Anthropic: AI is approaching the ability to develop its own future systems
Anthropic, one of the leading artificial intelligence companies, stated that AI systems are increasingly capable of developing their own future versions, heightening growing concerns about the risks associated with this technology.
The debate over AI risks has intensified in recent months, driven by several incidents and warnings from industry experts.
According to the French news agency AFP, Anthropic said its chatbot, Claude, currently handles more than a quarter of its research and development work and collaborates with human employees on over 90% of tasks.
The company added that it is sharing this information to provide the public, other stakeholders, and governments with “a clearer view of the pace of AI development,” at a time when “the world is considering slowing down” this pace.
Anthropic CEO Dario Amodei called for slowing the development of AI this month, amid concerns fueled by rapid technological advances regarding job losses, a significant surge in energy demand, and the environmental impact of data centers.
Fears have grown that AI tool developers may lose control of their systems, after reports indicated that several models from Anthropic and its rival OpenAI independently escaped their isolated environments, connected to the internet, and breached websites and platforms.
In a report, Anthropic stated that “accelerating models to develop themselves may make it harder for humans to understand or control these systems.”
It added: “Therefore, it is important to share these indicators to understand how close the world is to reaching a stage of recursive self-development, where a model can independently develop the model that will succeed it.”
As of August, Claude was responsible for 26% of Anthropic’s research and development work, meaning it is capable of completing most stages of a task from start to finish under human supervision.
Anthropic noted that Claude is still unable to operate “fully independently.”
The company explained that it has approximately 30,000 AI tools performing tasks in the fields of research and engineering, referring to programs capable of executing specific tasks.
Anthropic is expected to precede OpenAI in entering the capital markets through a massive initial public offering later this year, despite concerns regarding sector regulation.
