The Artificial Intelligence Security Summit, held this week in the United Kingdom, closed this Thursday with the commitment of the main technology companies to work together with governments in the future tests of their new models. And, in this way, limit the risks related to this technology, which the countries present have described as potentially catastrophic in a joint statement.
British Prime Minister Rishi Sunak highlighted that the United States, the European Union and other countries had agreed with a select group of companies to rigorously test artificial intelligence models, before and after their launch to the public. Representatives of OpenAI, Anthropic, Google, Microsoft, Meta and xAI They participated in the sessions this Thursday in which these tests were discussed.
To advance in this sense, they have asked Yoshua Bengio, a Canadian computer scientist and recognized as one of the godfathers of artificial intelligence, to lead the development of a “state of the science” report. The report will help understand the current capabilities of the technology and prioritize related dangers. Bengio is a Turing Award winner and a member of the UN Scientific Advisory Council.
“So far, the only people who have tested the safety of new AI models have been the same companies that develop them,” Sunak said in a statement. release. “We should not depend on them to correct their own homework, as many of them agree.”
The British government said that companies and governments have supported the creation of a new global testing center based in the United Kingdom. He announced that special attention will be paid to the dangers to national security and society.
China was left out of tests to reduce the risks of artificial intelligence
The first day of the summit had given a rare display of global unity. The meeting was opened by Michelle Donelan, Secretary of Technology of the United Kingdom, accompanied by the Secretary of Commerce of the United States, Gina Raimondo, and the Chinese Vice Minister of Science and Technology, Wu Zhaohui.
China, in fact, is one of the 28 countries that signed the Bletchley Declaration, in which governments agreed to demand transparency and accountability from companies that develop artificial intelligence systems. They also agreed to create plans to measure, monitor and mitigate potentially harmful capabilities.
China’s representative, however, was not present Thursday and did not ratify the testing agreement. Sunak had sparked controversy in his country’s parliament for having invited the Chinese government to the summit. He was criticized for ignoring the resistance of many Western countries to collaborating on technological matters with Beijing.
The prime minister has said in his defense that such an important issue requires the participation of the main actors. In addition, Sunak has insisted on the relevance that the United Kingdom could have as an intermediary between the economic blocs of the US, China and the European Union.
Existential risks and “real” risks
Vice President Kamala Harris, leader of the US entourage, said her administration’s “bold action” should be “inspiring and instructive” to other countries. Harris said the main focus should be on how artificial intelligence could widen inequality within societies and between countries.
The US official had already insisted on Wednesday on the need to adopt a “more practical approach”, which is not limited to evaluating existential risks. “There are additional threats that also demand our action, threats that are currently causing harm and that many people also consider existential,” Harris said.
The vice president cited the example of an elderly man who canceled his health care plan due to a faulty artificial intelligence algorithm. She also mentioned the case of a woman who was threatened by an abusive partner with modified photos. We have to work “across the spectrum,” Harris said, beyond existential threats about massive cyberattacks or biological weapons.
Identity cloning, autonomous driving through artificial intelligence and disinformation campaigns are among the most common dangers detected in the development of this technology. In this sense, Harris announced a new artificial intelligence security institute in the US. It will be aimed at developing standards for testing systems for public use and will collaborate with the UK test center announced by Sunak.
A fund to help African countries
Sunak’s office also reported that governments had given the go-ahead to share assessment results with other countries “where relevant”. He explained that part of the objective is to create shared standards to mitigate the risks of artificial intelligence.
The Foreign Secretary of the United Kingdom, James Cleverly, For its part, it announced a new fund of 100 million dollars. The initiative is backed by the US and other governments and aims to assist in the deployment of this technology in Africa. For example, including 46 African languages in the development of some artificial intelligence models.
The Secretary General of the UN, Antonio Guterres, also attended the summit, which was held in the Bletchley Park, headquarters of the United Kingdom’s code-breaking efforts during World War II. Guterres urged a coordinated global effort, comparing the risks of artificial intelligence to the Nazi threat, which British code breakers struggled to combat. “The threat posed by AI is more insidious, but could be just as dangerous,” he said.
The UN created the first global body on the governance of artificial intelligence last month. It is made up of an advisory group of 39 members, which belongs to institutions, governments and technology companies from around the world. Executives from Microsoft, Google and OpenAI are part of the group. This has proposed to deliver its own diagnostic report before the end of the year.