I think I understand the general premise of your argument and appreciate the appeal, but something still doesn’t sit well with me with regard to how much freedom this eliminates from the world. I’m trying to think it through and happy to have your feedback. First concern is that this tech would eventually get into the hands of people who have different moral and ethical standards, and who would apply it in a way that would result in a tyrannical dystopia with no possibility for meaningful rebellion. The second concern is when I take the idea further and extend the possibilities: what if AI can get accurately predictive enough to determine when the bad guy was about to pull the trigger and initiate a counter attack before he actually does. Beyond that, what if it can learn to predict his actions based on a real time brain scan, and stop him once he’s had just the thought of attack. A bit of Minority Report science fiction... but maybe not?
Someone, somewhere is going to do something like this with drones. I would prefer it be a freedom loving people who can code their values into the system.
I wish this were possible, but I'm pretty sure it's not, at least not in the near future.
* Ordinary bullets move at around 3000 feet per second, more than twice the speed of sound. By the time the sound reaches the microphone, that bullet will have already hit its target. Light is easily blocked by obstacles, and even if the camera does see a bullet, there's less than a hundredth of a second to do anything about it. You would need at least two frames of data to do any speed detection (much less figuring out 3d space from it), and while I could imagine the tech getting good enough at some point to calculate it in time, swivelling a gun around to fire in that amount of time is basically impossible.
* Hamas rockets are extremely cheap to make, and are often fired on timers, so that the terrorists are gone before it fires. They use grenades, bombs, drones of their own, whatever weaponry they can smuggle in via the tunnels to Egypt. These don't have easy counters.
* Drones are very expensive, and even if they're very high-end, they cannot move fast enough to dodge bullets. Easier would be to track direction of motion to move out of the line of fire in advance, but that would still probably be very difficult. These things don't have great manoeuvrability.
* Tunnel openings tend to be inside buildings (eg, into an inconspicuous child's bedroom, under the bed), and there's no easy way to locate the entrances.
I do like the general line of thinking though. Hopefully something good can come of it.
Would you be okay if I used this as part of another post? I’ll try to do a quick reply but high-level, I try to lay out a plausible vision more than the particulars when I write a post like this. I usually have a lot more detail up in my head than I have time to write down. I know you show up with knife and fork ready for the steak and I’d love to give the steak, but most of my time comes in five minute increments these days and it’s very hard to do anything long unless I happen to wake up extremely early one day. That said, I super value your time and don’t want to make too many excuses.
High level on stuff I think you agree on but didn’t mention that I’d like to point is still really big and valuable: it’s possible to make an AI model that would confer this level of marksmanship and targeting to an autonomous system and having the military deliberately go out to create a dataset to do so is a valuable use of taxpayer dollars and something that an organization on scale of a national military could reasonably accomplish. Just that alone is huge.
On timing, bullet speed, turret speed, etc.
I call this the “First Bullet Problem.” There’s lots of depth of field stuff with cameras I *think* you could do to address this. Kind of. That said, I’m almost certain we’ll never get to a scenario where you’d be able to shoot the first one of these bullets out of the air before it could hit its target. By the time the first bullet has struck the counter shots will have had time to take a predictive aim for where the next volley is going to be. But the problem isn’t really just “shoot a bullet with a bullet.” I think that’s what you could do for things like ship to ship battle and probably over long range for military engagements like we have in Ukraine but up close urban warfare is going to be tricky. The underlying problem statement is “I need to move an air column about a foot over to the left or right while a bullet is traveling through it.” If you can just make a very quick decision that a bullet is incoming and about to hit someone, you could do something like having an explosion of condensed air accomplish something like that. The particulars are of course challenging but I still think you could do something like that even if someone had to wear a really crazy personal “wind funnel and compressed air cartridge” device. I would see this entire thing as a ten or fifteen year project starting with small things like the laser pointer and building up to full scale drone manufacture.
On Hamas rockets being very cheap, timers, etc. This is one where I think quantity creates quality. If you have thousands of basically cheap cell phones flying over Gaze taking pictures at fairly quick intervals you can see who is going where to do what and then use that to target your drone swarms. You could also do something like have a thousand 19 year old military guys carry sensors around in a tunnel until you could teach a drone to fly inside of one.
On drones being expense: agreed. You have to develop specific tech for this and then you have to scale it. My point isn’t necessarily that 1,000 is the exactly target. It’s more that if you can move toward the extreme cheap end that you get to make easy trade offs between sacrificing equipment to save lives and you also have such overwhelming force your enemy even in his own mind has to admit you cannot be defeated. I agree on the bullet dodging and didn’t mean to imply they would dodge a bullet already in fire. Again, a first bullet problem. If you have line of sight on the barrel, you have a model that has the drone take an evasive maneuver. The 19 year old dudes in the military are going to be doing a lot of stuff to create this training data and at any given time there will be a few thousand dudes working at Gun Lab to make this data set.
On tunnel openings being inside buildings. Make a few drone variants that bust open doors. This is an engineering problem and can be overcome. The swarm doesn’t have to be completely homogenous, just the mass volume should be a certain and cheap type.
Do this for ten, fifteen, or twenty years and you end up with something much different than we have now.
In summary: we are moving into a world where you can start certain actions based on “wake sounds” and “wake signs” and we should rethink generating data instead of saying “wouldn’t it be nice if we had this data?” You can remove the human risk of your own soldiers completely from these enforcement decisions leading to more humane outcomes. Even the lower fidelity, less capable versions of this could prevent some kids from getting blown apart. I think I have a guess as to who you are behind the anon but even if I’m wrong I’d bet quite a lot of money you are the type of person who would like to do everything possible to save those kids.
This raises a secondary issue: if you can make a perfect system like this, that's great. A partially functioning system, or one that the enemy can eventually learn to bypass, is probably a net negative.
This relates to the issue where iron dome is tactically amazing but arguably a strategic mistake - it let Israel assume the problem was contained and basically ignore the problem until the terrorists found another, worse attack vector and used that instead.
I think this is actually more sound than something like iron dome because you can have constant improvement, test it in the real world, stakeholders, and everyone who is working at the business and manufacturing side of this thing wants to keep pushing out drones. You’ll have a constant need to turn over old drones and you’ll have a whole software side that needs new training data. The iron dome has to sit there until activated and you can’t really just “Test it out” cheaply because the munitions are so expensive. I do think in the future full deterrence will require stuff like iron dome as well as cheaper small arms fire but at least on the ground level, you could make human soldiers totally obsolete.*
*again, over a 10-20 year period of time. I know tech takes time to develop but if you take it in baby steps, start with the “locate gun by sound” feature which in my mind is pretty easy and doable, work up from there you could convince someone to keep funding you.
I really like the idea of no more school & soft targets, attracting crazed shooters. No mall shootings, no movie theater shootings, office & other work related shootings. It would be amazing. War could definitely be changed, I don’t know enough, to comment on that subject. Well thought out and written, ‘some guy’. Kids certainly change one’s perspective on life and death.
I think a non-lethal version of this would be perfect for school use. Give it a green light laser (you can reliably blind people for about fifteen minutes with a green light laser) a taser and maybe some smoke canisters and you’d have an immediate hard response for school shooter and public scenarios.
And it may be possible to show staff how to use this laser technology. Staff & school security teams wouldn’t need the assistance of police. Or were you thinking this would all be done by the laser Army you described? Either way it works. Zero casualties.
This all needs a long development period with some less impressive stuff at the start. I think you could very realistically train a model to point at a person shooting a gun based on sound. You could probably do that pretty quickly. That requires the shooter to shoot at least one bullet. Still, version one that’s better than where we are. Next you teach it to look for faces and then again to strobe a laser light across the eyes. Still very doable. Better than what we have now because it can respond right away.
Tesla vision chips I think have changed the game here. You can look for guns now in something like real time and you can look for guns before they fire and prime the systems to respond. That’s version two but probably more expensive since those chips aren’t produced at enough scale.
We don’t have drone tech good enough right now to have them just fly at people but I think we could make that tech of the military got involved.
Is the ultimate goal of war peace, or is it freedom?
I’m not a “just existing is enough” guy. Freedom is a condition of a meaningful peace. Stickies posts has my controls for that.
I think I understand the general premise of your argument and appreciate the appeal, but something still doesn’t sit well with me with regard to how much freedom this eliminates from the world. I’m trying to think it through and happy to have your feedback. First concern is that this tech would eventually get into the hands of people who have different moral and ethical standards, and who would apply it in a way that would result in a tyrannical dystopia with no possibility for meaningful rebellion. The second concern is when I take the idea further and extend the possibilities: what if AI can get accurately predictive enough to determine when the bad guy was about to pull the trigger and initiate a counter attack before he actually does. Beyond that, what if it can learn to predict his actions based on a real time brain scan, and stop him once he’s had just the thought of attack. A bit of Minority Report science fiction... but maybe not?
(Not sure what stickies posts are)
My plan to fix the government
https://extelligence.substack.com/p/introduction-to-a-boring-utopia
Someone, somewhere is going to do something like this with drones. I would prefer it be a freedom loving people who can code their values into the system.
I wish this were possible, but I'm pretty sure it's not, at least not in the near future.
* Ordinary bullets move at around 3000 feet per second, more than twice the speed of sound. By the time the sound reaches the microphone, that bullet will have already hit its target. Light is easily blocked by obstacles, and even if the camera does see a bullet, there's less than a hundredth of a second to do anything about it. You would need at least two frames of data to do any speed detection (much less figuring out 3d space from it), and while I could imagine the tech getting good enough at some point to calculate it in time, swivelling a gun around to fire in that amount of time is basically impossible.
* Hamas rockets are extremely cheap to make, and are often fired on timers, so that the terrorists are gone before it fires. They use grenades, bombs, drones of their own, whatever weaponry they can smuggle in via the tunnels to Egypt. These don't have easy counters.
* Drones are very expensive, and even if they're very high-end, they cannot move fast enough to dodge bullets. Easier would be to track direction of motion to move out of the line of fire in advance, but that would still probably be very difficult. These things don't have great manoeuvrability.
* Tunnel openings tend to be inside buildings (eg, into an inconspicuous child's bedroom, under the bed), and there's no easy way to locate the entrances.
I do like the general line of thinking though. Hopefully something good can come of it.
Would you be okay if I used this as part of another post? I’ll try to do a quick reply but high-level, I try to lay out a plausible vision more than the particulars when I write a post like this. I usually have a lot more detail up in my head than I have time to write down. I know you show up with knife and fork ready for the steak and I’d love to give the steak, but most of my time comes in five minute increments these days and it’s very hard to do anything long unless I happen to wake up extremely early one day. That said, I super value your time and don’t want to make too many excuses.
High level on stuff I think you agree on but didn’t mention that I’d like to point is still really big and valuable: it’s possible to make an AI model that would confer this level of marksmanship and targeting to an autonomous system and having the military deliberately go out to create a dataset to do so is a valuable use of taxpayer dollars and something that an organization on scale of a national military could reasonably accomplish. Just that alone is huge.
On timing, bullet speed, turret speed, etc.
I call this the “First Bullet Problem.” There’s lots of depth of field stuff with cameras I *think* you could do to address this. Kind of. That said, I’m almost certain we’ll never get to a scenario where you’d be able to shoot the first one of these bullets out of the air before it could hit its target. By the time the first bullet has struck the counter shots will have had time to take a predictive aim for where the next volley is going to be. But the problem isn’t really just “shoot a bullet with a bullet.” I think that’s what you could do for things like ship to ship battle and probably over long range for military engagements like we have in Ukraine but up close urban warfare is going to be tricky. The underlying problem statement is “I need to move an air column about a foot over to the left or right while a bullet is traveling through it.” If you can just make a very quick decision that a bullet is incoming and about to hit someone, you could do something like having an explosion of condensed air accomplish something like that. The particulars are of course challenging but I still think you could do something like that even if someone had to wear a really crazy personal “wind funnel and compressed air cartridge” device. I would see this entire thing as a ten or fifteen year project starting with small things like the laser pointer and building up to full scale drone manufacture.
On Hamas rockets being very cheap, timers, etc. This is one where I think quantity creates quality. If you have thousands of basically cheap cell phones flying over Gaze taking pictures at fairly quick intervals you can see who is going where to do what and then use that to target your drone swarms. You could also do something like have a thousand 19 year old military guys carry sensors around in a tunnel until you could teach a drone to fly inside of one.
On drones being expense: agreed. You have to develop specific tech for this and then you have to scale it. My point isn’t necessarily that 1,000 is the exactly target. It’s more that if you can move toward the extreme cheap end that you get to make easy trade offs between sacrificing equipment to save lives and you also have such overwhelming force your enemy even in his own mind has to admit you cannot be defeated. I agree on the bullet dodging and didn’t mean to imply they would dodge a bullet already in fire. Again, a first bullet problem. If you have line of sight on the barrel, you have a model that has the drone take an evasive maneuver. The 19 year old dudes in the military are going to be doing a lot of stuff to create this training data and at any given time there will be a few thousand dudes working at Gun Lab to make this data set.
On tunnel openings being inside buildings. Make a few drone variants that bust open doors. This is an engineering problem and can be overcome. The swarm doesn’t have to be completely homogenous, just the mass volume should be a certain and cheap type.
Do this for ten, fifteen, or twenty years and you end up with something much different than we have now.
In summary: we are moving into a world where you can start certain actions based on “wake sounds” and “wake signs” and we should rethink generating data instead of saying “wouldn’t it be nice if we had this data?” You can remove the human risk of your own soldiers completely from these enforcement decisions leading to more humane outcomes. Even the lower fidelity, less capable versions of this could prevent some kids from getting blown apart. I think I have a guess as to who you are behind the anon but even if I’m wrong I’d bet quite a lot of money you are the type of person who would like to do everything possible to save those kids.
This raises a secondary issue: if you can make a perfect system like this, that's great. A partially functioning system, or one that the enemy can eventually learn to bypass, is probably a net negative.
This relates to the issue where iron dome is tactically amazing but arguably a strategic mistake - it let Israel assume the problem was contained and basically ignore the problem until the terrorists found another, worse attack vector and used that instead.
I think this is actually more sound than something like iron dome because you can have constant improvement, test it in the real world, stakeholders, and everyone who is working at the business and manufacturing side of this thing wants to keep pushing out drones. You’ll have a constant need to turn over old drones and you’ll have a whole software side that needs new training data. The iron dome has to sit there until activated and you can’t really just “Test it out” cheaply because the munitions are so expensive. I do think in the future full deterrence will require stuff like iron dome as well as cheaper small arms fire but at least on the ground level, you could make human soldiers totally obsolete.*
*again, over a 10-20 year period of time. I know tech takes time to develop but if you take it in baby steps, start with the “locate gun by sound” feature which in my mind is pretty easy and doable, work up from there you could convince someone to keep funding you.
Nice text
Ty!
I really like the idea of no more school & soft targets, attracting crazed shooters. No mall shootings, no movie theater shootings, office & other work related shootings. It would be amazing. War could definitely be changed, I don’t know enough, to comment on that subject. Well thought out and written, ‘some guy’. Kids certainly change one’s perspective on life and death.
I think a non-lethal version of this would be perfect for school use. Give it a green light laser (you can reliably blind people for about fifteen minutes with a green light laser) a taser and maybe some smoke canisters and you’d have an immediate hard response for school shooter and public scenarios.
And it may be possible to show staff how to use this laser technology. Staff & school security teams wouldn’t need the assistance of police. Or were you thinking this would all be done by the laser Army you described? Either way it works. Zero casualties.
This all needs a long development period with some less impressive stuff at the start. I think you could very realistically train a model to point at a person shooting a gun based on sound. You could probably do that pretty quickly. That requires the shooter to shoot at least one bullet. Still, version one that’s better than where we are. Next you teach it to look for faces and then again to strobe a laser light across the eyes. Still very doable. Better than what we have now because it can respond right away.
Tesla vision chips I think have changed the game here. You can look for guns now in something like real time and you can look for guns before they fire and prime the systems to respond. That’s version two but probably more expensive since those chips aren’t produced at enough scale.
We don’t have drone tech good enough right now to have them just fly at people but I think we could make that tech of the military got involved.