AGI Will Not Spontaneously Generate Superintelligence
It won’t have to!
How long will it take for today’s AI technology to reach human-level intelligence? Known as Artificial General Intelligence, AGI has been the dream AI developers since the beginning. As far back as 1965, scientists worried that once AGI was achieved, it might spontaneously generate Superintelligence. Spontaneous Generation, the creation of life from non-living material, was hypothesized by Aristotle in 350 BC¹ but disproven by Pasteur in the late 19th century.²
Today’s AI-aware community believes that several more years of hard work and several tens of billions of dollars will drive LLMs (Large Language Models) to reach AGI level intelligence.
Renowned scholars such as Nick Bostrom (2014³) and Max Tegmark (2017⁴) have expressed fears that as soon as AGI is created it will experience a “fast breakout” to Superintelligence. The speed of this breakout — potentially seconds, minutes, or hours — is frightening because humans would not be able to mount a defense within that timeframe if it went in a dangerous direction.
Nine years after Bostrom’s publication and six years after Tegmark’s, AI evolution isn’t following this predicted sequence. AIs have already achieved superhuman proficiency in language understanding and translation, game playing…