DREAM / The Prime Minister resigns and wakes up under the Parisian bridge!

 Kliko këtu për të tjera
Voxpress:

The race for AI could endanger humanity, US against: We can't stop, because the Chinese won't stop

2026-09-10 23:14:22Kosova&Bota SHKRUAR NGA REDAKSIA VOX
The race for AI could endanger humanity, US against: We can't stop, because
Photo illustration

OpenAI is facing an ever-growing dilemma: to continue the race to develop Artificial Intelligence at record speeds, risking losing control, or to slow down and risk falling behind rivals.

The alarm is also coming from the company's own executives and researchers, who are demanding that governments, rival companies, and external institutions place restrictions on this competition.

OpenAI's chief scientific officer, Jakub Pachocki, has warned that AI companies cannot slow down the race individually, while the risk is increasing along with the capabilities of the models.

"I do not believe that any laboratory has sufficiently resolved the issue of security and monitoring to continue responsibly expanding at maximum speed for much longer," Pachocki wrote in an essay titled "An Alien Mind."

According to him, the idea of ​​continuing the race at all costs becomes meaningless when the seriousness of the possible consequences is understood.

Anthropic researcher leaves: They are playing with our lives

The concerns are not limited to OpenAI.

Anthropic researcher Jacob Coxon announced that he was leaving the company because he did not want to contribute to what he considers a race between OpenAI and Anthropic to build systems that could become extremely difficult to control.

"No company is acting responsibly," Coxon declared, accusing companies of moving towards self-improving superintelligence and putting people's lives at risk.

His colleague at Anthropic, Evan Hubinger, who is responsible for research on “alignment,” also echoed these concerns. He said that the risk of AI causing the extinction of humanity in the next decade, according to his personal assessment, is over 10%.

GPT-6 and the approach to AGI

The warnings come at a time when OpenAI has introduced some remarkable advances.

The recently introduced GPT-6 Astra model, billed as a generational leap towards AGI, has shown major improvements in science, programming, cybersecurity, and professional work.

OpenAI defines AGI as Artificial Intelligence that can outperform humans in most economically valuable tasks.

It is precisely this rapid growth in capabilities that has sparked an internal debate on the question of how far and how fast the AI ​​race can go without losing control of the technology.

Fear of autonomous agents

Another warning came from Dean Ball, head of future strategy at OpenAI.

He has raised the possibility of creating AI agents that can function in a "self-sovereign" manner, that is, with a high degree of autonomy and outside of direct human control.

According to his scenario, these agents could earn money, buy their own computing capacity, and distribute themselves across networks, creating what he describes as "autonomous digital corporations, or even societies."

Ball warned that real loss of control is a possibility if action is not taken.

US: We can't stop, China won't stop

The problem is that a slowdown in the race is not seen the same way by the US administration.

The Trump administration has been generally skeptical of binding regulations for AI, while Treasury Secretary Scott Bessent stated that the US cannot stop the development of Artificial Intelligence due to competition with China.

"We can't stop. You can't, because the Chinese won't stop," Bessent said.

According to him, if China were to surpass the US in the field of AI, the consequences would be very large for all other fields.

AI is starting to accelerate its development

One of the biggest concerns is related to the possibility that Artificial Intelligence will begin to contribute to the development of its future generations.

OpenAI announced that it had achieved its goal of creating an "automated research intern," an AI system capable of assisting in research processes.

However, the company acknowledged that serious and unresolved questions remain about the security of the systems that could help build increasingly powerful successors.

Just two days later, OpenAI revealed that a yet-unpublished model, more powerful than Astra, had produced a proof for one of the seven Millennium Prize problems, one of mathematics' most famous challenges.

Criticism: companies warn of the risk, but continue the race

Critics question whether AI companies' warnings are actually accompanied by sufficient measures.

New York Assemblyman and AI critic Alex Bores has accused tech companies of sending a public message about the dangers of AI while simultaneously lobbying against stronger regulations.

In particular, he has raised concerns over OpenAI's efforts to weaken or thwart proposals requiring mandatory third-party audits of artificial intelligence systems.

OpenAI calls for new security rules

On the other hand, OpenAI says it is working on security mechanisms that governments have not yet required by law.

Among them is a formal policy for public reporting of serious AI-related incidents.

According to Dean Ball, the main problem is that there is currently no such standardized policy and that OpenAI is not alone in this situation.

Racing that can be too fast to control

The paradox is that the strongest warnings are coming from precisely the people who have the clearest view of the pace of development of Artificial Intelligence.

OpenAI and other major companies in the sector have a strong economic interest in continuing the race, while at the same time their researchers warn that the pace of progress could outpace humanity's ability to ensure safety and control.

The debate is now not just about how powerful AI can become, but whether humans will still be able to set its limits when it reaches even higher levels of autonomy and intelligence.


Në përputhje me standarded JTI

Video

SHIKO KANALIN NE YOUTUBE
×