Dr. David Denkenberger co-founded and is a director at the Alliance to Feed the Earth in Disasters (ALLFED.info) and donates half his income to it. He received his B.S. from Penn State in Engineering Science, his masters from Princeton in Mechanical and Aerospace Engineering, and his Ph.D. from the University of Colorado at Boulder in the Building Systems Program. His dissertation was on an expanded microchannel heat exchanger, which he patented. He is an associate professor at the University of Canterbury in mechanical engineering. He received the National Merit Scholarship, the Barry Goldwater Scholarship, the National Science Foundation Graduate Research Fellowship, is a Penn State distinguished alumnus, and is a registered professional engineer. He has authored or co-authored 134 publications (>4400 citations, >50,000 downloads, h-index = 34, second most prolific author in the existential/global catastrophic risk field), including the book Feeding Everyone no Matter What: Managing Food Security after Global Catastrophe. His food work has been featured in over 25 countries, over 300 articles, including Science, Vox, Business Insider, Wikipedia, Deutchlandfunk (German Public Radio online), Discovery Channel Online News, Gizmodo, Phys.org, and Science Daily. He has given interviews on 80,000 Hours podcast (here and here) and Estonian Public Radio, WGBH Radio, Boston, and WCAI Radio on Cape Cod, USA. He has given over 80 external presentations, including ones on food at Harvard University, MIT, Princeton University, University of Cambridge, University of Oxford, Cornell University, University of California Los Angeles, Lawrence Berkeley National Lab, Sandia National Labs, Los Alamos National Lab, Imperial College, and University College London.
denkenberger
Yes, here is a fault tree analysis of nuclear war. And here is one for AI.
I have done some work on refuges. However, I am most interested in saving nearly everyone and preventing the loss of civilization. This turns out to be cost effective even if one only cares about the present generation. I am currently working on cost effectiveness from a far future perspective.
Thank you, Jennifer, for the introduction. Some more background on me: I have read the sequences and the foom debate. In 2011, I tried to do cost-effectiveness scoping for all causes inspired by Yudkowsky’s scope and neglectedness framework (the scope, neglectedness, and tractability framework had not yet been invented). I am concerned about AI risk, and have been working with Alexey Turchin. I am primarily motivated by existential risk reduction. If we lose anthropological civilization (defined by cooperation outside the clan), we may not recover for the following reasons:
• Easily accessible fossil fuels and minerals exhausted
• Don’t have the stable climate of last 10,000 years
• Lose trust or IQ permanently
• Endemic disease prevents high population density
• Permanent loss of grains precludes high population density
Not recovering is a form of existential risk (not realizing our potential), and we might actually go extinct because of a supervolcano or asteroid after losing civilization. Because getting prepared (research and development of non-sunlight dependent foods such as mushrooms and natural gas digesting bacteria, and planning) is so cost-effective for the present generation, I think it will be a very cost effective way of reducing existential risk.
Here is an analysis of nutrition of a variety of alternate foods. Leaf protein concentrate is actually more promising than leaf tea. No one has tried a diet of only alternate foods—that would be a good experiment to run. With a variety, the weight is not too high. Yes, we are hoping that some of these ideas will be viable present day, because then we can get early investment.
In the case of the sun being blocked by comet impact, super volcanic eruption, or full-scale nuclear war with the burning of cities, there would be local devastation, but the majority of global industry would function. Most of our energy is not dependent on the sun. So it turns out the biggest problem is food, and arable land would not be valuable. Extracting human edible calories from leaves would only work for those leaves that were green when the catastrophe happened. They could provide about half a year of food for everyone, or more realistically 10% of food for five years.
I also work on the catastrophes that could disrupt electricity globally, such as an extreme solar storm, multiple high-altitude detonations of nuclear weapons around the world creating electromagnetic pulses (EMPs), and a super computer virus. Since nearly everything is dependent on electricity, this means we lose fossil fuel production and industry. In this case, energy is critical, but there are ways of dealing with it. So the food problem still turns out to be quite important (the sun is still shining, but we don’t have fossil fuel based tractors, fertilizers and pesticides), though there are solutions for that.
Sorry for my voice recognition software error-I now have fixed it. It turns out that if you want to store enough food to feed 7 billion people for five years, it would cost tens of trillions of dollars. What I am proposing is spending tens of millions of dollars for targeted research and development and planning. The idea is that we would not have to spend a lot of money on emergency use only machinery. I use the example of the United States before World War II-it hardly produced any airplanes. But once it entered World War II, it retrofitted the car manufacturing plants to produce airplanes very quickly. I am targeting food sources that could be ramped up very quickly with not very much preparation (in months, see graph here. The easiest killed leaves (for human food) to collect would be agricultural residues with existing farm equipment. For leaves shed naturally (leaf litter), we could release cows into forests. I also analyze logistics in the book, and it would be technically feasible. Note that these catastrophes would only destroy regional infrastructure. However, the big assumption is that there would still be international cooperation. Without these alternative food sources, most people would die, so it would likely be in the best interest of many countries to initiate conflicts. However, if countries knew that they could actually benefit by cooperating and trading and ideally feed everyone, cooperation is more likely (though of course not guaranteed). So you could think of this as a peace project.
Grains are all from the same family-grass. It is conceivable that a malicious actor could design a pathogen(s) that kills all grains. Or maybe it would become an endemic disease that would decrease the vigor of the plants permanently. I’m not arguing that any of these non-recovery scenarios are too likely. However, if together they represent 10% probability, and if there is a 10% probability of the sun being blocked this century, and a 10% probability of civilization collapsing if the sun is blocked, this would be a one in 1000 chance of an existential catastrophe from agricultural catastrophes this century. This is worth some effort to reduce.
This could potentially help many decades in the future. But it would need to be an order of magnitude or more reduction in energy costs for this to produce a lot of food. And I am particularly concerned about one of these catastrophes happening in the next decade.
I like your succinct way of restating the case for spending some money on catastrophes other than AI.
It is possible that a loss of industry could be beneficial in the long term. One can adjust the moral hazard parameter to take into account this possibility. However, it does subject us to more natural risk like supervolcanic eruptions and asteroid/comet impacts. And if we actually lost anthropological civilization, we would not be doing any AI safety work. Even just losing industry for a long time I think would make most AI safety work not feasible, but I am interested in your thoughts. Without industry, we would not be able to afford nearly as many researchers. And they would just be doing math on paper.
Note that that statistic is how long people have been in their current job, not how long they will stay in their current job total. If everyone stayed in their jobs for 40 years, and you did a survey of how long people have been in their job, the median will come out to 20 years. I have not found hard data for the number we actually want, but this indicates that the median time that people stay in their jobs is about eight years, though it would be slightly shorter for younger people.
I have estimated global vitamin D3 production to be a few tons per year, so at US RDA of 600 UI, we could only provide about 3% of the global population. At your suggestion of 5000 UI/day, it would only be about 0.3% of people. This is why I looked into quickly scaling up vitamin D production. The most promising appeared to be seaweed, but we could not get anyone excited about doing it before there was a shortage. Fortunately, just mega dosing of those testing positive appears to be within our global D3 production capability at current infection rate. However, if we let it run through the population, I don’t think we would have sufficient supplies at current production.
Another interesting piece of evidence is a study on homeless people in Boston (who would likely not be vitamin D deficient because more outdoor time):
“100% of 147 COVID-19 positive subjects were asymptomatic.”
Source, which doesn’t really make the connection:
Baggett, T. P., Keyes, H., Sporn, N. & Gaeta, J. M. COVID-19 outbreak at a large homeless shelter in
Boston: Implications for universal testing. medRxiv 2020.04.12.20059618 (2020)
doi:10.1101/2020.04.12.20059618.
Maybe the most effective thing would be if there were a vitamin D futures market, to bid up the price to incite more production of it. But at the individual level, I think it makes sense to stock up to increase the price a little bit. If you don’t end up needing it, you could always give/sell it to those who do later. The one I bought is good for 1.5 years.
That was an exciting graph! However, the labeling would be more consistent if it were steam engines, piston engines, and turbine engines OR stationary, ship/barge, train, automobile, and aircraft (I assume you mean airplanes and helicopters and you excluded rockets).
this gives a paltry annual return on investment of 0.075%
which seems large until we note that it implies an annualized rate of return of 0.08%; far more than our estimate above, but a tiny rate of return.
Am I comparing the right numbers? It doesn’t seem like far more to me.
Great work!
The thing is, at some point it does mean the virus is unstoppable, in the sense that no reasonable or worthwhile attempt to stop it has any chance of success, outside of at most protecting particular vulnerable groups and doing mitigation. If the baseline transmission is higher than Delta and it’s mostly ignoring vaccinations, what is your plan exactly? Lock down much harder than we did in 2020? Close the grocery stores?
Naïve calculation: if surgical masks blocked 75% of aerosols going out and coming in, and if all transmission were through aerosols, and if people wore these masks all the time, this could theoretically overcome R0 = 16 (because the transmission would be reduced a factor of four in both steps). Of course it will not work out this well, but then there are many other measures that can be taken in addition. So I think it would be technically feasible to keep it under control. But if it were to run quickly through most of the population, omicron would have to be much less severe than other variants to not overwhelm hospitals (without massive scale up of treatments).
It’s hard to pin down a threshold of a specific time of exposure because it depends on the minimum infectious dose, which varies widely among people, at least for lots of diseases. Also, the rate of shedding varies widely based on the progression of the disease, whether the person is talking, how far away the person is, etc. Furthermore, the HVAC system causes additional variation. So I think when you add all these uncertainties, a 16 times reduction in emission/inhalation would correspond to very roughly a 16 times reduction in infection, but I would be very interested to see if someone has run the math on this.
The substantive complaint was that they [ALLFED] did an invalid calculation when calculating the annual probability of nuclear war. They did a survey to establish a range of probabilities, then they averaged them. One could argue about what kinds of ‘average them’ moves work for the first year, but over time the lack of a nuclear war is Bayesian evidence in favor of lower probabilities and against higher probabilities. It’s incorrect to not adjust for this, and the complaint was not merely the error, but that the error was pointed out and not corrected.
Tl; dr: ALLFED appreciates the feedback. We disagree that it was a mistake—there were smart people on both sides of this issue. Good epistemics are very important to ALLFED.
Full version:
Zvi is investigating the issue. I won’t name names, but suffice it to say, there were smart people disagreeing on this issue. We have been citing the fault tree analysis of the probability of nuclear war, which we think is the most rigorous study because it uses actual data. Someone did suggest that we should update the probability estimate based on the fact that nuclear war has not yet occurred (excluding World War II). Taking a look at the paper itself (see the top of page 9 and equation (5) on that page), for conditional probabilities of occurrence for which effectively zero historical occurrences have been observed out of n total cases when it could have occurred, the probability in the model was updated according to a Bayesian posterior distribution with a uniform prior and binomial likelihood function. Historical occurrences updated in this way were A) the conditional probability that Threat Assessment Conference (TAC)-level attack indicators will be promoted to a Missile Attack Conference (MAC), and (B) the conditional probability of leaders’ decision to launch in response to mistaken MAC-level indicators of being under attack. Based on this methodology, it would be double-counting to update their final distribution further based on the historical absence of accidental nuclear launches over the last 76 years.
But what we do agree on is that if one starts with a high prior, one should update. And that’s what was done by one of our coauthors for his model of the probability of nuclear war, and he got similar results to the fault tree analysis. Furthermore, the fault tree analysis was only for inadvertent nuclear war (one side thinking they are being attacked, and then “retaliating”). However, there are other mechanisms for nuclear war, including intentional attack, and accidental detonation of a nuclear weapon and escalation from there. Furthermore, though many people consider nuclear winter only possible for a US-Russia nuclear war, now that China has a greater purchasing power parity than the US, we think there is comparable combustible material there. So the possibility of US-China nuclear war or Russia-China nuclear war further increases probabilities. So even if there should be some updating downward on the inadvertent US-Russia nuclear war, I think the fault tree analysis still provides a reasonable estimate. I also explained this on my first 80k podcast.
Also, we say in the paper, “Considering uncertainty represented within our models, our result is robust: reverting the conclusion required simultaneously changing the 3-5 most important parameters to the pessimistic ends.” So as Zvi has recognized, even if one thinks the probability of nuclear war should be significantly lower, the overall conclusion doesn’t change. We have encouraged people to put their own estimates in.
Again, we really appreciate the feedback. Good epistemics are very important to us. We are trying to reach the truth. We want to have maximum positive impact on the world, so that’s why we spend a significant amount of time on prioritization.
*If the rapid test had some probability of success, like 70%, then if you took two test you might figure 1-(1-.7)^2 = 1-.3^2 = 1-.09 = 01% you have covid. But are the rapid tests independent?**
You need to start with a prior for this calculation. This paper also discusses independence of tests. And I think you meant to write 91%.
I am happy to do an AMA.