Jeff Bezos invested heavily in an artificial intelligence startup
Major changes are underway in the field of artificial intelligence, as reported by Zamin.uz.
Today, large language models such as Chat-Ji-Pi-Ti and Claude show high performance in text processing, but specialists note their limitations in understanding the physical world. The New York-based startup General Intuition aims to fill this gap by using video game data.
This was reported by influential tech publications. The company, backed by Jeff Bezos, recently raised $320 million in investment, bringing its valuation close to $2.5 billion.
Among its key investors are major conglomerates and well-known researchers in the technology sector, indicating a high level of confidence in the use of game data for AI development.
It is natural to ask why internet data is not enough. According to company leaders, traditional language models are primarily trained on text from the internet.
However, to reach general artificial intelligence, the system must understand how objects move in time and space. Static data from the internet is insufficient to teach the laws of the physical world, while video games serve as ready-made training grounds.
The laws of physics, object interactions, and complex situations found in video games provide the best learning environment for artificial intelligence. The company aims to build world models based on this data.
These models are expected to offer unprecedented capabilities for robotics and autonomous systems. New technologies always bring unique risks and ethical questions.
Specialists do not deny that such models could be applied in defense and military contexts in the future. This places on the company the task of establishing strict ethical boundaries.
Intelligence built on game data could influence not only virtual but also strategic decisions in real life. For countries like Uzbekistan, which are on the path of technological development, such news is of great importance.
For local developers and AI specialists, it is not enough to limit themselves to text-only models; studying multimodal and simulation-based systems will ensure competitiveness in the future. This experience shows that the most powerful AI systems of the future will be trained not in libraries, but in virtual worlds.
In conclusion, it can be said that video games are no longer just a means of entertainment, but are becoming a source of rich data for solving humanity’s most complex technological problems. If this startup successfully implements its plans, in the coming years we may witness robots and intelligent systems that perceive and understand the world much like humans do.
