According to CivDiv's channel about 70% of the drones used are still RC FPV drones. They are cheaper longer range and have a slightly simpler supply chain. Fiber optic cable prices have exploded because of the war.
It's not as easy as plopping a new SW on it. Have you wondered why, despite this conflict dragging on for this long, autonomous targeting is still not common? It's just much harder to do, and doing it reliably is actually super hard.
Hw access is also not as easy as you think. Try getting a Pixel 9 class SoC board that you can actually dev stuff on, you'd be surprised how hard it is to get such a thing. And when you get it, it will probably be 2x as expensive for just the board.
Oh, put the Pixel 9 on it then. Have you tried doing optical flow on rolling shutter cameras?
Then, whatever solution you get, you need sufficient volume and reliable supply of the thing. Oh, you can't find the chip anymore? Good luck porting the stack you have since the pipeline is different.
Whereas these rc FPV drones you can basically build what what you find in the back of your drawer.
I think one way communication should be sufficient together with a shared secret. It's probably less reliable compared to two way communication, but it should satisfy the criteria. IR visual pattern propagation will respect this. You don't have radio spectrum that is too prone to jamming, and if the drones are visual navigation/targeting based, if they can see you to acquire the you as a target, they can identify you as a friend as long as you can ensure emitters are not occluded from your side.
It's not perfect, but it's cheap and it should work.
Give me a couple hundred k and I can probably build the prototype, haha.
I would have expected the right wing parties to be against this, somehow, but... Nope, just checked my country's votes, and nationalists just love cc1.0 it seems.
I think you can do this relatively simply with infrared led lights. Imagine "Please don't kill me" remote controls. You are a bit more visible when broadcasting, but presumably you can be selective when turning it on.
You need hardware for these things. High speed image processing and high precision control and global shutter cameras. You can't just slap a raspi on it with any run of the mill camera and call it a day.
A bit amusing. I took Anna Karenina off Gutenberg and used that as test data for some of my flashing algorithms. I called it the "Anna test". I could have used random data, but where's the fun in that? Besides, during dev, structured text showed the kind of error I got much better than random data would have.
Super cool, what did you make? How large was the factory? How did it work out? One observation from my experience, the closer you are to production, the more stressful things are. But probably scale changes experience. For me, working on stuff that entered automotive production lines, anything that made the line stop or go slow was insanely stresful.
I would be a bit more charitable to OP here because it's not their fault, it's the Roborock app's fault. They convey in their UI the information of total surface cleaned in xxKm2 with the K bolded. But the km doesn't mean square kilometer it means thousands of square meters apparently. It's easy to just take the information at face value as OP probably did here. He went in app, saw the number and the very weird unit for the info and just posted it.
On the other hand, you are limited by having CD's which compared to streaming stuff from Spotify is much much less convenient, take up space and you need to create/buy them, your playlist don't synch up with your other devices. CD experience is much less streamlined than a smartphone. Perhaps nostalgia makes them seem cooler for you, but I am not sold.
I personally am somewhere between you and the author. I don't check _all_ the intermediary steps, but I do try to understand what it's doing [1] and follow the process. Mostly I let it do the changes itself without supervision at each step but when a coherent "chunk" of work is done, I go through it really thoroughly. In almost 90% of the cases after a chunk is done some adjustments are needed.
I find broad architectural design to be _better_ if you follow along in the process because you better understand the direction it's going earlier and you can shift the high level direction much earlier. Even if you check its steps, you can ask it for its take on high-level architectural aspects along the way, no problem. I think personal touch matters a lot though, because I naturally ask it and try to get the big picture image.
[1] I actually find it really instructive what tooling it uses to tackle a problem, I got to become a much better console user because of it
I'm not so sure. I think you can, you just need to intentionally drill into what you don't understand and it's exhausting. What I do agree with though is that I can't seem to build the ability to build it myself the same way as I would if I wrote it.
For example, I know my mental model works because I know what change I should do in order to get an effect and when I do the change, I get what I expect. But if I were to build myself something similar, I could not build it because the approach is somewhat out of my reach, I know it sounds weird, but it's hard to explain.
I am a bit confused which part you disagree with specifically. Reading AI responses and reviewing code seems to be what you propose as well.
Your example with MLIP is something that would not be prevented by this approach, during the planing phase, it would surface.
I guess the devil is in the details and the way you prompt it for starting the task matters.
But IMO you absolutely need to check the output, need to engage with what the model is doing, need to probe why something is built the way the model tries to build it.