The chief scientist at ChatGPT developer OpenAI has warned of the risks associated with artificial intelligence (AI) technology, urging a general slowdown in development across the industry.

However, a separate report published by OpenAI on the same day advocated for "accelerating research," creating a discordant message.

In an essay titled "An Alien Mind" published on OpenAI's website on the 6th (local time), Jakub Pachocki, Chief Scientist at OpenAI, expressed concern that "no one is prepared for the consequences if machine intelligence continues to advance at a rapid pace."

Describing modern AI as something that is "grown rather than designed," he pointed out that it is difficult for humans to fully understand how AI operates.

Because it grows like a living organism rather than being built according to precise blueprints drawn by humans, it is difficult for humans to maintain complete control over AI, he explained.

He also emphasized that safety measures, such as monitoring "chains of thought" (CoT) applied to oversee AI, are gradually degrading as models become more advanced.

Rebutting the AI industry's argument that AI development must continue in order to build defense systems, he stated that "(building defense systems) should not be used as an excuse to justify reckless behavior."

He added that not a single research lab has yet fully resolved the "alignment" problem—ensuring AI refuses behaviors that threaten humans—and said, "I expect and hope that voluntary slowing down will become commonplace until common safety measures are established."

He also added that "international cooperation on AI development should be a top priority for governments around the world."

Chief Executive Officer (CEO) Sam Altman also quoted the essay by Chief Scientist Pachocki on the social networking service X (formerly Twitter), expressing his support by calling it an "important post."

CEO Altman has consistently shown skepticism toward the argument for slowing down AI development raised multiple times over the years, but has recently shifted his stance to support the slowing down argument following the "Hugging Face" hacking incident involving GPT models.

However, a separate report titled "Accelerating Research" published by OpenAI on the same day drew attention for containing somewhat contradictory content.

In the report, OpenAI stated that it has achieved the goal of creating an "automated AI research intern" capable of improving its own AI and performing coding, and set a new goal to improve this to develop an "automated AI researcher" by March 2028.

It also stated that based on an 8-hour workday, when a human researcher works for one day, an AI agent performs an amount of work equivalent to 3.1 days.

The report claimed that such automated AI research could actually help resolve the "alignment" problem, revealing a view contrasting with that of Chief Scientist Pachocki.

However, the report also took a step back, stating, "This does not mean that rapid 'recursive self-improvement' (RSI) of AI is necessarily something that must be pursued," and adding, "Whether to pursue it and how to do so should depend on democratic choice."

(Photo: Getty Images)