A Tokyo-based technology company, Integral AI, has announced a breakthrough in Artificial General Intelligence, or AGI. The company says it has developed an AI system that can learn completely new tasks on its own, without using pre-existing datasets or receiving help from humans.
Researchers often describe Artificial General Intelligence as the ultimate goal of artificial intelligence research. Developers train AI systems for specific tasks such as writing text, recognizing images, or driving cars. Experts expect AGI to reason, adapt, and learn across many different tasks in a human-like way. For years, AGI has remained more of an idea than a real product.
Integral AI was founded by Jad Tarifi, a former Google engineer with more than a decade of experience in artificial intelligence. The company is based in Tokyo, Japan, and focuses on combining advanced AI systems with robotics. According to Integral AI, its new system represents a “fundamental leap” beyond current AI technologies.
In its announcement, the company explained that it defines AGI using three core conditions. The first is autonomous skill learning. This means the AI must be able to teach itself new skills in completely new situations without relying on existing data or human instructions.
The second condition is safe and reliable mastery. The system must learn without creating serious risks, unexpected behavior, or harmful side effects. The third condition is energy efficiency: the total energy used by the AI to learn a skill should be comparable to the energy a human would use to learn the same task.
Integral AI says these three conditions served as guiding principles throughout the system’s design and testing. Engineers at the company claim they were treated as strict benchmarks, not just ideas, during development.
The company also says it has already tested the system in real-world robot experiments. According to Integral AI, robots using the new AI learned new skills without any human supervision. The firm says this shows the system can adapt on its own, rather than following pre-programmed instructions. While the company has not fully shared technical details, it says the results support its claim of human-level learning ability.
Jad Tarifi described the announcement as a historic moment. “Today’s announcement is more than just a technical achievement; it marks the next chapter in the story of human civilization,” he said. “Our mission now is to scale this AGI-capable model system, still in its infancy, toward embodied superintelligence that expands freedom and collective agency.”
Tarifi has also spoken publicly about why he chose Japan as the base for Integral AI instead of Silicon Valley. After leaving Google, he said he wanted to work more closely with advanced robotics development.
Japan is widely known as a global leader in robotics. The country conducts strong research in humanoid robots, industrial automation, and intelligent machines. Tarifi believes this environment provides Integral AI with a stronger foundation. This foundation helps build AI systems that can interact with the physical world.
Despite the claims, many experts remain skeptical. The technology world has seen many “world’s first” announcements that critics later questioned or redefined. A well-known example is the debate over “quantum supremacy.” In this case, major companies made early claims. However, other researchers challenged those claims. They disagreed with the definition used.
Artificial General Intelligence is especially difficult to measure. There is no single, globally accepted test that proves an AI system has reached AGI. Integral AI has offered its own definition. However, experts say independent testing, peer-reviewed research, and open technical data needed before anyone can verify such claims.
Integral AI has also said its system mirrors the structure of the human neocortex. This part of the brain handles reasoning, perception, and language. Scientists caution that comparisons between AI models and the human brain are complex and often theoretical. Therefore, these comparisons remain difficult to confirm.
