Increased immigration of all kinds would help with scale. I don’t know that industrial policy is the key for defense technology. Drones or drone killers are not very high tech, for the cost of an F35 or F22 the defense department could contract for millions of these and our military industrial complex would quickly ramp up and produce them. The small electric motors and batteries are commodity items as are the CPUs, communications equipment and mechanical parts. The software would not be difficult to develop. It’s mainly a problem of old style thinking that large supersonic manned fighter jets are the backbone of security, while they are mostly useless for drone swarms. Also if we are using expensive Sidewinder missiles to shoot down every small balloon, I would expect China or even North Korea to respond by launching thousands of decoy balloons weekly just to keep seeing the show go on.
I would push back a little on the relevance of China's supposed lead in advanced batteries.
China is able to manufacture commodity lithium-ion batteries in vast quantities more cheaply than anywhere else in the world. But you don't need vast quantities of batteries to build a lot of small drones, and big drones are going to be powered by combustion engines or possibly fuel cells because of the order of magnitude improvement in capability a combustion engine offers compared to electric power.
Looking back to your 2014 QZ post, I'm not at all sure you were right about the end of the age of the gun, and the sunset of the place of the infantryman on the field of battle. Infantry are needed not just because they are cheap and easy to train, as you assert - and which I'd argue isn't even true (viz: Russia's mobilisation), but because they are a vital part of a combined arms team.
Russia's losses in tanks, IFVs, AFVs, basically any kind of vehicle throughout the war is largely because they didn't have enough competent infantry riding along to provide a shield for their armour. e.g. the NLAW has a max range of 800m. A strong infantry shield should be used to identify and suppress ATGM teams at that distance -- not all, of course, but certainly much better than Feb/March '22.
I'd also argue (on this narrow front of the place of the infantry) that countermeasures will be developed, or tweaked shortly to make drones just another tool on the battlefield. You can already see Ukraine's TB2s largely left the battlefield after the early months of the war once RU finally sorted out their AD network. More broadly, SPAAGs like the Gepard were decommissioned by most militaries because of the introduction of MANPADS, and because SPAAGs were less effective after the introduction of jet aircraft. But against low and slow drones, SPAAGs or DEWs will return to all combined arms units over the coming years.
Men with guns are still needed to clear trenches, call in artillery, hold territory, and I think we're a fair way away from that changing just because of the proliferation of aerial drones.
Modern warfare means combined forces. Everything has to work together.
P.S. Does anyone remember that 15-20 year old video of a Syrian tank heading down a street being followed by half the town? It was almost like something out of a cartoon, but the soldiers in the tank didn't realize that they had an entourage. Reportedly, the tank was destroyed by an RPG shortly after it turned the corner.
I thought all the unexplained UFO sightings by military pilots involved objects that exhibited capabilities beyond those of current aircraft, like high speeds and abruptly changing direction. Doesn't sound like balloons. Any explanation for the discrepancy? Was the whole UFO narrative just US govt disinformation (as is being proposed at WSJ Op ED)? Also, "cylindrical" describes anything from a roll of paper towels to the lid of a jar. What exactly are we talking about here?
If and object is actually accelerating that fast, it's not hard to detect - it'd be giving off a massive thermal signature and sonic booming left and right. Therefore it's not actually moving that fast and instruments are fibbing.
There's a lot of different ways to create the illusion of something fast. One way is the object is much closer and smaller than it appears. Another way is to send out a signal which appears to have a doppler shift. Another way is like a display of something moving fast (i.e. a visible object at one location vanishes just as similar object at a far away location appears).
If there are drones being tested for spying on sensors and/or confounding sensors, those would be good candidates for things that *look* like they have capabilities beyond current aircraft, but don’t actually.
Many of those can be explained by the motion of the camera and the angle of capture. Search for Mick West on YouTube or Twitter for analysis and debunking of UAP videos/photos.
At first I wanted to quote him here, but that'd be absurdly verbose, so I asked ChatGPT to summarize:
> The text discusses the impact of the US's restrictions on the sale of GPUs to China on the Chinese chip industry. The author argues that China is in a tough position as there does not seem to be any significant bailout plan for the Chinese chip industry. This neglect could be due to the Chinese President Xi Jinping choosing to let the industry take its lumps rather than bail it out.
> Gwern also highlights Xi Jinping's lack of interest in AI and chips and his emphasis on legible atom-heavy scientific projects. Furthermore, his reign has been marked by hostility towards and suppression of "big tech."
> Gwern notes that the chip ban does not necessarily doom Chinese AI research but makes it more prone to irrelevancy, brain drain, and Galapagos syndrome. The damage will be subtle, with researchers continuing to use existing clusters or assemble them one GPU at a time. The author argues that from an AI research perspective, this is an insane position that takes an L on one of the most important future technological capabilities.
> Gwern also notes that mistakes related to chips may be hard to see. Xi Jinping's lack of technical competence or expertise in the relevant area and his hatred of software and everything to do with it make it easy for him to neglect chips right now. This neglect may seem reasonable from his perspective as chips aren't that important and don't seem to be in that much worse trouble than anything else.
> Finally, Gwern argues that from a scaling perspective, the sane policy is for Xi to dump in as much subsidies as needed until China catches up in semiconductors with the US and its allies. Accepting a permanent straitjacket around GPUs and a tightened noose is tantamount to admitting defeat and abandoning the future to other countries.
“In battery manufacturing and commercial drones it will likely fail to catch up, because the U.S. economy is uniquely bad at subsidizing capital-intensive low-value-added manufacturing industries. But it will probably get a foothold, and its Asian allies will give it some added heft. Thus, it is very important that the U.S. maintain a respectful, constructive, cooperative relationship with South Korea and Japan, rather than alienating them with go-it-alone trade policies that pit the U.S. against its allies.”
The cheaper, safer silicon anode Ali-ion battery is being manufactured by Enovix. The smaller this battery is modeled, the greater the energy density. The current battery -- designed for wearables, smart phones, tablets, laptops -- has 100% more energy density than any Li-ion battery on the market. It also has Brakeflow technology, which mitigate s thermal runway, caused by short circuits and resulting in all-too-common fires that can reach 1,000F in a couple of minutes. ENVX’s Fab 2 is calendared for late 2023/early 2024 somewhere in the PAC Rim, related to Samsung as a likely client/partner. The Fab 3, which will produce EV batteries is calendared for 2025, location to be determined. Elon Muck has publicly stated: “A silicon anode is the way forward for lithium-ion batteries.” Co-Manufacturing Agreements with Samsung, et alia can scale-up relatively fast. I remember when the world doubted ASML could develop/commercialize EUV.
Good news: The battery chart lists Panasonic as Japanese, but the Nevada Gigafactory produced 39GWh in 2021, so it appears that at least half of that capacity is in fact American. Tesla has very high ambitions for battery production in the US and Germany (in the 100s of GWh per year).
I keep seeing 5G mentioned in various technology advancement/China competition articles and it's presence always vaguely surprises me. Why does it belong in the diagram with all of the other things? Why is it considered to be so important?
Tyler Rogoway said the stigma of talking about UFOs hindered our military response, but reading his work I had the opposite take-away! The fact that our military indulged the UFO categorization - a non-explanation explanation - rather than taking for granted these are terrestrial objects with erroneously reported physical properties, was horrifying to observe. Rational people were trying to communicate that these were likely electronic warfare tests, and that perspective was sidelined as overconfident and closed-minded.
My only consolation was that the idea that maybe military knew what was going on but was strategically playing dumb to avoid tipping off it's adversary. Increasingly, it looks like the military was just being dumb (because it sounds like they've gotten wise to it now).
Had to read this a few times to understand because you (and perhaps Tyler) write "UFO" when you mean aliens. It's completely correct to call them UFOs, but all talk of aliens is a dangerous distraction.
Good catch. Tyler is very careful in his words, but I mostly used UFO to mean aliens.
Or rather specifically I mean "treating UAPs as though they might be an alien". It's not that we should fool ourselves that we can instantly identify everything, it's that invoking aliens is an excuse not to identify things. And it encourages identification by credulous pattern matching, rather than questioning assumptions and using Bayesian reasoning.
Hi Noah - as a long time subscriber and COO of a data science company specifically focused on this space, I am glad to see you covering this topic. If you find yourself in need or want of more detail please feel free to ping me at m@enigmalabs.io.
Noah, a simple question about big balloons (I'll go dig myself, but maybe you've already found this out): how long do they last and how do they end? Do they pop when they get too high (simple physics question), or do they die on command where they will disappear in an ocean or over friendly soil to be collected?
The Chinese spy balloon had a large bus sized heavy payload, so the balloon envelope would need to be much larger and stronger than for typical weather balloons which have a small few kilogram payload.
Read up on Google's "Project Loon". Their balloons could stay up for more than 100 days. Sadly closed down two years ago. As far as I know they never carried cameras and they didn't operate over countries that didn't want them.
If they didn't already, the US military should try to hire some of the Loon team.
You don't say so explicitly but implicitly every breathless news story for the past decades about Amazon delivery drones bringing consumers their snacks *even faster* has been part of the problem attitudinally and culturally. Yes to US-made Skydio doing new things and for emphasizing infrastructure inspection, increasingly necessary. And hardware matters as much as software, just ask ChatGPT.
Hey Noah. Skilled immigration, industrial policy, and DARPA like programs seem to be the only way the US can compete.
But China wins with sheer scale and low cost manufacturing.
I know you mention this in the post. But any appetite to write more about the policy decisions the US has to make to catch up?
Essentially - what do we do about it?
Increased immigration of all kinds would help with scale. I don’t know that industrial policy is the key for defense technology. Drones or drone killers are not very high tech, for the cost of an F35 or F22 the defense department could contract for millions of these and our military industrial complex would quickly ramp up and produce them. The small electric motors and batteries are commodity items as are the CPUs, communications equipment and mechanical parts. The software would not be difficult to develop. It’s mainly a problem of old style thinking that large supersonic manned fighter jets are the backbone of security, while they are mostly useless for drone swarms. Also if we are using expensive Sidewinder missiles to shoot down every small balloon, I would expect China or even North Korea to respond by launching thousands of decoy balloons weekly just to keep seeing the show go on.
I would push back a little on the relevance of China's supposed lead in advanced batteries.
China is able to manufacture commodity lithium-ion batteries in vast quantities more cheaply than anywhere else in the world. But you don't need vast quantities of batteries to build a lot of small drones, and big drones are going to be powered by combustion engines or possibly fuel cells because of the order of magnitude improvement in capability a combustion engine offers compared to electric power.
They also don't care about the environmental costs associated and we, well, do
Wow! Just read the Warzone link https://substack.com/redirect/6f6f8945-538d-4fc4-bd52-8465a82acd9a?j=eyJ1Ijoicjg1eHQifQ.UuJTQp-A7blYGBImJqOMEiadKQvhrBfbn5tbPOPTq8Y that pretty much explains the whole UFO thing. Noah, thanks for bringing this source to our attention!
Yes, that is a great source I haven’t heard of.
Looking back to your 2014 QZ post, I'm not at all sure you were right about the end of the age of the gun, and the sunset of the place of the infantryman on the field of battle. Infantry are needed not just because they are cheap and easy to train, as you assert - and which I'd argue isn't even true (viz: Russia's mobilisation), but because they are a vital part of a combined arms team.
Russia's losses in tanks, IFVs, AFVs, basically any kind of vehicle throughout the war is largely because they didn't have enough competent infantry riding along to provide a shield for their armour. e.g. the NLAW has a max range of 800m. A strong infantry shield should be used to identify and suppress ATGM teams at that distance -- not all, of course, but certainly much better than Feb/March '22.
I'd also argue (on this narrow front of the place of the infantry) that countermeasures will be developed, or tweaked shortly to make drones just another tool on the battlefield. You can already see Ukraine's TB2s largely left the battlefield after the early months of the war once RU finally sorted out their AD network. More broadly, SPAAGs like the Gepard were decommissioned by most militaries because of the introduction of MANPADS, and because SPAAGs were less effective after the introduction of jet aircraft. But against low and slow drones, SPAAGs or DEWs will return to all combined arms units over the coming years.
Men with guns are still needed to clear trenches, call in artillery, hold territory, and I think we're a fair way away from that changing just because of the proliferation of aerial drones.
Modern warfare means combined forces. Everything has to work together.
P.S. Does anyone remember that 15-20 year old video of a Syrian tank heading down a street being followed by half the town? It was almost like something out of a cartoon, but the soldiers in the tank didn't realize that they had an entourage. Reportedly, the tank was destroyed by an RPG shortly after it turned the corner.
I thought all the unexplained UFO sightings by military pilots involved objects that exhibited capabilities beyond those of current aircraft, like high speeds and abruptly changing direction. Doesn't sound like balloons. Any explanation for the discrepancy? Was the whole UFO narrative just US govt disinformation (as is being proposed at WSJ Op ED)? Also, "cylindrical" describes anything from a roll of paper towels to the lid of a jar. What exactly are we talking about here?
If and object is actually accelerating that fast, it's not hard to detect - it'd be giving off a massive thermal signature and sonic booming left and right. Therefore it's not actually moving that fast and instruments are fibbing.
There's a lot of different ways to create the illusion of something fast. One way is the object is much closer and smaller than it appears. Another way is to send out a signal which appears to have a doppler shift. Another way is like a display of something moving fast (i.e. a visible object at one location vanishes just as similar object at a far away location appears).
If there are drones being tested for spying on sensors and/or confounding sensors, those would be good candidates for things that *look* like they have capabilities beyond current aircraft, but don’t actually.
Many of those can be explained by the motion of the camera and the angle of capture. Search for Mick West on YouTube or Twitter for analysis and debunking of UAP videos/photos.
From day one, the world has been run by and for sociopaths. The only thing that's changed is the weaponry at their disposal.
AI isn't really contested; China is f**ed. I recommend looking at Gwern's analysis here: https://www.lesswrong.com/posts/oBTkthd7h8sDpkiu2/analysis-us-restricts-gpu-sales-to-china?commentId=ugy4wdJisPHezLi35#ugy4wdJisPHezLi35
At first I wanted to quote him here, but that'd be absurdly verbose, so I asked ChatGPT to summarize:
> The text discusses the impact of the US's restrictions on the sale of GPUs to China on the Chinese chip industry. The author argues that China is in a tough position as there does not seem to be any significant bailout plan for the Chinese chip industry. This neglect could be due to the Chinese President Xi Jinping choosing to let the industry take its lumps rather than bail it out.
> Gwern also highlights Xi Jinping's lack of interest in AI and chips and his emphasis on legible atom-heavy scientific projects. Furthermore, his reign has been marked by hostility towards and suppression of "big tech."
> Gwern notes that the chip ban does not necessarily doom Chinese AI research but makes it more prone to irrelevancy, brain drain, and Galapagos syndrome. The damage will be subtle, with researchers continuing to use existing clusters or assemble them one GPU at a time. The author argues that from an AI research perspective, this is an insane position that takes an L on one of the most important future technological capabilities.
> Gwern also notes that mistakes related to chips may be hard to see. Xi Jinping's lack of technical competence or expertise in the relevant area and his hatred of software and everything to do with it make it easy for him to neglect chips right now. This neglect may seem reasonable from his perspective as chips aren't that important and don't seem to be in that much worse trouble than anything else.
> Finally, Gwern argues that from a scaling perspective, the sane policy is for Xi to dump in as much subsidies as needed until China catches up in semiconductors with the US and its allies. Accepting a permanent straitjacket around GPUs and a tightened noose is tantamount to admitting defeat and abandoning the future to other countries.
“In battery manufacturing and commercial drones it will likely fail to catch up, because the U.S. economy is uniquely bad at subsidizing capital-intensive low-value-added manufacturing industries. But it will probably get a foothold, and its Asian allies will give it some added heft. Thus, it is very important that the U.S. maintain a respectful, constructive, cooperative relationship with South Korea and Japan, rather than alienating them with go-it-alone trade policies that pit the U.S. against its allies.”
The cheaper, safer silicon anode Ali-ion battery is being manufactured by Enovix. The smaller this battery is modeled, the greater the energy density. The current battery -- designed for wearables, smart phones, tablets, laptops -- has 100% more energy density than any Li-ion battery on the market. It also has Brakeflow technology, which mitigate s thermal runway, caused by short circuits and resulting in all-too-common fires that can reach 1,000F in a couple of minutes. ENVX’s Fab 2 is calendared for late 2023/early 2024 somewhere in the PAC Rim, related to Samsung as a likely client/partner. The Fab 3, which will produce EV batteries is calendared for 2025, location to be determined. Elon Muck has publicly stated: “A silicon anode is the way forward for lithium-ion batteries.” Co-Manufacturing Agreements with Samsung, et alia can scale-up relatively fast. I remember when the world doubted ASML could develop/commercialize EUV.
Good news: The battery chart lists Panasonic as Japanese, but the Nevada Gigafactory produced 39GWh in 2021, so it appears that at least half of that capacity is in fact American. Tesla has very high ambitions for battery production in the US and Germany (in the 100s of GWh per year).
I keep seeing 5G mentioned in various technology advancement/China competition articles and it's presence always vaguely surprises me. Why does it belong in the diagram with all of the other things? Why is it considered to be so important?
"The Pentagon is trying to develop huge swarms of thousands of autonomous combat drones. China is creating a drone mothership."
G*d dammit haven't any of these people played _Horizon: Zero Dawn_?
Tyler Rogoway said the stigma of talking about UFOs hindered our military response, but reading his work I had the opposite take-away! The fact that our military indulged the UFO categorization - a non-explanation explanation - rather than taking for granted these are terrestrial objects with erroneously reported physical properties, was horrifying to observe. Rational people were trying to communicate that these were likely electronic warfare tests, and that perspective was sidelined as overconfident and closed-minded.
My only consolation was that the idea that maybe military knew what was going on but was strategically playing dumb to avoid tipping off it's adversary. Increasingly, it looks like the military was just being dumb (because it sounds like they've gotten wise to it now).
Had to read this a few times to understand because you (and perhaps Tyler) write "UFO" when you mean aliens. It's completely correct to call them UFOs, but all talk of aliens is a dangerous distraction.
Good catch. Tyler is very careful in his words, but I mostly used UFO to mean aliens.
Or rather specifically I mean "treating UAPs as though they might be an alien". It's not that we should fool ourselves that we can instantly identify everything, it's that invoking aliens is an excuse not to identify things. And it encourages identification by credulous pattern matching, rather than questioning assumptions and using Bayesian reasoning.
Hi Noah - as a long time subscriber and COO of a data science company specifically focused on this space, I am glad to see you covering this topic. If you find yourself in need or want of more detail please feel free to ping me at m@enigmalabs.io.
Noah, a simple question about big balloons (I'll go dig myself, but maybe you've already found this out): how long do they last and how do they end? Do they pop when they get too high (simple physics question), or do they die on command where they will disappear in an ocean or over friendly soil to be collected?
Sigh, even with a smattering of interesting facts from https://www.nytimes.com/2023/02/14/science/weather-balloons-stratosphere.html
the answer is All Of the Above.
US weather balloons are designed to reach 20-30 miles, do their thing and then pop, coming down as balloon dust. but it doesn't always work that way.
So the atmosphere will soon have floating junk clouds like the Pacific Ocean??
The Chinese spy balloon had a large bus sized heavy payload, so the balloon envelope would need to be much larger and stronger than for typical weather balloons which have a small few kilogram payload.
Read up on Google's "Project Loon". Their balloons could stay up for more than 100 days. Sadly closed down two years ago. As far as I know they never carried cameras and they didn't operate over countries that didn't want them.
If they didn't already, the US military should try to hire some of the Loon team.
You don't say so explicitly but implicitly every breathless news story for the past decades about Amazon delivery drones bringing consumers their snacks *even faster* has been part of the problem attitudinally and culturally. Yes to US-made Skydio doing new things and for emphasizing infrastructure inspection, increasingly necessary. And hardware matters as much as software, just ask ChatGPT.