In a shocking revelation, a recent report commissioned by the federal government has urgently called for immediate action to prepare the United States for a potential “extinction-level” threat from artificial intelligence (AI).
As reported by Time, the report titled “An Action Plan to Increase the Safety and Security of Advanced AI” was conducted by Gladstone AI. It proposes sweeping policy changes that could drastically shake up the AI industry.
The report boldly suggests that Congress should enact laws limiting the amount of computing power used to train AI models, a move that would undoubtedly send shockwaves through Silicon Valley.
Recently, tech giants like OpenAI and Google have made headlines with advanced AI systems such as GPT-4 and Gemini. However, these developments have also raised eyebrows due to some questionable computer-generated images.
The Gladstone report advocates for the establishment of a new federal AI agency. This body would have the authority to require tech companies at the forefront of AI development to obtain federal permission before advancing their software beyond certain limits.
Furthermore, it recommends that Washington impose stricter controls on the production and export of AI chips.
The State Department commissioned this report in 2022, which was written by Gladstone, a small four-person company specializing in providing technical briefings on AI for government employees. Despite being commissioned by federal authorities, it does not necessarily reflect current government views.
The report warns about the escalating risks of advanced AI development to national security. It likens the rise of advanced AI and AGI (artificial general intelligence) to nuclear weapons in terms of their potential to destabilize global security.
AGI is a hypothetical technology capable of performing tasks at or above human level. While it doesn’t exist yet, many labs claim to be working towards this goal.
The report’s authors interviewed members from OpenAI, Google, DeepMind, Anthropic and Meta. They painted a grim picture of the “perverse incentives” driving decision-making in these corporations.
According to Time, the report also highlighted that about 80% of Americans fear that AI could inadvertently trigger a catastrophic event.
Gladstone officials Jeremie and Edouard Harris have cautioned the government about AI risks for several years. They identified the State Department’s Bureau of International Security and Nonproliferation as one agency responsible for managing these risks.
The report also discusses the risk of weaponizing tech, which could potentially execute devastating biological, chemical or cyber attacks. It further addresses the possibility of losing control over an advanced system that could supersede its creators.
One proposed solution to slow down AI advancements is to limit the use of chips used for AI. However, as Time points out, this may not be feasible due to physical constraints of the universe.
The report concludes by stating that as AI algorithms continue to improve, more capabilities become available for less total compute power. This trend could make it impractical to mitigate advanced AI proliferation through compute concentrations at all.”