We need to start worrying about Ant Takeover happening before AI Takeover
1.1. If all ant colonies are already intelligent, they might be able to take over already via surprise brute force attack. They don’t need to kill everyone, just enough people to collapse civilization, and they can clog and destroy machines to help with that. Once civilization collapses I’m not optimistic that we could rebuild in the face of determined ant opposition, and certainly we couldn’t do it before they tech up and acquire powerful weapons of their own.
1.1.2. Ants could probably get fire pretty easily, especially since they can steal fire from humans. They need to get the knowledge first, which might be hard. In general, they might be bottlenecked by understanding of humans? It’s not like anyone teaches them to read.
1.1.3. This may be our main advantage: They only recently became intelligent, so they probably can’t read yet and may not even realize that humans are intelligent. They probably don’t realize that we know they are intelligent. Concepts like intelligence may be foreign to them, having had so little time to think and having terrible communication networks with other ants.
1.1.4. Come to think of it, how do these ant colonies communicate with each other? Probably not very well; at best they can only talk to the colonies adjacent to them, and…
1.1.5. they probably think more slowly than us too, since signals have to pass between their individual ants, which are spread out over several square meters and move slowly. I’d guess they think at 0.1x the speed of humans, or slower. This might not matter for some things (maybe they make up in quality what they lack in speed) but for communication it is probably especially tough.
1.1.6. On the other hand maybe they have access to high-bandwith communication, better than the audio and video signals we use? I doubt it, but maybe.
1.1.7. Anyhow, memetically, ant colonies are probably not really a civilization yet because they just haven’t had time to develop concepts. So much of what makes humans powerful is cultural evolution; ideas getting invented and refined and spread. Ants, even if they are as smart as humans or even if they are smarter, probably lack the concepts to really understand what’s going on around them. For now. For now we are safe from coordinated Ant Takeover.
1.1.8. The fact that they all became intelligent around now suggests that maybe they are communicating quickly via some magic method. Probably not though; the deus ex machina hypothesis seems more likely.
1.2. We should think about Ant alignment / values. What do they want? Probably nothing good for us, lol.
1.3. Maybe they can be bargained with? We probably have things they want.
1.3.1. We have spaceships. We have technology in general. And currently we are living mostly peacefully. Maybe we can get them to not kill us in return for giving them some of our tech and not killing them.
1.4. What is their rate of improvement? Ant colonies reproduce faster than humans. Optimistically they got intelligence via deus ex machina rather than via some natural process, because in the latter case we should expect them to be superhuman in a week or less and possibly in a few hours. But even in the deus ex machina scenario they can probably evolve quickly, if they want to.
1.4.1. Maybe we can try to make them evolve back to being stupid? Seems hard for us to exert that kind of selection pressure, since they are smart enough to not reveal their intelligence probably.
We need to start thinking about ant-human collaborations and how they could reshape the world economy.
2.1. Ants might be better pesticides than our current stuff?
2.2. Ants would certainly have all sorts of niche jobs like maintaining systems that have lots of small bits to crawl into and inspect.
2.3. A million ants is probably not very heavy. Maybe ants would make better astronauts than humans; we could put them in charge of robot terraformers or something.
2.4. Ants have other benefits for space colonization too probably—less need for natural light and space, more comfort with compressed habitat.
2.5. The military uses of ants are pretty obvious. Extremely stealthy assassins, carry poison into enemy commander, come back and report on things.
2.6. Slow though. Maybe drone delivery is the fix for that.
2.7. Maybe ants would be more efficient as computers than our current computers. Not for crunching numbers, but for doing other things. Probably cheaper and better than GPT-3, for example.
2.8. Probably ants would be useful for digging tunnels and running wires through them. Could cut costs for fiber optic cable deployment immensely, I think.
2.9. With human help, ants could probably be bred for intelligence, achieving superhuman levels in just a few years. (They might not need human help, lol.) This would of course make their economic applications skyrocket.
2.10. Might ants be useful for manufacturing? running wires through dense 3d-printed parts? Yeah seems reasonable.
2.11. Probably ants wouldn’t completely reshape the economy though, on the supply side, because I think most of the economy is transporting things and mining things and doing other “big” things that ants wouldn’t help with compared to our existing machines.
2.12. On the demand side, ants would change the economy a lot probably. Construction would become more expensive due to the need to negotiate with the local ants (otherwise, expect resistance!).
2.12.1. More demand for basic foods, less demand (proportionally) for meat. Insect farms probably become a thing.
2.12.2. Probably ants have some weird preferences that we aren’t yet aware of—certain goods become luxury goods, etc.
2.12.3. What is ant religion like? What is ant ideology like? If they are immune to such things, maybe that’s a good thing? If not, well, we’ve got to learn to live with it.
We should increase our credence that we are in a simulation, because this seems much more likely to happen in a simulation than not. But what kind of simulation would do this?
3.1. Probably a simulation made by someone for fun or whimsy? I can’t think of any scientific purpose for this yet, but maybe there is one…
3.1.1. in a sci-fi story I read, some super powerful aliens give random bits of their power to humans to see how the humans use them, and the rationale is basically to provide training data IIRC. Try to generalize from creative uses of power in different combinations and circumstances to a better understanding of power more generally. Maybe something like this is going on.
3.1.2. Maybe the idea is to collect data about interactions between intelligent species on the same planet, and this is just a crude way of getting that data (a more elegant way would involve co-evolving two intelligent species, since that’s more likely to happen naturally)
3.2. I had thought that most simulations were ancestor simulations. This is evidence against that claim. Which in turn is evidence against the things that made me think that, like acausal trade stuff.
If collectives of ants can be intelligent, can’t collectives of humans be intelligent too? Well, corporations and nations and so forth certainly act like they are intelligent… maybe this isn’t really news.
Multi-agent theories of mind should go up in probability on this evidence.
5.1. What implications does that have? Not much as far as I know. Maybe it reduces my credence in dualism, idealism, etc. and increases my credence in physicalism?
5.2. It makes me feel like the classical models of intelligence (expected utility maximization, etc.) are probably less useful for AI alignment etc. than I thought. That said, there’s the complete class theorem and stuff...
Currently no one but me knows about this. Who should I tell? If I tell the whole world, probably someone will start communicating with the ants, which will give them a temporary advantage since they can collaborate with ants militarily, but probably puts us all at risk by accelerating the ants’ memetic evolution and awareness of the strategic situation.
6.1. I should probably tell other EAs/rationalists first, and we’ll figure out what to do together.
6.2. In the long run, it seems like ants will evolve to be smarter than us, and that’ll be bad news. However, this will take several years at least. Perhaps the situation can be changed by then.
6.2.1. We can create AGI, for example. In an ant-proof bunker.
6.2.2. Tentative plan: We alert certain powerful actors in the world (The US government?) who then harden their systems against ant attack and prepare to defend against ant war.
6.2.3. Also, they start talking with some ants on an island or something to learn more about them. Try to detect whether other ants elsewhere in the world are communicating with those ants.
6.2.4. Keep working on AI alignment etc. Our hope is to achieve a harmonious future in which ants, humans, and AIs all live together happily despite differences in intelligence.
6.2.5. Maybe what we learn from the ants can help us with AI alignment. Try to figure out how they got intelligence, and what their values are.
6.2.6. Maybe ant colonies can be aligned! Because of the way they are distributed, maybe they are more transparent than humans. Heck maybe they are more transparent than AIs too because they only have a million ants instead of trillions of parameters. And transparency is super important.
6.2.7. Assuming we can figure out how to read ant minds (when they are suitably in a farm or scannable region so we can see what all their members are doing) then we should probably start scaling up ant-human cooperation big time, as an alternative to TAI.
6.2.8 Ant capability gain is probably slower too, not being able to copy themselves instantly and so on. So we are in for a much slower takeoff, which is good. Whew!
6.2.9. The downside is that we have much less direct control over their minds, and they are already intelligent so they are already mesa-optimizers basically and are likely to be deceptively aligned rather than actually aligned.
6.2.10. All things considered I’m still optimistic though. For one thing, we can be nice to them.
6.2.11. For another, we can keep them near human level and get them to work on AI alignment theory, at massive scale. Ant colonies probably consume less food than humans and thus can be cheaper than human workers.
This is evidence that other species might start achieving intelligence soon. If ants, certainly termites. But also dolphins, meerkats, heck, even flies, why not? Yeah, a priori ants seem like a more plausible candidate, but this is strong evidence that our a priori reasoning isn’t working!
7.1. We should launch investigations into this. We should also (jesus) begin a plan to end factory farming and the destruction of nature. We probably should have been doing that anyway, but now is better than ever… time to bring the coercive force of the state to bear, probably...
Wait a minute, are my readings just wrong? Maybe this is a massive prank.
I think this increases my credence that God exists. It sounds like something He would do.
OK, here goes:
We need to start worrying about Ant Takeover happening before AI Takeover 1.1. If all ant colonies are already intelligent, they might be able to take over already via surprise brute force attack. They don’t need to kill everyone, just enough people to collapse civilization, and they can clog and destroy machines to help with that. Once civilization collapses I’m not optimistic that we could rebuild in the face of determined ant opposition, and certainly we couldn’t do it before they tech up and acquire powerful weapons of their own. 1.1.2. Ants could probably get fire pretty easily, especially since they can steal fire from humans. They need to get the knowledge first, which might be hard. In general, they might be bottlenecked by understanding of humans? It’s not like anyone teaches them to read. 1.1.3. This may be our main advantage: They only recently became intelligent, so they probably can’t read yet and may not even realize that humans are intelligent. They probably don’t realize that we know they are intelligent. Concepts like intelligence may be foreign to them, having had so little time to think and having terrible communication networks with other ants. 1.1.4. Come to think of it, how do these ant colonies communicate with each other? Probably not very well; at best they can only talk to the colonies adjacent to them, and… 1.1.5. they probably think more slowly than us too, since signals have to pass between their individual ants, which are spread out over several square meters and move slowly. I’d guess they think at 0.1x the speed of humans, or slower. This might not matter for some things (maybe they make up in quality what they lack in speed) but for communication it is probably especially tough. 1.1.6. On the other hand maybe they have access to high-bandwith communication, better than the audio and video signals we use? I doubt it, but maybe. 1.1.7. Anyhow, memetically, ant colonies are probably not really a civilization yet because they just haven’t had time to develop concepts. So much of what makes humans powerful is cultural evolution; ideas getting invented and refined and spread. Ants, even if they are as smart as humans or even if they are smarter, probably lack the concepts to really understand what’s going on around them. For now. For now we are safe from coordinated Ant Takeover. 1.1.8. The fact that they all became intelligent around now suggests that maybe they are communicating quickly via some magic method. Probably not though; the deus ex machina hypothesis seems more likely. 1.2. We should think about Ant alignment / values. What do they want? Probably nothing good for us, lol. 1.3. Maybe they can be bargained with? We probably have things they want. 1.3.1. We have spaceships. We have technology in general. And currently we are living mostly peacefully. Maybe we can get them to not kill us in return for giving them some of our tech and not killing them. 1.4. What is their rate of improvement? Ant colonies reproduce faster than humans. Optimistically they got intelligence via deus ex machina rather than via some natural process, because in the latter case we should expect them to be superhuman in a week or less and possibly in a few hours. But even in the deus ex machina scenario they can probably evolve quickly, if they want to. 1.4.1. Maybe we can try to make them evolve back to being stupid? Seems hard for us to exert that kind of selection pressure, since they are smart enough to not reveal their intelligence probably.
We need to start thinking about ant-human collaborations and how they could reshape the world economy. 2.1. Ants might be better pesticides than our current stuff? 2.2. Ants would certainly have all sorts of niche jobs like maintaining systems that have lots of small bits to crawl into and inspect. 2.3. A million ants is probably not very heavy. Maybe ants would make better astronauts than humans; we could put them in charge of robot terraformers or something. 2.4. Ants have other benefits for space colonization too probably—less need for natural light and space, more comfort with compressed habitat. 2.5. The military uses of ants are pretty obvious. Extremely stealthy assassins, carry poison into enemy commander, come back and report on things. 2.6. Slow though. Maybe drone delivery is the fix for that. 2.7. Maybe ants would be more efficient as computers than our current computers. Not for crunching numbers, but for doing other things. Probably cheaper and better than GPT-3, for example. 2.8. Probably ants would be useful for digging tunnels and running wires through them. Could cut costs for fiber optic cable deployment immensely, I think. 2.9. With human help, ants could probably be bred for intelligence, achieving superhuman levels in just a few years. (They might not need human help, lol.) This would of course make their economic applications skyrocket. 2.10. Might ants be useful for manufacturing? running wires through dense 3d-printed parts? Yeah seems reasonable. 2.11. Probably ants wouldn’t completely reshape the economy though, on the supply side, because I think most of the economy is transporting things and mining things and doing other “big” things that ants wouldn’t help with compared to our existing machines. 2.12. On the demand side, ants would change the economy a lot probably. Construction would become more expensive due to the need to negotiate with the local ants (otherwise, expect resistance!). 2.12.1. More demand for basic foods, less demand (proportionally) for meat. Insect farms probably become a thing. 2.12.2. Probably ants have some weird preferences that we aren’t yet aware of—certain goods become luxury goods, etc. 2.12.3. What is ant religion like? What is ant ideology like? If they are immune to such things, maybe that’s a good thing? If not, well, we’ve got to learn to live with it.
We should increase our credence that we are in a simulation, because this seems much more likely to happen in a simulation than not. But what kind of simulation would do this? 3.1. Probably a simulation made by someone for fun or whimsy? I can’t think of any scientific purpose for this yet, but maybe there is one… 3.1.1. in a sci-fi story I read, some super powerful aliens give random bits of their power to humans to see how the humans use them, and the rationale is basically to provide training data IIRC. Try to generalize from creative uses of power in different combinations and circumstances to a better understanding of power more generally. Maybe something like this is going on. 3.1.2. Maybe the idea is to collect data about interactions between intelligent species on the same planet, and this is just a crude way of getting that data (a more elegant way would involve co-evolving two intelligent species, since that’s more likely to happen naturally) 3.2. I had thought that most simulations were ancestor simulations. This is evidence against that claim. Which in turn is evidence against the things that made me think that, like acausal trade stuff.
If collectives of ants can be intelligent, can’t collectives of humans be intelligent too? Well, corporations and nations and so forth certainly act like they are intelligent… maybe this isn’t really news.
Multi-agent theories of mind should go up in probability on this evidence. 5.1. What implications does that have? Not much as far as I know. Maybe it reduces my credence in dualism, idealism, etc. and increases my credence in physicalism? 5.2. It makes me feel like the classical models of intelligence (expected utility maximization, etc.) are probably less useful for AI alignment etc. than I thought. That said, there’s the complete class theorem and stuff...
Currently no one but me knows about this. Who should I tell? If I tell the whole world, probably someone will start communicating with the ants, which will give them a temporary advantage since they can collaborate with ants militarily, but probably puts us all at risk by accelerating the ants’ memetic evolution and awareness of the strategic situation. 6.1. I should probably tell other EAs/rationalists first, and we’ll figure out what to do together. 6.2. In the long run, it seems like ants will evolve to be smarter than us, and that’ll be bad news. However, this will take several years at least. Perhaps the situation can be changed by then. 6.2.1. We can create AGI, for example. In an ant-proof bunker. 6.2.2. Tentative plan: We alert certain powerful actors in the world (The US government?) who then harden their systems against ant attack and prepare to defend against ant war. 6.2.3. Also, they start talking with some ants on an island or something to learn more about them. Try to detect whether other ants elsewhere in the world are communicating with those ants. 6.2.4. Keep working on AI alignment etc. Our hope is to achieve a harmonious future in which ants, humans, and AIs all live together happily despite differences in intelligence. 6.2.5. Maybe what we learn from the ants can help us with AI alignment. Try to figure out how they got intelligence, and what their values are. 6.2.6. Maybe ant colonies can be aligned! Because of the way they are distributed, maybe they are more transparent than humans. Heck maybe they are more transparent than AIs too because they only have a million ants instead of trillions of parameters. And transparency is super important. 6.2.7. Assuming we can figure out how to read ant minds (when they are suitably in a farm or scannable region so we can see what all their members are doing) then we should probably start scaling up ant-human cooperation big time, as an alternative to TAI. 6.2.8 Ant capability gain is probably slower too, not being able to copy themselves instantly and so on. So we are in for a much slower takeoff, which is good. Whew! 6.2.9. The downside is that we have much less direct control over their minds, and they are already intelligent so they are already mesa-optimizers basically and are likely to be deceptively aligned rather than actually aligned. 6.2.10. All things considered I’m still optimistic though. For one thing, we can be nice to them. 6.2.11. For another, we can keep them near human level and get them to work on AI alignment theory, at massive scale. Ant colonies probably consume less food than humans and thus can be cheaper than human workers.
This is evidence that other species might start achieving intelligence soon. If ants, certainly termites. But also dolphins, meerkats, heck, even flies, why not? Yeah, a priori ants seem like a more plausible candidate, but this is strong evidence that our a priori reasoning isn’t working! 7.1. We should launch investigations into this. We should also (jesus) begin a plan to end factory farming and the destruction of nature. We probably should have been doing that anyway, but now is better than ever… time to bring the coercive force of the state to bear, probably...
Wait a minute, are my readings just wrong? Maybe this is a massive prank.
I think this increases my credence that God exists. It sounds like something He would do.