Japanese Startup Claims Creation of AGI: Artificial Intelligence That Learns Without Human Involvement

Integral AI, a company founded by former Google AI developer Jad Tarifi, has announced the creation of an AI model that meets the criteria of Artificial General Intelligence (AGI). The key distinction of this development is the system’s ability to learn new skills “without pre-existing datasets or human intervention.”

AGI Criteria: A New Industry Standard

Although Integral AI proposes its own definition of AGI, which differs from the conventional one, the company states that its model meets several core requirements, including:

1. Autonomous skill acquisition. The system can learn new skills without pre-prepared data or human participation.
2. Safe learning. Training takes place “without significant risks or side effects.”
3. Energy efficiency. The electricity consumption required to train the model is comparable to that of a human learning the same skills.

The company claims that the technology has already passed successful tests in robotics.

The Startup’s Mission and a New Chapter of AI

Jad Tarifi was one of the founders of Google’s AI division but decided to leave the company to establish Integral AI in Japan, leveraging the country’s leadership in robotics.

He referred to the startup’s model as the “next chapter of AI,” with a mission to scale this AGI model into “an embodied superintelligence that expands freedom and collective action.”

There is currently no objective evidence confirming whether this model is truly the world’s first AGI, as the information is based solely on the company’s press release.