From Language to World: The Evolution of AI Towards Understanding Our Environment

At the forefront of AI technology, there is a paradigm shift underway from mere language models to "world models" that comprehend our environment. This shift is subtly reflected in the conflict observed between Sam Altman and Elon Musk.

This is highlighted by the emergence of a robot venture named Figure, which has gained attention due to OpenAI's investment.

Competing against Tesla, this company is developing humanoid robots with AI technology at its core.

The latest engine, implemented with the cooperation of OpenAI, incorporates a multimodal approach and demonstrates smooth conversational abilities in its demonstrations, symbolizing the era of multimodal AI.

In the industrial sector, Covariant's development of RFM1 is garnering attention.

As global leaders in the Factory Automation (FA) sector, like Japan's Yaskawa Electric and FANUC, accelerate their investments in AI, RFM1 stands out for its unique versatility.

RFM1's distinctive feature lies in its use of 8 billion parameters for pre-training, grounded in Transformer technology. This capability allows it to handle not just text but also multimodal data, including videos. It represents a significant leap beyond traditional language models, enabling AI to function as an interface with the physical world.

Furthermore, RFM1 also addresses methods of collaboration between humans and robots, anticipating more natural interactions. Future AI is expected to transcend mere text analysis, gaining a broader context understanding and advancing real-world applications.

Thus, in the realm of AI, we are witnessing an evolution from language models to world models. This transformation is poised to fundamentally alter the impact of AI on our lives and work. We are currently living through this historic moment.