Shulou(Shulou.com)11/24 Report--

CTOnews.com, July 7 (Xinhua)-- OpenAI announced that it is forming a new team, led by its co-founder and chief scientist Ilya Sutskever and Alignment director Jan Leike, to find ways to guide and control the next generation of "super-intelligent artificial intelligence systems."

▲ source OpenAI is reported that the team currently has the highest priority and access to 20 per cent of the company's computing resources, with the goal of addressing the core technical challenge of "controlling super-intelligent AI" within the next four years. The team claims that they need time to make "scientific and technological" breakthroughs to guide and control AI systems that are much smarter than humans.

Researchers Sutskever and Leike believe that the "next generation of super-intelligent AI" may be born within a decade, which will be the most influential technology ever invented and can help solve many problems that previously require a lot of human resources to solve. But at the same time, the researchers believe that the technology also has the possibility of eating human beings, and its great power may lead to the loss of human sovereignty or even extinction.

The team said that in terms of "ensuring that smarter AI systems follow human intentions", they do not have a reasonable solution to guide or control potential super-intelligent AI. At present, humans can only adjust the generation process of AI through feedback, but humans cannot reliably monitor AI systems from the root.

OpenAI plans to build an AI "automated alignment researcher" at a roughly human level, then use a great deal of deep learning to expand AI capabilities and iteratively adjust the super intelligence.

To this end, researchers need to develop scalable training methods, then verify the generated model, and finally conduct a series of controllable tests on the model.

CTOnews.com found that researchers are now planning to use the AI system to help evaluate other AI systems (scalable oversight), while also hoping to understand and control how their models extend oversight to tasks that human researchers cannot monitor.

Sutskever and Leike said that as the research deepens, it is expected that their future research focus may also change, and there may even be new research areas to follow.

At present, the new team is also recruiting machine learning researchers and engineers. Sutskever and Leike believe that recruiting more machine learning experts is crucial to solving the problem of the next generation of super-intelligent AI. They plan to widely share the team's research results and will promote the common progress of the industry.

The new team's work complements OpenAI's existing work to improve the security of current models such as ChatGPT and to understand and mitigate other risks posed by artificial intelligence, such as abuse, economic disruption, misinformation, prejudice and discrimination, addiction and overdependence. While the new team will focus on the machine learning challenge of combining super-intelligent AI systems with human intentions, we are actively working with interdisciplinary experts to solve related socio-technological issues to ensure that our technological solutions take into account a wider range of human and social problems.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.