Criticism of the scientific basis of percentage estimates of existential AI risks
Scientist Heidy Khlaaf from AI Now Institute criticizes percentage estimates of existential AI risks (e.g., “10% chance of humanity being wiped out”) as unfalsifiable and scientifically unfounded due to the lack of concrete catastrophe scenarios.
Heidy Khlaaf, chief scientist at the research institution AI Now Institute, is among the experts who question the meaningfulness of assigning percentage probabilities to existential scenarios involving artificial intelligence—for example, claims of a ten-percent chance of humanity being wiped out. According to her, such claims are “neither falsifiable nor verifiable, and therefore inherently unscientific.” In her view, the problem is compounded by the lack of concrete examples of how “superintelligent” AI could actually cause a catastrophe.
A source familiar with the thinking at Anthropic said that “the exact odds of any particular outcome are probably unknowable.” Khlaaf adds that if there is no substantiated evidence for such estimates, this means that the figures given for “extinction” also have no scientific basis. Critics argue in this sense that such figures borrow the authority of science without being backed by corresponding scientific work.
The source text is part of a broader comparison of different views on AI risks, which also included other extreme claims (for example, comparing AI to nuclear weapons or calling the entire topic a “psyop”); this summary does not cover those other claims because they do not concern the specific criticism of the scientific validity of the estimates. Details can be found in the source article.
Why it matters
The debate on AI safety often relies on specific risk percentages that the public and policymakers perceive as scientifically grounded. Critics, however, say these are estimates without a verifiable basis, which is relevant to anyone who uses such figures as an argument in discussions about AI regulation or development.
Two audiences, two different impacts
What this means
For individuals
When reading claims such as “X% chance of humanity being wiped out by artificial intelligence,” critics say it is appropriate to take such figures with a grain of salt because they can be neither verified nor falsified.
More practical updates →For a business
Companies working with numerical estimates of existential AI risks (for example, in statements associated with Anthropic) face criticism that such estimates lack a scientific basis, which weakens their weight in regulatory and public debate on AI safety.
Risks and compliance More business impacts →Check the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.