Society must learn to accept certain negative aspects of artificial intelligence in exchange for its advantages and broad accessibility. This sentiment was expressed by Sam Altman, CEO of OpenAI, during an interview with Politico.

"We believe the world must accept some bad things for the benefits of this technology and the freedom it gives people," he stated.

Among the potential downsides Altman identified are significant breaches, misuse of AI, and fraud. He mentioned that completely eliminating these risks would require stringent restrictions, a trade-off he is unwilling to make. According to him, humans will achieve "far more good than bad" through the use of artificial intelligence.

However, Altman acknowledged that there are limits to acceptable risks, noting that he would not tolerate catastrophic outcomes, particularly a severe loss of control over AI systems.

Differences with Anthropic

The interview focused largely on the contrasting approaches of OpenAI and Anthropic regarding AI safety and regulation.

Altman emphasized that there is a "fundamental difference in perspective" between the two companies. OpenAI advocates for broader user access to powerful systems and less stringent regulations, while Anthropic takes a more cautious stance.

He recognized the rationale behind proponents of strict limitations, who argue that the technology could become so powerful and dangerous that it should be managed by a single lab in San Francisco, which would oversee its use and control access to its benefits.

Altman disagreed with this viewpoint, labeling it an unacceptable compromise. He believes that the pursuit of completely eliminating harm should not result in concentrated control of the technology within a single entity.

Convergence of Views

Despite their differences, OpenAI and Anthropic have seen some convergence in their views on safety in recent months.

On September 12, Anthropic CEO Dario Amodei urged developers to slow the pace of advancements in AI models to ensure that safety measures keep up with the technology's evolution.

He suggested that independent experts should have ongoing access to these systems to evaluate AI behavior and verify compliance with developer commitments.

On the same day, Altman publicly endorsed the idea of regulating the speed of advanced model development and agreed on the need for independent assessments.

OpenAI has also increased its support for mandatory regulations. On September 9, the company advocated for federal safety requirements linked to the capabilities of neural networks.

Before national regulations are established, it backed four bills in California concerning independent AI assessments, auditor standards, youth protection, and defenses against biological threats.

The ChatGPT developer specifically highlights serious threat categories: offensive cyber capabilities, biological and chemical risks, harmful manipulations, and loss of control over systems. Special procedures for risk assessment and mitigation are designated for these threats.

On September 23, during a United Nations Security Council meeting, Altman stated the necessity of maintaining powerful AI systems under human control and fostering international collaboration in this area.

It is worth noting that a U.S. Senate subcommittee has begun investigating OpenAI's actions following an incident involving Hugging Face in July.

Follow ForkLog on social media: Telegram (main channel), Facebook, X