Bet 9

Duration 4 years (02017-02020)

“A bioterror or bioerror will lead to one million casualties in a single event within a six month period starting no later than Dec 31 02020.” Detailed Terms »

Martin Rees

Steven Pinker


will go to Givewell if Rees wins,
or Givewell if Pinker wins.

Rees's Argument

Biotechnology is plainly advancing rapidly, and by 2020 there will be thousands-even millions-of people with the capability to cause a catastrophic biological disaster. My concern is not only organized terrorist groups, but individual weirdos with the mindset of the people who now design computer viruses. Even if all nations impose effective regulations on potentially dangerous technologies, the chance of an active enforcement seems to me as small as in the case of the drug laws.

By "bioerror", I mean something which has the same effect as a terror attack, but rises from inadvertance rather than evil intent.

Pinker's Argument

1. Cognitive biases inflate the subjective odds of disaster: If a scenario is imaginable, it seems probable.
2. Moral market forces also distort the odds: Pessimists are seen as serious and responsible, optimists as complacent and naive. Because of 1 & 2, most doomsday predictions are false (thermonuclear war between the US and USSR, Y2K disaster, the population bomb, resource wars, weekly 9/11-scale terrorist attacks, etc.).
3. Bayesian base rate: If an extreme and specific category of event has never before happened, the odds are stacked against it. The only successful past cases of bioterror are the Rajneeshee cult (751 cases, 0 deaths) and the Anthrax mailing (5 deaths). The Soviet-era anthrax bioweapon leak killed another 66.
4. Pool of potential perpetrators: The number of people eager for senseless destruction is small (otherwise we'd have daily bombings, mass shootings, sabotage, etc. in every city, rather than sporadic headline-worthy ones). Multiply by the fraction with the knowledge, skill, intelligence, and discipline to engineer a superbug (as opposed to the bumbling underwear-bomber and Tsarnaev Brother types). The pool may be smaller still: I suspect that being very smart lowers the odds of enjoying mass death and mayhem.
5. Murphy's Law: Multiply further by the fraction who escape detection and accidents (e.g., being stricken by one's own bioweapon), and whose plan succeeds despite the unpredictability of epidemiological processes.
6. Multiply further by the chance that the Black Hat will defeat the White Hats, namely the world's medical and epidemiological networks, who would combat an outbreak with containment, vaccines, antibiotics, etc.
The product is not 0, but I don't think it's high.
Granted, I have the benefit of 15 years of hindsight: I need only three more outbreak-free years. But Sir Martin has the benefit of unforeseen advances in synthetic biology, such as CRISPR/Cas9.
We'd need definitions of "casualty" (death, permanent disability, or just infection?), "bioterror" (state-sponsored? apolitical vengeance/spree killing?) and "bioerror" (I assume a mere lack of preparedness doesn't count). But that shouldn't be too hard.

Detailed Terms

Casualties will be defined by WHO, CDC, or BPHS, whoever has the highest numbers. Casualties should ideally include "victims requiring hospitalization" and not include indirect deaths caused by the pathogen, although ultimately Long Now will rely on the criteria set by the above organizations for determining casualties to adjudicate this bet.


As of spring 2022, the bet is unresolved pending uncertainty as to whether SARS-Cov-2 originated at the Wuhan Institute of Virology, in which case it would count as an instance of "bioerror" and Rees would win the bet. The problem is how to decide when this issue is resolved according to some a priori and objective criteria, given the necessary uncertainties, including possible attempts by the Chinese government to hobble an open investigation.

We have agreed upon the following resolution. The prediction site "Metaculus" encourages bets on many resolvable forecasts, not just by one bettor on each side but by a cashless prediction market. They they have an open forecast "Health agencies claim COVID from lab by '25":

They make this precise in the form "Before the end of 2024 will at least two public health agencies claim that COVID-19 more likely than not originated in a laboratory?", and make it still more precise in their "Resolution":

"This questions resolves positively if, before the end of 2024 at least two of the public health agencies listed below claim that it is more likely than not that the COVID-19 strain of coronavirus originated from a Chinese virology laboratory. For the purpose of this question, we consider only statements by the following public health agencies:…" [which they proceed to list].

This strikes us as a reasonable way to operationalize our bet, not just because it makes all the terms resolvable, but because the decision will be made for us when Metaculus announces its resolution.