Prominent leaders in the AI industry, including Dario Amodei from Anthropic, Sam Altman from OpenAI, and Elon Musk from xAI, have recently agreed on the need to slow down the development of artificial intelligence. This agreement came after the resignation of one of Anthropic's researchers and his warning about the threats posed by this technology to humanity.
Challenges Facing AI Development
Despite this agreement, various obstacles such as domestic and global competition, profit motives, and political pressures from U.S. officials, particularly Donald Trump, stand in the way of a coordinated effort to control the pace of AI development. Nick Ries, former director of emerging technology policy at the Department of Homeland Security, believes that this moment could prompt the industry to rethink its definition of success.
Read more: The trial of Huawei's conspiracy to commit crimes has begun in New York
Ries states, "If we consider success in AI similar to success in the aviation industry, this success means establishing on-time and accident-free flights." He also points out that while we do not tolerate commercial airplane crashes, significant failures in AI seem to be more acceptable.
Amodei's Plan to Slow Development
Amodei has recently published a detailed paper in which he presents a plan to slow down the development of advanced AI. One of his suggestions is that leading AI companies, known as frontier labs, commit to providing continuous and similar access for independent evaluators as their employees. Evaluation teams should have offices, access cards, and company laptops to monitor safety practices.
Amodei has announced that Anthropic has unilaterally committed to this step. Altman has also described this plan on social media as a "great idea" and stated that OpenAI will follow this decision.
Altman has also welcomed a "federal framework" for setting safety standards for advanced AI and believes that companies do not need to wait for legislation or antitrust exemptions before addressing AI safety.
In Amodei's paper, he also mentions the need for regulations and government intervention so that collaboration among leading AI companies can be facilitated with the help of the U.S. government and other democratic countries. He expressed hope that leading labs could establish common safety standards and limitations on the unchecked rate of AI advancement.
International Challenges and Practical Barriers
Amodei also referred to the cooperation of democratic countries with "autocratic governments" and acknowledged the challenges, particularly in collaborating with China. He outlined various levels of agreements that global leaders could sign, stating that the simplest agreement might be a ban on "certain narrow and clearly dangerous uses of AI," such as the production of biological weapons.
However, he also noted that reaching an agreement on model testing before release, especially in areas like cybersecurity or biological threats, will be very challenging. He stated that setting a "speed limit" for AI models that can produce improved versions of themselves will be a significant challenge.
Altman announced on social media that "regulating the speed" of AI development does not mean stopping it, and progress in this field will continue. He also emphasized that "any competitive pressure from the U.S. should not be an excuse for recklessness." While experts view this agreement as promising, many questions remain unanswered.
Read more: Deep concerns among AI developers: Is speed coming at the cost of safety? · Key leaders issued a warning about AI development
Source: abcnews.com
