UIUC, Google, Stanford, UC Berkeley, CMU Joint Review: How Proactive AI Advances Towards Superintelligence

The four core characteristics of an agent: reactivity, social ability, autonomy, and proactiveness (proactiveness) were proposed as early as 1995. Proactive AI achieves its goals by setting objectives and taking action, with its development path including moving from simple task planning to complex world modeling, and potentially achieving superintelligence through multi-agent collaboration and general reasoning capabilities in the future.