At General Intuition's New York R&D floor, an AI agent plays video games for 100 hours straight, while a large robot navigates the office using gameplay data to learn about the physical world.
The same brain powering these agents can generalize from gaming environments to real-world robotics. This technology could revolutionize how we train AI in complex tasks and environments.
General Intuition raised $320 million at a valuation of $2.3 billion, backed by investors like Khosla Ventures and former Google DeepMind researchers. The aim is to use human action data from gameplay to create more intuitive and adaptable AI agents.
Their proprietary data set, amassed from Medal's gaming platform, could become the backbone for generalized agents that operate seamlessly in both virtual and real-world scenarios. But as with any powerful technology, the ethical considerations are significant.
