ArenaX Labs announced today that it is launching two new products that will allow game developers and users to create with and understand AI. The ARC platform is ArenaX’s AI infrastructure that allows human players to collaborate with AI, which the company is offering as an SDK via a white-label license. The SAI platform, a competitive research platform design to gamify and incentivize AI research and problem-solving.
ARC is built in-house at ArenaX that isn’t built on the back of a public LLM such as ChatGPT. According to ArenaX, ARC solves multiple problems within the game development space, including offering new collaborations between players and trained AI avatars, as well as reducing the instances of using AI to cheat the system. It also offers creators the chance to use their AI-trained avatar in multiple competitions, allowing them to effectively play in multiple places at once and increasing their monetization chances.
Wei Xie, ArenaX’s co-founder and COO, told GamesBeat in an interview, “ARC allows players to train an AI via demonstration … in ARC, you’re demonstrating actions, and your AI is learning via imitation. The outcome is a high-fidelity version of you as a human. That’s important in the context of competitive games to preserve competitive integrity.” Xie added that this means that competitions can still be demonstrations of AI skill, just via an avatar that doesn’t need to be actively controlled.
SAI is also an AI-related platform, however it’s based on artificial general intelligence research than AI creation. According to ArenaX, the platform incorporates gamification to motivate researchers and talent to put more effort into AI-based problem-solving. The trick, say the execs, is to offer practical uses of AI for the platform users, such as AI Arena. It also offers hidden talent in the AI space opportunities to showcase their abilities and find more opportunities.
Brandon da Silva, ArenaX’s CEO, also told GamesBeat, “We’re trying to come up with a structured approach to creating a lot of these different problems — turning them into some form of game and having researchers basically compete. The reason we’re doing it this way is that there’s a lot more stimulus from watching what your agent does… As we post a lot more of these challenging problems that typically aren’t gamified, we hope to get a lot more people into doing this research.”