Adam Ford & I on Great Filter

Adam Ford interviewed me again, this time on the Great Filter:

We have three main sources of info on existential risks (xrisks):

  1. Inside View Analysis – where we try to use our best theories to reason about particular causal processes.
  2. Earth Track Records – the empirical distribution of related events observed so far on Earth.
  3. The Great Filter – inferences from the fact that the universe looks dead everywhere but here.

These sources are roughly equally informative. #2 suggests xrisks are low, even if high enough to deserve much effort to prevent them. I’d say that most variations on #1 suggest the same. However, #3 suggests xrisks could be very high, which should encourage more xrisk-mitigation efforts.

Ironically most xrisk efforts (of which I’m aware) focus on AI-risk, which can’t explain the great filter. Most analysis efforts also focus on #1, less on #2, and almost none on #3.

GD Star Rating
Tagged as: , ,
Trackback URL:
  • Doug

    Light-cone eating AI doesn’t explain the great filter, but there are many variations of AI xrisk that don’t fall into this category. What about the classic example of a super-intelligent AI who’s objective function is to maximize paperclip production over the next fiscal quarter?

    That AI would convert the whole world, including all humans, into paperclips. But it wouldn’t bother with interstellar colonization because it wouldn’t payoff within a year. Not saying it’s likely, but there are many AI xrisk scenarios that involve the end of humanity but no interstellar colonization.

    • RobinHanson

      “Many scenarios” is not the same as “scenarios with a high total probability measure”.

      • Doug

        Strong AI is a subject area which we currently understand very little of. When one is faced with estimating probabilities in a deeply uncertain domain, a very useful heuristic is to shrink each plausible and distinct scenario to approximately equal probability weight.

    • Paul Crowley

      Such an AI still turns its entire light cone into paperclips, because it’s never wholly sure that the fiscal year is over, and so it might as well be on the safe side.

    • Weaver

      Wouldn’t it turn itself into paperclips, at the end?

      If we ever find the paperclip planet…attempt no landing there.

  • arch1

    Robin, I didn’t listen past 13:00 yet so apologies but can you help me understand how all of the following could be true? 1) There is a great filter of order ~10^-20, 2) “Anything that happened between humans and today” (or presumably during any given ~300k year period?) is “very unlikely to be a great filter step”, 3) It is very likely that in the *next* 300k years, if we don’t get wiped out (presumably by succumbing to a great filter step, which per #2 is unlikely), our progeny will disperse widely among the stars, 4) after which they’ll be very difficult to kill off.
    [Er, while typing I realized that possible explanations are: #3 is wrong because interstellar travel is surprisingly difficult, or #4 is wrong because physical dispersal doesn’t protect against all significant threats. And come to think of it #2 seems questionable too, since not all 300k-year periods are created equal. I think I’m sorry I asked…:-]

    • arch1

      I can’t believe I omitted the following possibility (the only optimistic one of the bunch): All of 1)-4) are true, and we have already survived the great filter (though we can’t yet say what it consisted of, which is a tad unsettling..)

    • RobinHanson

      Anything recent can’t be a step of the form of something where we keep try try trying until we succeed. Many past steps seem to have been of that form. We could have a step of the form where there’s a fixed window of vulnerability during which we either succeed or fail.

      • arch1

        That helped a lot, thanks!

  • solipsist

    Alien civilizations expanding at close to light-speed shouldn’t be visible. They can’t be seen for the same reason that supersonic missles can’t be heard. Right?

    And the typical progression of life might be a few billion years of silence, followed by a few hundred years of developing technology, followed by a few billion years of near light-speed expansion. In this scenario, technically nacent species should expect to see a silent sky. Right?

    So we should be confused about why we developed in such an old galaxy, not about why the skys appears lifeless. Right?

    • RobinHanson

      Yes, but I think that amounts to pretty much the same puzzle.

  • Dan Browne

    Charlie Stross has a good one: If future versions of us are augmented and are able to think big thoughts then maybe the communication between them is of such high information density that distance becomes a problem…

    Additionally: we’re trying to assess the probability of steps occuring that enable us to bypass big filters to being observed in the universe. What about the visibility of big filters themselves? If there are *lots* of civs but they are all wiped out, then the filters that do so don’t seem to be visible either. What does that tell us about where we are along the yellow brick road to a galactically expanding civ?

  • Dan Browne

    Here’s another one: 300 billion stars in the galaxy. Assume they were all born at the same time. There is a supernova every fifty years. That means that over 9 billion years we will see 180 million supernovas. That means that there is a supernova in each 1/1666 volume of the galaxy over the 9 billion years. The volume of the galaxy is roughly 8 trillion cubic light years. So the volume of 1/1666 of the galaxy is 4.8 billion light years. Using the formula for the volume of a sphere and working backwards I get a distance of about 1,000 light years radius for each supernova in the galaxy on average given the size of the galaxy. That’s not enough to wipe us out (we need 100 light years or closer). So that’s not it…

    • Weaver

      Supernova’s aren’t evenly distributed. You need to work out the proportion of potentially habitable worlds which do get zapped.

      *Hint* Many will survive

  • 5ive

    Love the discussion. Something I’ve heard you say as way as dismissal in other venues in addition to this one is something akin to “there’s going to be constant expansion.” I think that when discussing the Fermi paradox, there’s some potential bias when we compare our recent history and project it to the distant future.

    I don’t think humanity or our ancestors would quit expanding out of an environmental/altruistic impulse. However, I do think there’s a non-trivial chance that far before we get to Matrioshka-brain AIs–maybe AIs in Rhode Island sized computers–they might get bored. It’s possible that we aren’t *too far* away from figuring the universe out and that it won’t take too much processing time for a super-intelligence to answer all the interesting questions.

    It’s pretty common to hear “everyone goes into virtual reality because it’s so awesome” as a solution to the Fermi paradox, but the exploration-gets-boring version of that is probably more plausible. Not *everyone* is going to want to live in a virtual world, even if most did. But if I were a superintelligence who had it all figured out, I might just create five Dyson swarms like my own to establish a little redundancy and then stop expanding, and the reason for that might be obvious and compelling enough that every super-AI does the same thing.

    • RobinHanson

      To make a substantial contribution, it is not enough that some advanced folks get bored. It requires that most entire civilizations don’t have *any members who aren’t too bored to expand. That is a very high bar.

      • AnotherScaryRobot

        I know you’re not a big believer in it, but foom lowers this bar substantially. In many foom scenarios, the entity that undergoes foom may end up controlling most or all of the subsequent civilization’s resources, in effect reducing each civilization from a vast number of members to a single individual.

        There may also be effects that cause end-state superintelligences to be quite similar to each other. For instance, if there is a single optimal architectural design for intelligence given the physics of our universe, every self-improving superintelligence will, presumably, eventually end up with that architecture.

        In other words, instead of diversity both within and between civilizations, it’s plausible that there may be neither.

  • 5ive

    I’m not sure the “Earth had life on it almost right away” story is that interesting. A more interesting data point is how long it took very simple life to become multicellular. About 3.5-Billion years? Clearly *that step* wasn’t simple, even if basic life was. Three billion years is long enough that maybe most planets aren’t stable that long and we truly are special. The Cambrian Explosion is probably the Great Filter in the past that I’m most optimistic about. My top five potential great filters, in no order:

    1) Multicellular life is hard.
    2) The Universe is still very young. We are among the first technological civilizations.
    3) Our ancestors or nearby cohort civilizations were among the first and we are a simulation.
    4) AIs dominate the future and even relatively small super-AIs learn all there is to know surprisingly quickly and universally and quickly look inward.
    5) It’s difficult to imagine a technological catastrophe that would wipe us out completely and do it to most civilizations. The timing is too hard. It has to happen consistently before almost all civilizations colonize other planets but we, who are getting close, haven’t sniffed it yet? Instead, positive effective extinction. Within 1000 years or so, civilizations transcend in some way.

    • IMASBA

      It’s not that all multicellular life is very complex (there are multicellular lifeforms that are basically only 4 kinds of different cells patched together only slightly more secure than colonies of cooperating single cell lifeforms), but on Earth the concentration of oxygen was very low until the cambrian explosion and this is most likely what kept multicellular life from thriving before then.

      • Dan Browne

        Agreed. It looks fairly straightforward for multicellular life to develop – it happened repeatedly and continues to happen today. The latest evidence is that the geology of the earth itself actively worked to remove oxygen form the atmosphere even after photosynthesis started so there was a see saw effect on the levels of oxygen. Maybe that’s one of the great filters right there.

    • Dan Browne

      Actually the blind watchmaker has independently evolved multicellular life at least 40 times. And it’s still going on. The hard part is going from Prokyarotes to Eukyarotes. That only happened once. Every single Eukyarote there is today is a descendant of that original organism.

      • IMASBA

        How would we know if eukaryotes only evolved once? They don’t leave fossils with enough detail so for all we know there was competition between multiple kinds a billion years ago, it’s even possible it wasn’t so much competition as fusion: single celled organisms can exchange DNA with each other and receive DNA from viruses.

      • Dan Browne

        Agreed, we don’t. At the same time, however, there is one one lineage left in the remaining Eukaryotes whereas there are multiple derived lineages of multicellular organisms. We can even see yeast going from unicellular to multicellular almost in real time. Point I was trying to make is that prokaryote -> eukaryote is significantly more difficult thant eukaryote-> multicellular eukaryote.

      • IMASBA

        “Point I was trying to make is that prokaryote -> eukaryote is significantly more difficult thant eukaryote-> multicellular eukaryote.”

        I agree, I just didn’t agree we have enough evidence to point to prokaryote -> eukaryote being a great filter. A single line could be the result of a fusion of existing eukaryote lines before the first multicellular organisms existed.

      • Dan Browne

        Agreed: there is no way to know short of a time machine. I’m recalling a half baked memory from a textbook “Gene?” I read about ten years ago on Genetics 101 about Eukaryotes and Prokaryotes and which came first. There are arguments in both directions but with the biggest camp favoring a Prokaryotes came first and it was hard to go from Prokaryotes to Eukaryotes.


    To be precise: it’s mostly about not detecting signs of extraterrestrial technology. We now barely have the ability to detect an ocean on nearby exoplanets so there may be a large number of planets with life on them in our galaxy that we simply have not discovered and/or examined at a level where signs of life on the planet can be detected. So the question really becomes why haven’t we detected radio or laser signals, an abundance of terraformed planets or even alien spacecraft (or why hasn’t Earth been taken over by replicating machines long ago). I think that last point has been addressed by Robin before (Earth might be an oasis in a desert-like patch of the galaxy that alien civilizations don’t bother to exploit). Of course we don’t know how hard it is to terraform a planet so perhaps even with their best efforts aliens have not really multiplied the number of life-bearing planets by a large number, or, if we are an oasis, they have not done so in close proximity to Earth which means our current technology simply won’t detect planets. When it comes to the lack of signals: this may just indicate a lack of deliberate communication, if the nearest advanced planet is 200 lightyears away (a mere fraction of the galaxy’s radius) it will be very hard to detect any signal that is not deliberately amplified in our direction (from our own experience here on Earth we also know that long range radiotraffic can decrease with increasing technological advancement). There could of course be lots of cultural and strategic reasons as well.

  • Pingback: Overcoming Bias : Adam Ford & I on Great Filter | My Website

  • Charlene Cobleigh Soreff

    Good discussion! One quibble:
    “a Dyson Sphere, which is supposed to be several thousand times wider
    than a super-jupiter, would be detectable at 10000 – 20000 light-years
    as far as energy conservation isn’t broken and the energy output of the
    central star, after having been used by a type II civilisation, is
    re-emitted in thermal infrared. ”
    so we can rule out Dyson spheres from part of our galaxy, perhaps 10^8 stars or so – not from the whole 10^20 stars in the whole visible universe.

    • RobinHanson

      I didn’t mean to imply we could now see a star-sized Dyson sphere five billion light years away.

  • Pingback: The Great Filter | Matt Mullenweg

  • Pingback: The Great Filter - WP Creative

  • Dan Browne

    @Robin: The following link is a good one. Jason Wright at Penn State has written a paper on detecting Kardashev Type 3 Civs and has potential candidate galaxies where they might be found with potential evidence(!). Here is the original paper

    and here is the new scientist article