Until now, a doomsday scenario or the idea of a catastrophic disaster wiping out the human race appeared far-fetched, confined to our screens. However, with the rapid advent of AI, these fears are gaining traction, according to a report sanctioned by the United States government and founded on hard facts.
A report published by Gladstone AI that describes and assesses the potentially destabilizing effects of advanced AI on global security “in ways reminiscent of nuclear weapons” is provided. Concerns regarding AI’s potential replacement of human labor appear negligible in light of the report’s subsequent findings.
The US government hired Gladstone AI to do research on the security risks and proliferation of weaponized and misaligned AI in October 2022, one month before the ChatGPT debut. The $250,000 federal contract required a four-person firm, Gladstone AI, to complete the 247-page report within a year. Gladstone AI’s mission is to brief government employees on AI developments.
Following extensive consultations with more than two hundred stakeholders, such as government officials from the United States and Canada, prominent cloud service providers, and AI safety groups with a specific focus on computing and security experts, the report strongly advises the United States government to take prompt and decisive action in order to avert the escalating threats that AI poses to national security.
Advertisement
Concerns were also expressed during these discussions by a number of AI safety experts regarding the possibility that adverse incentives could influence the decision-making processes of company executives.
The report states that the competition within the industry to attain AGI and, consequently, obtain the majority of economic benefits exacerbates the dangers posed by AI. According to Jeremie Harrison, co-founder of Gladstone, the current inclination is to develop systems with sufficient capabilities to be weaponized in a catastrophic manner or to be uncontrollable.
Furthermore, he stated that a calamitous event that completely halts AI research for everyone would be the worst-case scenario, meaning that not even the extraordinary benefits of this technology for scientific research or healthcare could be realized. Numerous technology executives have discussed AGI, or Artificial General Intelligence, which pertains to machines that are capable of executing tasks that are on par with, if not surpassing, human capabilities. Although AGI systems are not currently in existence, it is highly probable that they will be implemented within the next five years, if not sooner.
In an effort to proactively tackle these challenges, the report additionally presents an Action Plan that delineates an all-encompassing and unparalleled collection of policies—their implementation of which would wreak havoc on the AI industry. It also notes, however, that the current rate of AI development is so rapid that events may have already transpired by the time conventional policymaking concludes.
The action plan suggests that the US government contemplate the implementation of legislation that would criminalize the training of AI models using computational power that surpasses a predetermined threshold. Additionally, it advises that AI companies, such as Google and OpenAI, secure authorization from the government prior to training and deploying new models that surpass a lower computational threshold.
Advertisement
Moreover, it underscores the critical nature of investigating the feasibility of implementing regulations that forbid the public from learning the complex intricacies or “weights” of powerful AI models, possibly via open-source licenses, and impose severe penalties such as incarceration. Furthermore, the production and exportation of AI chips must be subject to stricter government supervision. Additionally, federal funding should be directed toward research endeavors that aim to standardize sophisticated AI technologies with safety protocols.
Meanwhile, the world’s tech behemoths are rapidly building infrastructure to train the next generation of more powerful systems, with some planning to employ 10 or 100 times more computing power. At this watershed moment, the involvement of governments and regulatory bodies is important in establishing safety standards based on ethical considerations. Equally crucial is worldwide coordination to guarantee that AI is used to advance humanity rather than destroy it. Smartphones have made technology accessible to everyone, but as the saying goes, with great power comes great responsibility. AI’s importance, applications, threats, and prospective ramifications must all be included in our educational framework, not only to attain balance, but also to live as a species!
