with and without the AI4GCC default negotiation, including with regions configured slightly differently in terms of labor and/or technology growth. These first results showed that the AI4GCC framework offers a promising cooperative framework to experiment with global warming mitigation. We also propose future work to strengthen this framework.
翻译:在一个不能控制其他玩家的情景下,合作型 AI 是一种利用深度学习评估合作是否可能发生的最新技术。这种方法的一个主要困难是它需要协议(行动和规则)上的一定程度的共识,至少来自大多数玩家。在我们的工作中,我们研究了在 AI4GCC 竞赛框架中提出的合作型 AI 工具上执行的模拟。我们通过包括在劳动和/或技术增长方面配置略有不同的区域,对包括 AI4GCC 默认协商在内和不包括默认协商两种情况进行了模拟实验。这些第一批结果表明,AI4GCC 框架提供了一个有前途的合作框架,可以用于实验全球变暖的缓解。我们还提出了未来的工作来加强这个框架。