To crack down on China’s development of AI, it is reported that the United States plans to restrict the export of artificial intelligence model software

Reuters reported on Wednesday (8th), citing people familiar with the matter, that the Biden administration is preparing to open a new front to protect U.S. artificial intelligence (AI) from China, and initially plans to use the most advanced artificial intelligence (AI) models such as ChatGPT Wait for guardrails to be set up around the core software of the AI ​​system.

Reuters reported on Wednesday (8th), citing people familiar with the matter, that the Biden administration is preparing to open a new front to protect U.S. artificial intelligence (AI) from China and initially plans to use the most advanced artificial intelligence (AI) models such as ChatGPT and other AI systems. Set up guardrails around core software.
People familiar with the matter said the U.S. Commerce Department is considering new regulatory measures to restrict the export of proprietary or closed-source AI models, whose software and training data are confidential.
The report pointed out that any action by the U.S. government would be in addition to a series of measures implemented over the past two years that were aimed at blocking the export of complex AI chips to China and slowing down Beijing\’s development of cutting-edge technology for military purposes.
Even so, regulators will struggle to keep up with the industry\’s rapid growth.
The U.S. Department of Commerce declined to comment before publication time, and the Chinese Embassy in Washington did not immediately respond to a request for comment.
It is understood that the United States is currently unable to prevent local AI giants such as OpenAI invested by Microsoft (MSFT-US), Google DeepMind owned by Alphabet (GOOGL-US), and Anthropic from selling their technologies to others.
Each of these companies has developed some powerful closed-source AI models.
As a result, researchers in the U.S. government and private sector worry that competitors may use these models to conduct aggressive cyberattacks or even create powerful biological and chemical weapons.
These models can mine large amounts of text and images and summarize information and generate content.
In order to enact export controls on AI models, the U.S. may turn to thresholds included in an executive order on AI issued last October that are based on the computing power needed to train the models, people familiar with the matter said.
When this level is reached developers must report their AI model development plans and provide test results to the U.S. Department of Commerce.
Two U.S. officials and another person briefed on the discussions said the computing power threshold could become the benchmark for determining which AI models are subject to export restrictions.
If this threshold comes true, it will only restrict the export of unreleased models because no existing technology has reached this threshold yet, although Google\’s Gemini Ultra may meet the criteria for restricting exports, according to EpochAI, a research organization that tracks AI trends.
People familiar with the matter emphasized that the government has not yet finalized the proposed rules.
But the fact that such a move is being considered signals that the U.S. government is further cracking down on Beijing’s efforts to develop AI despite the difficulties of enforcing a strong regulatory regime on this rapidly evolving technology.
\”As the Biden administration looks at competition with China and the dangers of sophisticated AI, AI models are clearly one of the tools and potential bottlenecks to consider,\” said Peter Harrell, a former official at the National Security Council.
” He believes it remains to be seen whether the technology will actually turn into an export-controllable bottleneck.
In addition, people familiar with the matter also revealed that any new export regulations may also target other countries.
However, no matter how the United States sets the export threshold for AI models, it will be difficult to control because many models adopt an open source model.
Tim Fist, an artificial intelligence policy expert at the Washington think tank CNAS, said enforcing controls even on more advanced proprietary models will be a challenge because regulators may struggle to determine the right criteria to determine which models should be subject to controls.
In addition, he pointed out that China may be only about two years behind the United States in developing its own AI software.

Like (0)
Previous May 8, 2024 1:59 pm
Next May 9, 2024 9:42 am

Related posts