SEOUL, South Korea (AP) – Major artificial intelligence companies renewed their commitment to safely develop AI at a mini-summit on Tuesday, while world leaders called for public support to advance research and testing of the technology. agreed to establish a network of safety agencies.
Google, Meta, and OpenAI are among the companies that have taken voluntary safety initiatives. AI Seoul SummitThis includes unplugging state-of-the-art systems if the most extreme risks cannot be contained.
The two-day meeting will be a continuation of the November meeting. AI Safety Summit Bletchley Park in the UK is at the forefront of efforts by governments and global institutions to design guardrails for this technology amid concerns about the potential risks it poses to both daily life and humanity. It was done inside.
The UK government, which co-hosted the event, said in a statement that leaders from 10 countries and the European Union would “build a common understanding of AI safety and coordinate efforts on AI research.” The network of safety agencies will also include those already established by the UK, US, Japan and Singapore since the Bletchley conference, it said.
Seven months after the Bletchley talks, United Nations Secretary-General António Guterres said in the opening session that, “From disinformation to mass surveillance to the possibility of autonomous lethal weapons, we are witnessing life-changing technological advances and life-threatening advances.” “We are facing new risks.”
In a video address, the UN Secretary-General said there is a need for universal guardrails and regular dialogue on AI. “We cannot sleepwalk into a dystopian future where the power of AI is controlled by a small number of people, or worse, by algorithms beyond human comprehension,” he said.
The 16 AI companies that signed on to the safety commitment also include Amazon, Microsoft, Samsung, IBM, xAI, and companies from France. Mistral AIChina's Zhipu.ai, and G42 in the United Arab Emirates. They committed to responsible governance and public transparency, and vowed to ensure the safety of cutting-edge AI models.
This isn't the first time an AI company has made an exaggerated but non-binding safety commitment. Amazon, Google, Meta, Microsoft Registered last year Follow voluntary safeguards brokered by the White House to ensure products are safe before they are released.
The Seoul conference revealed that some of these companies newest version their AI model.
The Safety Pledge includes publishing a framework for how companies will measure the risk of their models. In extreme cases where the risks are severe and “unbearable” and the risks cannot be mitigated, AI companies will need to hit the kill switch and stop developing and deploying their models and systems.
Since last year's UK conference, the AI industry has “increasingly focused on its most pressing concerns, including misinformation and disinformation, data security, bias, and the constant surveillance of humans,” one of the conference's leaders said. said Aiden Gomez, CEO of Cohere. AI companies that have signed the agreement. “It is important to continue to consider all possible risks, prioritizing those that are most likely to cause problems if not properly addressed.”
Governments around the world are scrambling to develop regulations around AI, even as AI technology advances rapidly and threatens to transform many aspects of daily life, from education and the workplace to copyright and privacy. I am. Although there are concerns that advances in AI will lead to job losses, spread false information Or it could be used to create new biological weapons.
This week's conference is just one of many initiatives around AI governance. united nations general assembly Approved the first resolution on the safe use of AI systems US and China We recently held the first high-level consultation on AI and a European Union global first. AI law It is expected to take effect later this year.
__
Chan contributed to this report from London. Associated Press writer Edith M. Lederer contributed from the United Nations.
