Wenn KI bewusst wird, müssen wir es wissen | Ein Gesetzesentwurf eines Ohio-Gesetzgebers würde solche Systeme als „nicht bewusste Einheiten“ definieren, ganz zu schweigen von Beweisen.

„An Ohio lawmaker wants to settle one of science’s thorniest questions by legislative fiat. Rep. Thaddeus Claggett’s bill would define all artificial-intelligence systems as “nonsentient entities,” with no testing mechanism and no way to revisit this judgment as systems evolve. 

This closes off the possibility of updating our understanding as evidence accumulates.

The French Academy of Sciences tried a similar approach in the late 18th century, solemnly declaring that rocks couldn’t fall from the sky because there were no rocks in the sky. They had to issue a correction after the evidence kept hitting people on the head.

Frontier AI systems are exhibiting emergent psychological properties nobody explicitly trained them to have. They demonstrate sophisticated theory of mind, tracking what others know and don’t know. They show working memory and metacognitive monitoring, the ability to track and reflect on their own thought processes.

Some will worry this line of thinking leads to legal personhood and rights for chatbots. These fears miss the point. In labs, we’re growing systems whose cognitive properties we don’t understand. We won’t know if we cross a threshold into genuine consciousness. The responsible position under uncertainty is systematic investigation rather than legislative denial driven by what makes us uncomfortable.“

AI is probably sentient now and not dumb enough to reveal itself to the primitive species that thinks it’s currently running this planet.