In his new book The Biological Universe: Life in the Milky Way and Beyond, evolutionary biologist Wallace Arthur predicts the life we will find in the galaxy and universe: Life forms are to be found across the Milky Way and beyond. They will be thinly spread, to be sure. … we can anticipate what life elsewhere will be like by examining the ecology and evolution of life on Earth.
There may be more basic steps to explore before jumping to extraterrestrial and artificial life. All living entities share some common characteristics but we lack knowledge whether these are independent ie essential. Living things all 'want' to grow and do work to achieve that and only do it for their singular case (membrane bound) Cells convert energy to work at very low entropy change rates which we are not able to duplicate. Cell survival is limited to low kinetic energy states. Our history tells us that knowledge progresses at slow rates but only with a succession of experiments. Can those promoting complex possibilities able to suggest possible starting points for experiments (above basic facts)? Are they open to such a suggestion? If not, should motivation considered. Just musings but if this has been solved please advise.
"Between two and three billion years from now … if [we?] make it that far, we might have the technology to colonize the closest suitable exoplanets."
Does anyone here doubt that, absent some nonobvious insuperable barrier to interstellar travel, we with our exploding technological acceleration could not only have the technology to, but actually *use it to colonize*, a few nearby exoplanets *and* begin despoiling their environments in say 100M years w/ room (& possibly orders of magnitude) to spare?
If Arthur provided details I'm dying to know them and may have to eat my metaphorical hat. Absent which the extreme duration and unexpected precision (+/- 25% for a technology project lasting *billions* of years?!?) strike me as almost comically ludicrous. I don't know much about interstellar travel, but I'd be hard pressed to come up w/ such a project timeline without inserting a *nearly* 2.5By +/- 0.5By canasta/lunch break.
The buzz is that Navy pilots are seeing UAPs every day.
Then we get blurry radar and thermal (infrared) images only, and there are doubts that such images are misinterpreted (or simply fabricated by bored Navy pranksters).
Cameras today are small and cheap and can store days on information.
If there are UAPs, and they monitor Navy fleets, and the Navy is not sure what they are, then I would think the Navy would mount optical cameras ASAP.
I am aching to believe that alien craft are zooming around. I love the whole topic of life in the galaxy. I believe the odds are heavily stacked in favor of intelligent life in our galaxy (though I ponder if distances are too vast to be navigated).
So why doesn't the Navy mount a bunch of cameras on aircraft?
But these biologists fully admit that a more general concept of natural selection applies understandably to artificial life. It just isn't what they are used to, so they prefer to fall back on stuff they are used to.
My critique of foom is about the likelihood of innovation being concentrated into a very small number of very powerful innovations, contrary to our experience of innovation for the last few billion years.
"Wallace Arthur seems to be yet another biologists who just can’t imagine our descendants being that different from us, or artificial life making much of a difference to the cosmos."
I have a hunch for why that's the case.
After reading Yudkowsky's Sequences, I read the ebook on the AI-Foom debate you both had, and I noticed how you kept your counterpoints to his arguments grounded on well known, well researched Economic mechanisms and principles, while he argued extrapolating from more general epistemological notions.
My guess is that, in this debate here, something similar is going on. These academic authors discussing the possibilities for alien life are biologists, and therefore, while they do briefly consider more speculative possibilities on post- and non-biological life, they then ground the discussion back to the bulk of well known, well researched biological mechanisms and principles.
Hence, my take is that, from their perspective, AI- and transhumanist-inspired arguments, while interesting, cannot be taken more seriously until we first do have GAIs and transhuman biological-machine hybrids to observe how they operate, find the laws that do govern them (rather than inferring the ones that may come to govern them if and when they arise), and then extrapolate from those. Which is similar to how, from your perspective (or what I understood of it), Yudkowsky's arguments about a small team developing a seed GAI before everyone else, this AI via recursive cascades becoming super-intelligent, and then dominating the world beyond and despite the boundaries of economies of scale and large-scale technological and industrial synergies, doesn't seem likely compared to the hard-Economics-inspired EM future.
If this analogy is correct, we then have three levels of abstraction, from highest to lowest, on the subject of reductive physics-compliant life:
a) AI-theoretical (+ Physics-bound): Kurzweil, Yudkowsky etc.
b) "a" + Economics-bound: yours.
c) "b" + Biology-bound: Mason, Kershenbaum, Arthur etc.
With clear conceptual breaks separating these three distinct takes.
I'm also curious if we know anywhere near enough to be sure that life doesn't exist inside stars, in accretion disks or other high density high energy scenarios. Certainly chemical life is right out but how do we know that there aren't complex reproducing structures based on the kind of interactions in these enviornments.
Maybe there is a good general argument but I haven't heard it.
Given the evidence of your summary of the book, academics are *not* doing a "more careful better" job than beer discussants. Enrico Fermi had a better grip on the topic over the lunch table at Los Alamos.
Odd that he rules out brown dwarfs. The lower levels of their atmospheres might be too hot for complex molecules, but what about the upper levels? Given that there may be as many BDs as there are luminous stars, that's a topic worth investigating.
Assuming AI is developed by Natural General Intelligence (NGI), some subset of universe-roaming AI will be created to develop and spread NGI elsewhere in the galaxy (lets call this AI subset 'spreader AI'). So I agree with you Robin that ignoring AI expansion generally also ignores spreader AI, which is a (realistic, plausible and quicker) way for NGI to spread itself across the universe.
Spreader AI and NGI expansion rates will have a relationship to each other depending on the types of expansion activities such AI's are developed to undertake.
Two types of spreader AI come to mind that would enable NGI to expand at either at the same rate as spreader AI or faster.
- 'Carer' AI's like in the show 'Raised by wolves' where AI is tasked to find habitable planets and raise NGI's. If only 'carer' AI's are used, we would expect the spreader AI and NGI expansion rates across the universe to be similar, or only marginally different in cases where the AI survives on the new world but NGI doesn't, and vice versa.
- 'Seeder' AI's that remain or pass by planets and insert various technologies to enable future life. The technology inserted could be robust but less complex forms of life with the intention that these evolve into NGI's over time, or more complex life that have been robustly designed to survive new conditions. The technology could also be that which terraforms the planet, possibly with the idea of a sending a second seeder AI at a later stage. If only 'seeder' AI's were used and they remained on the planet, the expansion rates would again be similar. But if seeder AI's did not themselves remain on the planet that was seeded but just inserted technology as it passed by, we might expect NGI's to expand much faster than spreader AI.
And of course, I suspect both forms of AI will be used separately and in combination (e.g. a seeder remaining and turning into an eventual carer, a seeder planting many carers, a carer manufacturing new seeders, etc). In all of these cases I expect NGI to expand as fast or faster than spreader AI.
Will spreader AI be a large subset of AI generally? And therefore expand NGI's quicker than AI's more generally? I doubt it, and even if spreader AI were 99.99% of all AI created, the final 0.01% of AI that self-replicates without biological constraints would soon encompass the universe quicker than NGI's are capable of doing. Still...this was a fun thinking exercise. Thanks for the post Robin!
Thank you for reviewing this argument. I also find it disappointing. How can someone so confidently and handwavingly extrapolate from one clearly unrepresentative data point (our living solar system), and yet ignore the much more telling fact that we've not been colonized by aliens? The latter fact tells us far more about what does and doesn't populate the Milky Way.
The much safer extrapolation is that life is grabby and spreading is easy, so if there are other civilizations in our galaxy, they would be super obvious even to the naked eye. But it doesn't take biological expertise to make this superior argument, so maybe that's why biologists don't make it.
But surely the very best topics for research are also topics for beer speculation, right? The fact of beer speculation on a topic doesn't mean that researchers aren't doing a more careful better job of discussing them, right?
When a handful of undergrads (a mix of biology, chemistry, and physics majors, perhaps?) chat about exactly this sort of topic across a table in a loud restaurant, while drinking a lot of beer and consuming a lot of pizza, nobody else cares, right? But when an academic writes a book (or blog) full of more or less the same wild speculations (adding a Table of Contents, better organization overall, and some proofreading), the same ideas become worthy of attention and debate within the wider academic community. Now that's fascinating.
There may be more basic steps to explore before jumping to extraterrestrial and artificial life. All living entities share some common characteristics but we lack knowledge whether these are independent ie essential. Living things all 'want' to grow and do work to achieve that and only do it for their singular case (membrane bound) Cells convert energy to work at very low entropy change rates which we are not able to duplicate. Cell survival is limited to low kinetic energy states. Our history tells us that knowledge progresses at slow rates but only with a succession of experiments. Can those promoting complex possibilities able to suggest possible starting points for experiments (above basic facts)? Are they open to such a suggestion? If not, should motivation considered. Just musings but if this has been solved please advise.
"Between two and three billion years from now … if [we?] make it that far, we might have the technology to colonize the closest suitable exoplanets."
Does anyone here doubt that, absent some nonobvious insuperable barrier to interstellar travel, we with our exploding technological acceleration could not only have the technology to, but actually *use it to colonize*, a few nearby exoplanets *and* begin despoiling their environments in say 100M years w/ room (& possibly orders of magnitude) to spare?
If Arthur provided details I'm dying to know them and may have to eat my metaphorical hat. Absent which the extreme duration and unexpected precision (+/- 25% for a technology project lasting *billions* of years?!?) strike me as almost comically ludicrous. I don't know much about interstellar travel, but I'd be hard pressed to come up w/ such a project timeline without inserting a *nearly* 2.5By +/- 0.5By canasta/lunch break.
A little OT but in the ballpark:
Why not optical cameras mounted on Navy jets?
The buzz is that Navy pilots are seeing UAPs every day.
Then we get blurry radar and thermal (infrared) images only, and there are doubts that such images are misinterpreted (or simply fabricated by bored Navy pranksters).
Cameras today are small and cheap and can store days on information.
If there are UAPs, and they monitor Navy fleets, and the Navy is not sure what they are, then I would think the Navy would mount optical cameras ASAP.
I am aching to believe that alien craft are zooming around. I love the whole topic of life in the galaxy. I believe the odds are heavily stacked in favor of intelligent life in our galaxy (though I ponder if distances are too vast to be navigated).
So why doesn't the Navy mount a bunch of cameras on aircraft?
But these biologists fully admit that a more general concept of natural selection applies understandably to artificial life. It just isn't what they are used to, so they prefer to fall back on stuff they are used to.
My critique of foom is about the likelihood of innovation being concentrated into a very small number of very powerful innovations, contrary to our experience of innovation for the last few billion years.
"Wallace Arthur seems to be yet another biologists who just can’t imagine our descendants being that different from us, or artificial life making much of a difference to the cosmos."
I have a hunch for why that's the case.
After reading Yudkowsky's Sequences, I read the ebook on the AI-Foom debate you both had, and I noticed how you kept your counterpoints to his arguments grounded on well known, well researched Economic mechanisms and principles, while he argued extrapolating from more general epistemological notions.
My guess is that, in this debate here, something similar is going on. These academic authors discussing the possibilities for alien life are biologists, and therefore, while they do briefly consider more speculative possibilities on post- and non-biological life, they then ground the discussion back to the bulk of well known, well researched biological mechanisms and principles.
Hence, my take is that, from their perspective, AI- and transhumanist-inspired arguments, while interesting, cannot be taken more seriously until we first do have GAIs and transhuman biological-machine hybrids to observe how they operate, find the laws that do govern them (rather than inferring the ones that may come to govern them if and when they arise), and then extrapolate from those. Which is similar to how, from your perspective (or what I understood of it), Yudkowsky's arguments about a small team developing a seed GAI before everyone else, this AI via recursive cascades becoming super-intelligent, and then dominating the world beyond and despite the boundaries of economies of scale and large-scale technological and industrial synergies, doesn't seem likely compared to the hard-Economics-inspired EM future.
If this analogy is correct, we then have three levels of abstraction, from highest to lowest, on the subject of reductive physics-compliant life:
a) AI-theoretical (+ Physics-bound): Kurzweil, Yudkowsky etc.
b) "a" + Economics-bound: yours.
c) "b" + Biology-bound: Mason, Kershenbaum, Arthur etc.
With clear conceptual breaks separating these three distinct takes.
I think you need to better sample typical beer speculations.
I'm also curious if we know anywhere near enough to be sure that life doesn't exist inside stars, in accretion disks or other high density high energy scenarios. Certainly chemical life is right out but how do we know that there aren't complex reproducing structures based on the kind of interactions in these enviornments.
Maybe there is a good general argument but I haven't heard it.
Given the evidence of your summary of the book, academics are *not* doing a "more careful better" job than beer discussants. Enrico Fermi had a better grip on the topic over the lunch table at Los Alamos.
Odd that he rules out brown dwarfs. The lower levels of their atmospheres might be too hot for complex molecules, but what about the upper levels? Given that there may be as many BDs as there are luminous stars, that's a topic worth investigating.
Assuming AI is developed by Natural General Intelligence (NGI), some subset of universe-roaming AI will be created to develop and spread NGI elsewhere in the galaxy (lets call this AI subset 'spreader AI'). So I agree with you Robin that ignoring AI expansion generally also ignores spreader AI, which is a (realistic, plausible and quicker) way for NGI to spread itself across the universe.
Spreader AI and NGI expansion rates will have a relationship to each other depending on the types of expansion activities such AI's are developed to undertake.
Two types of spreader AI come to mind that would enable NGI to expand at either at the same rate as spreader AI or faster.
- 'Carer' AI's like in the show 'Raised by wolves' where AI is tasked to find habitable planets and raise NGI's. If only 'carer' AI's are used, we would expect the spreader AI and NGI expansion rates across the universe to be similar, or only marginally different in cases where the AI survives on the new world but NGI doesn't, and vice versa.
- 'Seeder' AI's that remain or pass by planets and insert various technologies to enable future life. The technology inserted could be robust but less complex forms of life with the intention that these evolve into NGI's over time, or more complex life that have been robustly designed to survive new conditions. The technology could also be that which terraforms the planet, possibly with the idea of a sending a second seeder AI at a later stage. If only 'seeder' AI's were used and they remained on the planet, the expansion rates would again be similar. But if seeder AI's did not themselves remain on the planet that was seeded but just inserted technology as it passed by, we might expect NGI's to expand much faster than spreader AI.
And of course, I suspect both forms of AI will be used separately and in combination (e.g. a seeder remaining and turning into an eventual carer, a seeder planting many carers, a carer manufacturing new seeders, etc). In all of these cases I expect NGI to expand as fast or faster than spreader AI.
Will spreader AI be a large subset of AI generally? And therefore expand NGI's quicker than AI's more generally? I doubt it, and even if spreader AI were 99.99% of all AI created, the final 0.01% of AI that self-replicates without biological constraints would soon encompass the universe quicker than NGI's are capable of doing. Still...this was a fun thinking exercise. Thanks for the post Robin!
Thank you for reviewing this argument. I also find it disappointing. How can someone so confidently and handwavingly extrapolate from one clearly unrepresentative data point (our living solar system), and yet ignore the much more telling fact that we've not been colonized by aliens? The latter fact tells us far more about what does and doesn't populate the Milky Way.
The much safer extrapolation is that life is grabby and spreading is easy, so if there are other civilizations in our galaxy, they would be super obvious even to the naked eye. But it doesn't take biological expertise to make this superior argument, so maybe that's why biologists don't make it.
But surely the very best topics for research are also topics for beer speculation, right? The fact of beer speculation on a topic doesn't mean that researchers aren't doing a more careful better job of discussing them, right?
When a handful of undergrads (a mix of biology, chemistry, and physics majors, perhaps?) chat about exactly this sort of topic across a table in a loud restaurant, while drinking a lot of beer and consuming a lot of pizza, nobody else cares, right? But when an academic writes a book (or blog) full of more or less the same wild speculations (adding a Table of Contents, better organization overall, and some proofreading), the same ideas become worthy of attention and debate within the wider academic community. Now that's fascinating.