AI or ‘super intelligence’? Trump isn’t the only one who is confused
US President Donald Trump has declared that artificial intelligence should now be called “super intelligence”. Columnist Jacob Aron explains where that term came from and explores whether its many historical meanings apply today
Artificial intelligence (AI) has been a subject of debate since its inception, with much focus on whether machines can think like humans. Recently, President Donald Trump has taken a stance, advocating for the term "super intelligence" to replace "artificial intelligence." This shift in terminology aims to rebrand AI, but what exactly does it entail?
Trump's move isn't unique; Meta CEO Mark Zuckerberg has also expressed views that "superintelligence" is nearing realization. These calls for rebranding stem from a desire to emphasize the potential and rapid advancements of AI technologies.
To differentiate these terms, we must look back to the 1940s and 50s when AI was first conceptualized. The term "superintelligence," originally popularized by philosopher Nick Bostrom, was defined as an intellect exceeding human cognitive performance in virtually all domains. This definition goes beyond current AI systems and suggests a level of cognitive capability that surpasses all human minds.
Historically, AI researchers distinguished between weak and strong AI. Weak AI refers to computer simulations that mimic human minds but lack actual cognition. Strong AI, on the other hand, describes actual cognition emerging from software. Searle's notion of strong AI, or "actual understanding," has been dismissed as unrealistic. Instead, weak AI, now known as narrow AI, has become a reality, powering various applications from spam filters to driverless cars.
General AI, a system capable of human-level performance across all tasks, has given way to artificial general intelligence (AGI). This concept, popularized by Shane Legg, aims to create an AI capable of transferring learning across domains, much like humans. Most major AI companies, including DeepMind and OpenAI, are striving to achieve AGI.
Despite the seemingly clear-cut definitions, the reality of AGI remains elusive. The standard measure of intelligence, the IQ, is insufficient for defining AGI. AGI, as Bostrom defines it, must surpass human cognitive capabilities in virtually all areas of interest. Currently, we have not achieved this level of cognitive performance in AI.
Written by urgent.news from New Scientist's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
