OpenAI’s Q-Star Initiative
In late 2024, OpenAI’s Q-Star project garnered attention for its ambitious aim to create a 125 trillion parameter multimodal model, signaling a substantial leap toward AGI. Although the launch was postponed due to high inference costs, the project’s scale underscores OpenAI’s commitment to advancing AI capabilities.
MIT’s Q-Star 2.0: Self-Evolving AI
Researchers at the Massachusetts Institute of Technology (MIT) introduced Q-Star 2.0, an AI model with real-time self-evolving capabilities. This model distinguishes itself by learning from new challenges, exhibiting human-like cognitive abilities, and excelling in complex problem-solving tasks. Its performance on the ARC AGI benchmark, surpassing previous models, marks a significant milestone in AGI research.
Google DeepMind’s World Modeling AI Team
In early 2025, Google DeepMind announced the formation of a new AI research team dedicated to developing “world models” capable of simulating physical environments. Led by Tim Brooks, this initiative aims to create realistic training scenarios for robots and AI systems, contributing to the broader pursuit of AGI.
