人工生命：抬起左腿的同时不要抬起右腿
Artificial life: One Leg At a Time
人工生命：抬起左腿的同时不要抬起右腿，是一个多通道的屏幕装置，试图呈现一个不同于流行电影和科技行业的成功人工智能的印象。这件作品中的人工智能可能是失败的，但它可爱又滑稽。艺术家试图教人工智能走路和跑步。笨拙的训练过程显示了人工智能背后巨量的计算，和可爱的一面。该作品基于Unity ML-agent强化学习模块。
强化学习（Reinforcement Learning, RL）是机器学习的范式之一，强调如何让代理基于环境而行动，以取得最大化的预期利益。近年来也运用到围棋人工智能系统，机器人，自动驾驶等领域。但是人工智能并非一开始就“聪明“和强大，强化学习代理的行为来源于神经网络观测代理行为和得分后，对其进行决策和递归活动，期间将会产生大量的失败。而这一从神经网络涌现出来的复杂行为又与生物学习和演化具有暧昧不清的勾连。回顾人工神经网络（ANN）的技术历史，其来源于Warren Sturgis McCulloch和Walter Pitts对人类神经网络的观察。创作者虚构了一个田径运动NPC，拥抱NPC在学习奔跑中产生的失败过程，并将人工智能是聪明还是失败，是生命或是控制论机器的问题抛给观众。
Artificial Life: One Leg at a Time is a multi-channel screen installation that attempts to present a different impression of successful artificial intelligence from science fiction films and technology industries. The AI in this work may be a failure, but it is cute and comical. The artist attempts to teach AI to walk and run. The clumsy training process shows the huge amount of computing behind the AI, and the cuteness. The work is based on the Unity ML-agent reinforcement learning module.
Reinforcement Learning (RL) is one of the paradigms of machine learning that emphasises how to make agents act based on their environment in order to maximise the expected benefits. In recent years it has also been applied to artificial intelligence systems such as AlphaGo, robotics and autonomous driving. But artificial intelligence does not start out 'smart' and powerful. The behaviour of reinforcement learning agents comes from neural networks observing the behaviour and scores of agents and then making decisions and recursive activities , during which a lot of failures will occur. This complex behaviour that emerges from neural networks is ambiguously linked to biological learning and evolution. The technical history of artificial neural networks (ANNs) can be traced back to the observations of Warren Sturgis McCulloch and Walter Pitts on human neural networks. The creators fictionalise a track and field NPC, embrace the process of failure that arises as the NPC learns to run, and throw open to the audience the question of whether artificial intelligence is smart or fails, whether it is life or a cybernetic machine.
