As a case in point, the self-replication of misinformation on social media networks has become a substantial physical risk in the early 21st century causing not only swings in elections, but riots, take overs, swings in the stock market (GameStop short squeeze January 2021), and mob killings. It is incredibly difficult to create risk assessments for these sorts of future harms.

In biology, we see major damage to a wide variety of species as the result of uncontrolled self-replication. We call it cancer.

We also see programmed processes in biological settings including apoptosis and necrosis as means of avoiding major harms. What might these look like with respect to artificial intelligence?