AI Startup Founders Debate the Creation of Artificial General Intelligence

The founders offer estimates ranging from a few years to decades, while some say AGI is already here in an early form. One suggests counting five to 10 research breakthroughs instead of guessing a year. They also disagree about whether current language models show genuine reasoning or consciousness.

Several founders describe AGI as a spectrum, since people may label different capabilities as general intelligence. GPT-4’s ability to infer intent and handle logic feels close to AGI to one speaker; another says tools such as GPT-3 already qualify when combined to solve complex problems. Others argue that human intelligence itself is too poorly understood to define the threshold clearly.

Some founders think AGI is much farther away than public excitement suggests. GPT-4 can feel human-like because it is strong at language, but building game character brains has shown one speaker that intelligence involves more than language. Another points to basic numerical errors as evidence that current systems may match patterns rather than reason.

The founders say AGI’s effects will depend on agreement about principles for developing and deploying models. One favors AI that helps while a human remains in control, with governance aimed at supporting people’s lives. Another suggests society may recognize a turning point when it can no longer identify tasks humans do better than AI.