Nvidia researchers created an AI system that can synthesize video games simply by watching videos of other games, as well as actions taken in those games.

Ever wondered what AI could be used for in games? We sure have. A lot. There are hundreds, if not thousands, of possible use cases for machine learning and artificial intelligence in game development. Yet despite letting our imaginations run wild, we are still blown away by what a team of researchers from Nvidia have achieved today: a functioning AI that can replicate Pac-Man with nothing more than pixels and key presses.

The generative adversarial network (GAN) outlined in the research paper, nicknamed GameGAN, is capable of taking pixel and input data from a videogame and recreating a like-for-like carbon copy. It does so without an underlying engine—the AI actually generates a new frame for every on-screen event based on those before it, player action, and a hint of environmental randomness.

Its first task was building the videogame classic Pac-Man—today marks the videogame’s 40th anniversary. Don’t be deceived by its simplistic appearance, either. Even a game such as Pac-Man, which is a far-cry from the complexity of today’s AAA titles, makes for a daunting task. That’s because in order to build the game the AI must first understand its rules.

New chip designs are coming, and the biggest announcements aren’t for video games.

Sony is bringing machine intelligence to its image sensors. The electronics and entertainment giant announced this week a sensor that applies artificial intelligence while processing imagery without the need for extrema hardware or assistance.

Players in the photography industry have been focusing on increasingly greater numbers of pixels for ever-sharper enlargement and reproduction capabilities and on increasingly compressed devices for lighter weight and greater portability.

But Sony’s new IMX500 sensor is the first to improve efficiency though AI implementation.

Living organisms are capable of sensing and responding to their environment through reflex‐driven pathways. The grand challenge for mimicking such natural intelligence in miniature robots lies in achieving highly integrated body functionality, actuation, and sensing mechanisms. Here, somatosensory light‐driven robots (SLiRs) based on a smart thin‐film composite tightly integrating actuation and multisensing are presented. The SLiR subsumes pyro/piezoelectric responses and piezoresistive strain sensation under a photoactuator transducer, enabling simultaneous yet non‐interfering perception of its body temperature and actuation deformation states. The compact thin film, when combined with kirigami, facilitates rapid customization of low‐profile structures for morphable, mobile, and multiple robotic functionality. For example, an SLiR walker can move forward on different surfaces, while providing feedback on its detailed locomotive gaits and subtle terrain textures, and an SLiR anthropomorphic hand shows bodily senses arising from concerted mechanoreception, thermoreception, proprioception, and photoreception. Untethered operation with an SLiR centipede is also demonstrated, which can execute distinct, localized body functions from directional motility, multisensing, to wireless human and environment interactions. This SLiR, which is capable of integrated perception and motility, offers new opportunities for developing diverse intelligent behaviors in soft robots.