AI-rewritten: This is a summary of an article from New Scientist, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

---

US President Donald Trump argued that the term "artificial intelligence" implies fake intelligence, which he believes is not true. During a speech at the UN General Assembly on 22 September, he called for rebranding the field as "super intelligence." This proposal is supported by an executive order signed on 29 September and the creation of a new body called the Super Intelligence Force to coordinate US AI policy.

Philosopher Nick Bostrom popularized the term "superintelligence" in his 2014 book, defining it as any intellect that greatly exceeds human cognitive performance in virtually all domains. In contrast, Trump’s definition focuses on the promise and potential of current technologies to cement US dominance. The article notes that definitions of advanced AI have shifted over time, moving from weak AI (simulations) to strong AI (actual cognition), then to narrow AI (specific tasks), and finally to artificial general intelligence or AGI.

AGI is defined by OpenAI as highly autonomous systems that outperform humans at most economically valuable work. However, definitions vary; a leaked 2024 agreement between OpenAI and Microsoft defines it as a system capable of generating $100 billion in profit. Bostrom’s superintelligence requires exceeding the cognitive abilities of every human who ever lived, not just those alive today. The article states that we have not yet achieved superintelligence, though many companies treat AGI as a near-future business goal.

Measuring intelligence remains difficult. While some AI models score between 130 and 150 on IQ tests, this measure is considered flawed because it ignores many aspects of human capability and may be influenced by training data. Human scores can exceed 200 in young children, but these are hard to measure reliably. Computer scientists have proposed alternative tests like ARC-AGI, which considers both capabilities and computational cost. The article concludes that while companies are certain AGI will arrive, it is impossible to say when or if it will happen, noting that definitions and methods for evaluating intelligence are constantly shifting.