Three-part series covers some of the GDC 2018 highlights: mobile gaming, Epic’s move on game development, and gaming at Unity.
The message at GDC 2018 comes through loud and clear, AI/ML and realtime rendering with a little blockchain thrown in for no good reason.
Mobile gaming is a powerful force at GDC this year too. In the near future, games will know no platform. The emphasis is playing any game, any time, on any platform. And Epic’s Fortnite, the darling of GDC 2018, showed how it’s done.
As an obvious given, the panel agreed that AI has long been a part of gaming. As an industry, gaming has nurtured AI during the long AI winter that started in the mid-80s when slow progress in AI research suggested traditional approaches were inadequate, and in fact, in some ways they were. Rather than trying to teach an AI everything they need to know, the industry has moved on to teaching machines how to learn, which lies at the heart of the current AI boom. Gaming is now getting the advantage of AI technological advances in hardware and software. First with rules-based behavior for NPCs (Non-Playing Characters) and today with enhancements driven by machine learning and inference.
Google’s Darin Briskman said that we’re living in a golden age for AI due in no small part to the power afforded us by the cloud. AI is cheap, affordable and accessible and certainly that was one of the overall messages of GDC this year.
What came up during the panel and should be emphasized over and over again, is that AI is an indefinite term and furthermore what used to be referred to as AI is really a pretty primitive use of rules-based responses. Briskman called for a shift in PC gaming from rules-based reaction to behavior based.
Unity’s Meuleau commented that Unity is trying to make AI more accessible to developers by automating it. The company there is working on incorporating learning into games so the game learns from players. As a result, NPCs will be better able to perform like reasonably good game players rather than neophytes, or fixed function robots. The idea of keeping the game play at a level consistent with that of the player was also brought up. It’s not really fun for players to be crushed by games, bots, or even other players so it’s possible that AI can actually slow down game play so the not-so-great players will stay engaged.
Magic is happening today with the arrival of machine learning and inference and its wide use in disparate fields. Nvidia’s Edelsten said this is such a great time to be working in these fields because there’s so much cross over. The work being done in AI for automotive can be used in game development for such capabilities as interpreting surroundings, identifying hazards, etc. SLAM can now be used to generate environments in a game. But also, truly magical possibilities are being explored such as realtime translation during game play, bringing gamers much closer together in a community. NPCs, those aimless bots wandering the landscapes of games, can now be capable of behaving in unexpected ways. They can lie, or switch allegiances if that serves the game.
“Game development is a lot harder than astrophysics. I have a pretty good idea where Mars is going to be. I had no idea they were going to set the cats on fire.”
Machine learning and AI are rapidly becoming larger than any segment of industry or any platform. The learning that has been happening as machines learn to identify cats, and people, and traffic patterns is that the information is breaking free of silos. Sure, companies may hope to be able to keep pools of data and knowledge safe and segmented, but it seems obvious that information is going to be free. Just as it would be monstrous for automotive companies to try and tie up the information that is coming in from cars about road conditions and safety, it’s not going to be possible to lock up information about health and disease, or standards for highways and building materials. So, these early wild west days of learning and building information stores is going to benefit everyone. It’s as if the Tower of Babel is being built again, but Machine Language is now the mother tongue.
Peter McGuinness, JPR’s VPU expert, has written a more detailed article about the panel and brings his own insight and experience to the story. See Five guys and Jon: Torching the cat, by Peter McGuinness.