Forecasting Revival Dates

When might biostasis patients be revived?

Photo by Kévin JINER on Unsplash

During the dozen years I worked at the Alcor Life Extension Foundation I conducted visitor tours hundreds of times. When we reached the Patient Care Bay, perhaps the most frequent question was: “When do you think they will be revived?”

My first response to this query is to say that I do not know. Much research is needed for that to happen, with technologies yet to be invented. How can anyone pretend to know when it will happen, assuming it ever does? It would a little like asking the Wright Brothers in 1903 in what year the first supersonic jet will be flown. Orville and Wilber’s first flight lasted 12 seconds and covered 120 feet. Even though their fourth flight of that first day extended these numbers to 59 seconds and 852 feet at 6.8 miles per hour, they had practically no basis for making such an extrapolation.

We are in a better position than the Wright brothers when it comes to prognostication. Orville and Wilbur were visionary engineers and innovators but did not spend time pondering how to estimate the trajectory of technological development. Even with vastly more thought devoted to forecasting, we remain largely unable to accurately forecast the date of developments more than a very few years ahead and we consistently fail to forecast the details of future developments. We have a few successful multi-decade but very specific observations such as Moore’s Law but the specifics of how we kept on Moore’s exponential curve have often been surprising.

Sooner with AI?

“I don’t know” is a deeply unsatisfying answer. I often follow up with an equally honest but barely more satisfying answer: Sometime between 50 and 150 years seems plausible, but it is still very much a guess. 50 years seems unlikely given previous rates of technological development. But we are living in the still-early days of a revolution in Artificial Intelligence that might kick us up onto a much steeper curve of technological progress. Certainly, plenty of people hold this expectation.

Personally, while I see much room for current large language model LLM AIs to advance, I have doubts about their ability to do genuinely innovative research in areas such as the biology of aging. That kind of development, and the development of real SAI (super-intelligent AI) will most likely require a different approach. It is hard enough to predict the pace of developments within the current AI paradigm. Forecasting the success of a yet-unproven approach is far more error-prone. I hope to be wrong about this. I hope very badly, since it could make the difference between underdoing legal death and having to be placed into biostasis and living until aging has been conquered.

So, with AI assistance, it seems possible that we will fix the aging problem and figure out the technologies needed to repair and revive human patients from biostasis by 2075. I pick 150 years (or 2175) as the outer edge of my guess for merely intuitive reasons. 150 years is a long time, especially with the now-acknowledged reality of AI. That does not mean I rule out the problem taking longer than that to be solved. Basically, that hundred-year stretch is my confidence interval but, with lower confidence, I grant that revival could come earlier or later.

Sorry. That is the most precise I will get. False precision is not helpful. Making biostasis succeed means planning for the indefinite future.

Bet on it

One way to attempt to figure out plausible dates for revival is provided by prediction markets such as Manifold and Metaculus. For instance, for the question “Before 1 January 2050, will any human cryonically preserved for at least 1 year be successfully revived?” the Metaculus market gives only a 2% chance. (However, the proposition “What will the earliest preservation date of any resuscitated cryonics patient be” gives a consensus estimate of 2061 but with only 36 market participants.)

Essentially the same question at Manifold – “Will a human preserved through cryonics be successfully restored by end of 2050?” – puts the odds at 20%. (Similar bets put the odds between 15% and 25%.) The first of the two is the outcome of 269 forecasters whereas the second is the result form only 60 bettors. The outcome of the deeper market seems far more plausible to me.

Manifold has two different markets asking the same thing with an insignificant difference in wording: Will a human preserved through cryonics be successfully restored by the end of 2075? One market pegs the probability at 39%, the other at 35%. Although the proximity of the two estimates is superficially encouraging, the number of bettors is only 19 in one case and 46 in the other.

Both Metaculus and Manifold also feature bets on the odds that cryopreserved people will be revived by the end of 2100. The 69 forecasters in the Metaculus market put the odds at 12%. Manifold again has two distinct bets on the same thing. One finds odds of 52% at one and the other at 34%.

Finally, a market on Metaculus asks: “When will the first person that has been cryopreserved for more than 1 year be resuscitated or uploaded, if this occurs by year 2200?” 64 forecasters result in an average answer of 2096. This seems wildly at odds with another market that asks: “If you die today and get cryonically frozen, will you ‘wake up’?” This gives odds of only 6%.

These results are interesting but I would not take them very seriously. Most of the markets have fewer than one hundred participants and the largest still has only 437. The fewer the number of bettors, the less reliable the outcome is likely to be. Also, due to restrictive laws, instead of real money being at stake, bettors use tokens. The whole point of decision or prediction markets is to put real money at risk, incentivizing careful thinking and adjusting your confidence to your level of knowledge and thoughtfulness about the issue.

Superforecasters

What about superforecasters? These are individuals who have repeatedly performed near the top across multiple bets. They may have hard to define skills that explain their repeated success. So far as I know, revival from biostasis is not yet a topic on which superforecasters have bet. Although superforecasters seem to be real, two factors make me doubt their reliability on this topic.

One is the general concern that those successful in forecasting over multiple bets might just have been lucky. We know there are fund managers who manage to outperform the market for years in a row but they practically always eventually fall behind. The more bets on which a superforecaster excels, the more I would trust that their ability is genuine.

But there is another problem. Superforecasters have placed bets on propositions with a fairly short time horizon. These are often no more than one year. At most they look ahead just a handful of years. When we are wondering about the application of yet to be invented technologies that are decades and perhaps a century or more away, even superforecasters will struggle to be accurate.

No simple answer

Here is something I think we can be confident about: Biostasis patients will not all be revived at the same time. I do not mean on the same day or month. I expect that years and possibly decades will separate revivals. Two factors explain this view, both reflecting the quality of preservations.

First, other things being equal, patients preserved in earlier years will be more damaged than patients preserved later by more advanced technologies and procedures. The very earliest cases were essentially “straight freezes.” Although James Bedford, the first cryonaut, theoretically was perfused with crude cryoprotectants, in practice they did not get circulated and had no effect. Other cases were straight freezes because too long had passed before a patient could be reached and perfusion was not possible.

First in, last out

Throughout the 1980s and 1990s, increasingly high concentrations of cryoprotectant were used, resulting in less ice formation. If cryonics organizations such as Alcor and Cryonics Institute were to do CT scans of all their patients, we would learn a lot about the degree of ice formation. This is entirely feasible for an organization that has its own CT scanner.

Early this century, vitrification was introduced, further reducing ice formation. Over time, response to a member in critical condition may become more rapid, although that has so far been challenging. In the case of chemical preservation, although this is a recent development, later cases might take longer until revival is possible since all the cross-links have to be reversed. These cases might take longer than good pure cryopreservation cases. That will depend on the types of repair capabilities we develop in the future.

We can aim for a reduction in the number of cases of unattended clinical death through two means: better monitoring and alert systems and spreading of jurisdictions allowing you to choose when to undergo legal death. It seems likely that the least damaged patients will be repaired and revived first. Whether the gap between earlier and later patients will be months or years or decades, I would not venture to guess.

Second, the spectrum of damage does not align directly with the year of preservation. Sometimes earlier cases will have better structural preservation than some later cases because of individual circumstances. A 1990 case with minimal delay and no complications in transport or surgery may have resulted in a better preservation than a 2020 case with serious delays in beginning procedures.

Delay is not the only factor that can worsen the quality of preservation. The cause of legal death also matters. Someone who dies of cancer but whose cardiovascular system is in good condition may perfuse faster and more thoroughly than someone who died of heart disease or brain cancer or an aneurysm. (The simple idea of “first in, last out” is not a rule, but only an approximation and generalization.)

We can each take a guess at how long it will be before biostasis patients start being revived. We will differ widely in our estimates. It doesn’t matter. Whether it is 20 years or 50 or 100, we need to support the same strategic push – organizational resilience and longevity.

I wrote about the related issue of estimating the probability that cryonics will work for you.