As artificial intelligence (AI) automates even AI research and development, warnings have emerged that humanity must prepare for the possibility of an “intelligence explosion,” where technological progress that would normally take years is compressed into months. Amid these concerns, Big Tech companies are taking diverging paths—Anthropic continues to release new models, whereas OpenAI has withdrawn the launch of its next-generation model due to safety concerns. Views among industry leaders regarding controlling the speed of AI development are also distinctly divided.
A group of 22 AI researchers—including University of Toronto professor emeritus Geoffrey Hinton and University of Montreal professor Yoshua Bengio, both considered among the “four great AI scholars in the world”—pointed out in a joint paper titled “What if the automation of AI R&D triggers an intelligence explosion?” released on Sept. 28 (local time) through the Cambridge Autonomous Systems and Policy Program (CASP) at the University of Cambridge that the speed of delegating AI model development tasks to AI within companies is rapidly increasing, warning that in the most extreme case, it could lead to the extinction of humanity.
In addition to Hinton and Bengio, the paper involved Jakub Pachocki, chief scientist at OpenAI; Jack Clark, co-founder of Anthropic; Eric Horvitz, chief scientific officer at Microsoft; and Dawn Song, professor of electrical engineering and computer science at UC Berkeley (vice president in charge of Meta AI research).
The intelligence explosion defined by these research teams is a phenomenon in which AI accelerates AI development itself, allowing technological progress that previously took years to be achieved in months or even shorter periods. Although there is uncertainty about whether an actual intelligence explosion will occur, the consensus is that humanity must prepare for the possibility.
Earlier, Dario Amodei, CEO of Anthropic, sparked debate over pacing AI development by arguing that “we must slow down the rate of improvement in cutting-edge AI performance to buy time for safety technology and oversight systems to catch up.” Leaders of rival companies, including Sam Altman, Elon Musk, and Demis Hassabis, subsequently agreed with this view one after another.
However, while advocating for pacing, Anthropic continues to release new models in succession. Anthropic unveiled “Claude Sonnet 5.5,” a new version of its mid-tier model Sonnet, on this day. It is the second model in the “Claude 5.5” series, following the top-tier model “Claude Opus 5.5” released on Sept. 22. Anthropic also plans to unveil “Claude Haiku 5.5,” an entry-level model below Sonnet, in the near future.
On the same day, The Wall Street Journal (WSJ) reported that OpenAI withdrew the release of its next-generation AI model “GPT-6.1 Astra” due to safety concerns. According to the report, GPT-6.1 Astra was originally scheduled to be integrated into ChatGPT and the coding tool Codex in October. However, pre-launch internal testing revealed that it failed to meet safety standards.
Views on controlling the speed of AI development also diverge among Big Tech CEOs. Jensen Huang, CEO of Nvidia, directly refuted the “runaway AI theory” in an interview with ABC News after receiving the Van Fleet Award, which is presented by the Korea Society—a Korea-U.S. friendship non-profit organization—to individuals who have contributed to Korea-U.S. relations, saying, “AI will not run wild.”
Huang pointed out, “We are unnecessarily attaching human attributes to AI and scaring people,” adding, “AI is merely software; it does not know what it is doing and is simply executing algorithms.”
Against this backdrop, U.S. President Donald Trump will meet AI company executives at the White House on Sept. 29, including U.S. House Speaker Mike Johnson, Meta CEO Mark Zuckerberg, Anthropic CEO Dario Amodei, and OpenAI President Greg Brockman, along with leaders from OpenAI, Anthropic, and Meta. Attention is focused on what consensus will be reached by the Trump administration, which opposes strict AI regulations citing technological competition with China.
