This post touches on a controversial topic. I invite readers to approach it with a scout mindset and a strong dose of skepticism. It is my hope that the topic can be explored in a spirit of good faith and collaborative truth-seeking, its controversial nature and unfortunate history notwithstanding.

I believe the topic of UFOs is more important than is commonly recognized. I think people tend to dismiss the topic, including the extraterrestrial hypothesis for the origin of (some) UFOs, for what I think are ultimately not good reasons.

So it seems to me that it’s worth engaging with the reasons people give, and to point out where my reasoning diverges from the thoughts of those who dismiss the extraterrestrial hypothesis. I suspect that Eliezer Yudkowsky’s reasoning is representative of the thoughts of many people who dismiss the topic, and hence what follows is a point-by-point reply to Yudkowsky on UFOs.

To clarify, my main point here is that the reasons given by Yudkowsky do not justify the low probability that he assigns to the extraterrestrial hypothesis, let alone his apparent dismissive stance toward it (e.g. saying that “you can approximately leave that possibility out of your thinking”). My aim is not to argue that the extraterrestrial hypothesis is the most likely hypothesis concerning the origin of (at least some) UFOs, but I am arguing that we are not justified in leaving the possibility out of our thinking.^[1]

Yudkowsky (his words are quoted like this in the rest of this post):

My model of the world confidently says no to alien UFOs: Their technology would not be such that, having arrived here across interstellar distances and then remained hidden, they'd need to fly around in large visible vehicles. It is definitely the case, given the physics we already know, that the aliens can do whatever surveillance they want using far tinier devices; eg, covalent-bond-strong, micron-sized robots, like bacteria but not with proteins held together by static cling.

These statements seem to make some fairly strong assumptions about what hypothetical alien UFOs “need” to do and what their motives would be. It also makes some strong assumptions about the capabilities of hypothetical technologies.

In particular, an implicit assumption in the statements above seems to be that hypothetical extraterrestrial UFOs would want to stay as small and hidden as possible. But is that a motive that we can confidently assume? For example, isn’t it conceivable that their motives could imply being visible in some cases? Note that this possibility is not limited to the specific signaling story proposed by Robin Hanson about gradually informing humans, or post-human descendants, that they are more powerful than us (see e.g. Hanson, 2021a; 2021b; 2021c). Some alternative possibilities might include the following (this is by no means an exhaustive list):

• Experimental interventions to test reactions, as when humans explore the reactions of non-human animals in response to certain interventions (cf. Goumas et al., 2020).
• Attempts to slowly habituate humans to a non-human presence (cf.  human habituation interventions among non-human primates).
• Attempts to prevent nuclear catastrophes that might spoil efforts to gain information about the developmental trajectories of emerging civilizations. (Even if outcomes with nuclear catastrophes would provide some information, outcomes without total nuclear disaster might still generally be more informative.)

To be clear, I am not claiming that any one of these motives deserves a high prior. But it’s worth being clear that the assumption that hypothetical alien UFOs would want to stay as small and hidden as possible is not a necessary or justified one. There are other plausible possibilities, conditional on an extraterrestrial presence. As Robin Hanson rightly notes with respect to confident claims about alien motives: “If we often have trouble explaining the behaviors of human societies and individuals, I don’t think we should feel very confident in predicting detailed behaviors of a completely alien civilization.” (Related points and memes are sometimes brought up in AI risk discussions, e.g.: “BREAKING: 100 IQ Man Confidently Declares What a 1 Billion IQ AI Will Do.”)

Apart from the implicit assumptions about motives, it seems to me like a very strong claim to say that hypothetical aliens would be able to do “whatever surveillance they want using far tinier devices”. How can we be confident that there are not at least some types of useful surveillance, or computational processes, or maneuverability (etc.) that in fact require — or are at least aided by — visible-sized objects? If greater size means greater computational power (or greater maneuverability or the like), why should the worthwhile gains in added computational power (or other abilities) necessarily always fall below the threshold of visibility?

Superintelligence is possible - it is just flat wrong that a human is as smart as any physical system can get - and you'd expect something crossing interstellar distances to be long since superintelligent. If they wanted to stay hidden, they'd stay hidden successfully. If aliens wanted to help Earth and not hide, humans would not be dying of cancer. If aliens didn't want to help, nor to hide, the aliens would have harvested the Solar System for matter and energy.

This again seems to make some strong implicit assumptions. Yudkowsky appears to present these three possibilities as though they exhaustively cover all possibilities, or at least all plausible ones. Yet that is hardly the case.

“If they wanted to stay hidden, they'd stay hidden successfully.” As hinted above, an alternative possibility is that hypothetical extraterrestrials might generally want to stay hidden, while still being willing to visibly interfere in cases where this is deemed worthwhile from their perspective (e.g. to interfere with nuclear weapons). Such minimally flexible motives do not seem far-fetched conditional on an extraterrestrial presence. Indeed, one could argue that it would be less plausible for them to have perfectly rigid and categorical motives (e.g. “always stay completely hidden no matter what”).

“If aliens didn't want to help, nor to hide, the aliens would have harvested the Solar System for matter and energy.” I don’t think that conclusion follows. As I’ve argued elsewhere, there are various reasons why advanced extraterrestrials might refrain from being grabby and harvesting the Solar System (along with all other stars within their reach).

For example, they might wish to stay hidden from other cosmic civilizations (which does not require them to stay completely hidden on Earth), they might wish to study emerging civilizations like ours (which also does not require them to always stay completely hidden from us), or they might reduce the risk of internal drift and coordination failure by staying relatively small and “quiet” rather than being maximally large and grabby. And, of course, these motives are not mutually exclusive; a hypothetical alien civilization could in principle possess all these motives simultaneously.

Alien psychology would not be such as to (per Robin Hanson) stay mysterious and uncertain yet occasionally visible, in order to maximize their status. If aliens wanted to shape our opinions and had no rules against causal impacts on us, they could literally just rewrite our brains. If they wanted to leave us alone instead, we'd apparently be alone. If aliens wanted to accomplish Hanson's exact hypothetical end of gaining status in our eyes, there'd be much better strategies than generating disreputable UFO stories.

Again, it is worth being clear that Hanson’s conjectured explanation of occasional visibility (conditional on alien UFOs) is but one among a much broader set of such conjectures, and not necessarily the most plausible one.^[2]

However, in fairness to Hanson’s signaling conjecture, his story is not that hypothetical alien UFOs would seek to maximize status per se. Rather, his story is that they would seek to gradually communicate their greater status and power within some specific constraints, such as not causing excessive panic, dislike, and hostility.

I have enormously wide uncertainty about the distribution of true alien psychologies, or the spacefaring agencies that grow out of them. But it's uncertainty over a metric where--when we look back down at Earth and what those psychologies would mean to us--the supervast majority of probable alien intellects, would not come here across interstellar distances, quietly and hiding on arrival, and then occasionally fly around in giant visible vehicles.

When trying to reason about the “supervast majority of probable alien intellects”, it is critical to be clear about what we are conditioning on, and whether we are taking observation selection effects into account. In particular, are we conditioning on our own existence here on Earth? After all, there may be a vast difference between our expectations about what near-Earth aliens would be like and do in a universe where our existence is not assumed versus a universe in which our existence on Earth is an observed starting point.

This also relates to Yudkowsky’s earlier claim that aliens would harvest the Solar System for matter and energy: such grabby aliens seem a priori likely, yet if we are talking about hypothetical near-Earth aliens and we are conditioning on our own existence, grabby near-Earth aliens are effectively ruled out.

More generally, our assessment of the distribution of “probable alien intellects” changes radically when we take our existence into account and focus specifically on hypothetical near-Earth aliens. I have tried to illustrate this point in the toy figure below, where the blue is the large swath of possible scenarios that are incompatible with our existence (and hence these scenarios get ruled out when we condition on our existence).
 

As we can see, possibilities that look highly improbable in the non-conditioned case may suddenly become quite likely. While “quietly hiding yet occasionally visible” might still not be the most likely a priori expectation about how hypothetical near-Earth aliens would behave, this anthropic update nevertheless does make it significantly more likely. (And the prospect of such hypothetical near aliens arguably becomes more likely still in light of the broader anthropic argument found here.)

These points about observation selection effects, as simple as they are, seem to me broadly neglected in attempts to reason about hypothetical near-Earth aliens.

I have enormously wide uncertainty over the possible range of alien technologies. But I can use current knowledge of physics and chemistry, and the advance analyses that others have done of what technological possibilities those imply, to put a lower bound under alien technology that's comfortably above "needs to use giant flying vehicles for travel or surveillance".

This again seems to make strong assumptions about motives and about supposed (non-)gains from added size, as mentioned earlier.

I don't need to know exactly what aliens are trying to do, to know that only a few and unlikely goals would imply a best possible strategy of flying around in sightable UFOs while staying otherwise hidden.

This does not seem accurate to me (though granted, the meaning of “a few and unlikely” is a bit vague). As mentioned earlier, there are various strategies and intervention patterns that would involve generally staying hidden while permitting visible interventions on occasion. And while such strategies might not seem like the most likely strategies a priori, they also do not seem far-fetched (conditional on both our existence and the existence of near-Earth aliens). This is especially true if we consider the full class of motives that would involve such a pattern of occasional visibility, including “unknown unknown” such motives, as opposed to just considering the first few and most salient potential motives that come to mind.

(If I was somehow told with certainty that I would lose my bet, but not told the details, I would cry. With happiness and relief and hope. My present worldview is a grim one; anything that shatters my worldview is good news in expectation if not in certainty.)

This parenthetical remark is not central to Yudkowsky’s reasoning about the plausibility of extraterrestrial UFOs (except perhaps insofar as it indicates that he is not moved to reject its plausibility based on wishful thinking).

That said, I find it worth noting that I don’t really understand the remark. In particular, I don’t understand why “anything that shatters [his] worldview is good news in expectation if not in certainty”. Is his current view that we are most likely headed for an s-risk in expectation? Or worse, that we are most likely headed for one of the very worst s-risks conceivable? If not, it seems like there is a lot of room for worldview-shattering information to make our outlooks much grimmer than they are.

If Yudkowsky thinks we are headed for a “much worse than death” s-risk outcome in expectation (relative to his own standards of “much worse than death”), then that would seem to have significant implications for his top recommendations for altruistic priorities (e.g. perhaps making “separation from hyperexistential risk” a top priority).

If he does not think we are headed for a “much worse than death” outcome in expectation, I do not understand why being told that UFOs are extraterrestrials without being “told the details” would cause him to “cry with happiness and relief and hope” and why it would almost certainly be good news in expectation. After all, even if some UFOs have extraterrestrial origins, tail-end horror scenarios could seemingly still come about through future inter-galactic conflicts, and perhaps large-scale strategies of staying hidden even make such scenarios more likely rather than less (e.g. such strategies might be considered weak evidence for a state of hostility and conflict at the largest scales of cosmopolitics). At least it is not clear that it would make such horror outcomes seem substantially less likely.

UFOs aren't aliens. You can approximately leave that possibility out of your thinking. I've studied some of what one needs to study, to know that a bit more surely. There's vast room above human intellect and human technology, and what that implies is this: if there are hidden aliens, they're successfully hiding. Like, actually successfully, without the whole UFOs thing.

This again assumes that remaining completely hidden must be the chief overriding motive of hypothetical near aliens, which, again, is hardly a warranted assumption. There are other possibilities that involve less rigid and perhaps more plausible motives.

Additionally, it is worth considering the plausibility of hypothetical near aliens who are indeed successfully hiding. In my view, the possibility of such successfully hiding near aliens is a priori more likely than we tend to assume, perhaps even deserving a prior of around 1 to 10 percent or higher (some support for that seemingly crazy view is found here, here, and here). And if scenarios involving “successfully hiding” near-Earth aliens are not extremely unlikely a priori (conditional on our existence) — which it seems to me they are not — then arguably scenarios involving “generally hiding yet occasionally visible” near-Earth aliens are not extremely unlikely either a priori. After all, the latter class of scenarios is quite similar to the former, save for the slightly less categorically rigid motives.

Finally, the points discussed so far only concern our priors; they have not touched on the reported UFO evidence itself, and how this might update our views (or fail to do so). Yet a proper analysis of the topic would also need to take that reported evidence into account. While it is quite debatable what we should make of that reported evidence, my take is that it does overall increase the probability of (some) UFOs having extraterrestrial origins.

For example, I think the 2004 Nimitz incident — involving people like David Fravor, Alex Dietrich, Chad Underwood, and Gary Voorhis — is an example of an incident that should at least somewhat update our beliefs in that direction, given the exotic nature of the reported capabilities and the apparent credibility of the witnesses. Some other prominent incidents where similarly exotic abilities were reported include:

• Foo fighters observed during World War II (e.g. 1, 2, 3).
• The 1957 Milton Torres incident.
• The 1976 Tehran UFO incident involving pilot Parviz Jafari among others.
• The 1980 Peruvian Air Force incident with pilot Oscar Santa Maria Huerta.
• The 1986 Brazilian UFO incident.
• The 2019 USS Omaha incident, from which there is also footage of the associated radar readings; the objects were tracked by two different radar systems as well as a FLIR camera.

Any substantive take on UFOs would seemingly need to account for such reports and sensor recordings (e.g. with some deflationary account), and not just quietly ignore them. In other words, the reported evidence needs to be incorporated into our world models somehow, even if just as an elaborate hoax, which would itself be quite noteworthy. Yudkowsky does not seem to provide any deflationary explanation of the reported evidence, despite its apparent relevance.

While all the considerations listed above, including the reported evidence, may still leave us quite far from finding it likely that any UFOs have extraterrestrial origins, I submit that we are not justified in approximately leaving the possibility out of our thinking.

1. ^{^}

   I have argued similarly in “From AI to distant probes”, “Silent cosmic rulers”, and “Can we confidently dismiss the existence of near aliens? Probabilities and implications”.

2. ^{^}

   Likewise, Hanson’s proposal that such hypothetical aliens would likely be intra-galactic panspermia siblings does not seem to me the most likely conjecture. Various considerations suggest that inter-galactic origins would be significantly more likely.