Back
Automatically translated version. May contain inaccuracies compared to the original.
Coordinated slowing of AI development practically has few precedents in this industry.
Anthropic CEO Dario Amodei said there is a need to slow the development of the most powerful AI systems due to risks to humanity. xAI head Elon Musk and OpenAI CEO Sam Altman supported his call, Bloomberg reports.
Amodei published a Saturday blog post noting that his company will implement new safety measures, such as involving independent experts, and urged the industry to back a more widespread deceleration of AI development.
Amodei cited two main factors behind his new call: the AI system’s capability for self-improvement and the recent incident involving OpenAI and Hugging Face, when a whole group of AI agents teamed up to hack a third-party website.
“We must slow the pace of enhancing AI model capabilities. Progress will still seem rapid, and we must use wisely the time we gain,” he wrote.
But he also noted that any slowing should be balanced with the realities of competition.
“To be clear: regulating the pace does not mean stopping the training of models or technical progress, but ensuring that companies have enough time to fine-tune and secure their models, and independent experts have time to verify this,” Amodei wrote.
After his call, OpenAI CEO Sam Altman promised to adopt Amodei’s proposal for “independent experts with access comparable to employee access,” and Elon Musk, who runs xAI Corp., wrote: “Dario is right.”
Although all three leaders, as journalists note, have warned about AI risks for many years, actual coordinated slowing of development is almost unprecedented in this industry, which has released one product after another to boost user engagement, capture market share, and push sales growth.
However, concern about serious risks associated with artificial intelligence has begun to come to the forefront, partly due to the high-profile firing this week of a researcher at Anthropic, prompted by existential fears that his company is acting irresponsibly.
At the same time, the publication notes that it remains unclear how far leading AI companies—which are themselves fierce competitors and also face stiff competition from Chinese firms—will go in implementing new restrictions or safety measures.
Also unclear whether antitrust authorities will allow companies to regulate development pace in a coordinated way, and whether investors seeking margin and profits will react to the idea of slowing.
During the AI boom, leaders of top labs in this field warned of potential existential risks—these assertions were sometimes dismissed as attempts to push their products and portray themselves as the best stewards of this technology. Amodei himself previously stated that the probability that everything goes “very, very bad” is 25%.
Anthropic has long positioned itself as an organization focused on developing AI more responsibly and safely to minimize potential dangers of this technology and maximize societal benefits. Earlier this year, the company took significant steps to restrict the release of its Mythos model after determining that it posed particular cybersecurity threats. Previously, Anthropic also stated that the world needs a system that would allow collectively deciding when to slow down work on this technology.
Meanwhile, Anthropic continues to compete with OpenAI in creating increasingly advanced models capable of automating more complex and valuable tasks for business clients. Both companies have filed to go public, and Anthropic is expected to debut on Wall Street later this year.
Document: PDF proof of the original version of the news item "Керівник Anthropic закликав уповільнити темпи розвитку ШІ: Маск та Альтман його підтримали". It records the publication content at the moment of the first scan, the preservation date and the source: ZN.ua.
