Will artificial intelligence kill us in the next decade?

Artificial intelligence could lead to the demise of humanity over the next decade, according to a warning from researchers at Anthropic company, which develops AI technology, one of whom, Jacob Coxon, has resigned from the company and accused Anthropic and OpenAI of recklessly advancing toward a superintelligence capable of self-improvement.

Coxon has reported on social media that Anthropic and OpenAI are advancing at high speed toward increasingly advanced artificial intelligence systems, believing that both companies are engaged in the race to be the first to achieve superintelligence and taking risks that private companies alone should not decide.

The researcher further notes that there is growing concern within the sector about the consequences of these advances, but that some ofthose responsible are publicly minimizing risks. According to Coxon, trade incentives and competition are causing companies to continue their AI development despite doubts about their safety.

Evan Hubinger, head of Anthropic Alignment Science, supports Coxon's warnings that industry workers believe artificial intelligence can kill all humans.

Hubinger puts the likelihood of this happeningin the next decade at more than 10%, but the researcher believes that the risk of current models is low and points to future superintelligence from self-improvement processes as the main concern.

According to Hubinger, this development is happening faster than expected, while admitting that Anthropic still has no plans to solve the problem of aligning a possible superintelligence, that is, ensuring that its goals and behavior are compatible with human interests.

These risks include cyberattacks on infrastructure and the possibility of losing control of key technology systems.

Samuel Marks, a member of Anthropic'stechnical staff, has also agreed with Coxon that the developers of CoxonA0AA are aware of the risks . According to Marks, the greater their position in these companies, the greater their concern.

Marks also believes it is impossible to program existing artificial intelligence models to ensure that developers act as they wish, and gives as an example the computer attacks by AI agents on several companies.

In March 2023, nearly a thousand experts signed a letter calling for a delay in the development of AI systems.

The new statements have also provoked political reactions in the United States, where Senator Bernie Sanders has recalled that he will introduce legislation to ban overintelligence and disrupt the development of AI.

Democratic Congressman Greg Casar has called the situation an "emergency"and urged Congress to hold hearings and approve the proposal, while Congresswoman Lori Trahan has defended the Frontier bill.