This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
I can only assume that you haven't of the Vogons.
I am in no way specialist on how paperclip maximizers work, but to me the actual capabilities of the models are mostly dressing over the underlying tech. And the fact that they use the inefficient human language is mostly artefact of the google babelfish experiments.
To make drones properly lethal you need meshed swarm coordination, ability to create and adjust tactics on the fly, without much centralization in conditions of severe jamming. So they do have incentive to cram as big models as possible in as limited hardware.
Surely that sort of thing is in the domain of the CIA!
You're sort of missing my point - the goal here is to cram the most optimized models - that is, the models that are going to best allow them to do meshed swarm coordination, or whatever. Larger is not always better; better is better.
I have no particular reason to think that the direction LLMs research is currently developing is good at all for that sort of thing. Cramming Astra into a drone is going to result in something that's much slower and less decisive than a simple deterministic script, because LLMs are compute-hungry. That's not to say that neural networks won't be helpful - computer vision is great, for instance.
But LLMs are not the only form of artificial intelligence, and for most military tasks a strong general-purpose LLM is both overengineered and poorly suited for the mission, especially the mission of terminal target determination. For intelligence synthesis, they likely have some promise.
You have the right of it.
The additional issue is data, to a shock of nobody paying attention. Military data is near non-existent. And the data that does exist is low-resolution, or tightly controlled by non-ML/AI folks who have been trained to have negative desire to share (The Security Clearance process over time selects for a type). LLMs are the Ford F150 of ML models. Half the battle is often getting a very specific tank configuration that is actively trying to hide in foliage from the very CV models you are trying to detect it with. You have N=3 samples of this tank sitting on a tarmac somewhere and it turns out the ability of your model to extrapolate that picture to 3.0 GSD satellite or drone footage is abysmal. And that is before it even starts attempting adversarial countermeasures.
There are of course solutions, but they don't work as well as folks imagine they do, including generative image generation. Most of the examples of drone based hits are FPVs or FPVs with a "wait" situation where the drone hovers, detects a general class and is then given approval by a human analyst who manually reviews the vid/image.
Very interesting! I actually thought that open-source imagery would be better resourced than that (although of course it's pretty easy to mess with the profile of a tank with a can of spray paint or a camouflage mesh, which I guess is part of the problem).
Always very heartening to have an industry professional tell me I haven't gone completely off the rails, thanks.
There are what a couple hundred grainy low res videos of Ukr drones targeting Russian tanks + other vehicles? It's well below the threshold to train an ATR model on data quantity, without even getting into quality. Stock images are from non-operational angles (you for the most part aren't level with targets, its an isometric or overhead view, NERF to solve that is expensive and has dubious performance improvements, and having worked in the generation of low-sample military images, it can actually be quite hard to secure exemplars in EO imagery. SAR imagery just sucks, speckle is multiplicative noise which makes it nearly impossible to remove and hard to detect around.
Humans are obviously much easier to target, however then we get into friendly fire and the fairly disturbing optics of AI drones targeting your own soldiers... There are logistic issues around distributing RFID and similar tags to prevent it, that to my knowledge are being worked through. But the main goal of drone warfare is destroying expensive vehicles with cheap drones.
Training datasets for the ATR stuff are being developed but its sort of an incumbent's advantage, with active known defense contractors having most of the customer connections to be able to get the data, and the resources to label it. Unfortunately they are also slow AF, which is why defense tech startups sorta eat their lunch in certain spaces. Transfer learning is something I've seen recently, taking civilian jeeps with SAM-esque models and applying them to humvees etc. It's always much easier to train on your own nations hardware anyway. Though it gives bad optics to Brass, who don't like it, and an open question on whether it will transfer.
For Russian (or American) kit, I figured you would have good most-angle photos from parades, demos, and such. You're right though that these are mostly taken from the human level.
I imagine that another obvious issue distributing RFID tags to your troops is that they can be used by both sides. I realize that in theory they will only respond to coded radio signals that in theory only one side can transmit, but if you're putting them everywhere and then putting the transponder in something as failure prone as a kill-drone your enemy's going to be reverse-engineering your IFF on like day 2 of the war.
I've speculated, based on my knowledge of machine imaging, that we'll see "wartime camo" become a thing in future wars, with "peacetime camo" designed to be something maximally different from operational paint schemes to confuse ATR. Apparently the Russians are already using a variety of disruptive paint schemes in Ukraine. We can only hope it results in making razzle-dazzle great again!
Yes, I've heard that point a lot. I know an RF guy who's trying to do some passive emitter thingy that goes wayyy over my head. No clue how it works. But he has some pithy comment for it that I am forgetting, its quite funny.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Wow my field has made it onto TheMotte, spooky. It's not quite so simple as big LLM models on limited hardware. There are a lot of networking problems on swarm interfacing if you want something that is truly adaptable. A lot of swarm robotics is essentially "dumb" control algorithms with software permutations, because LLM-agent hallucinations are rough + LLM sizes require truly monstrous hardware on drones. There's work on fitting them to Jetson Orin Nanos' but it's a slog just from the amount of compute needed + performance of smaller quantized models etc.
This is generally a case against the LLM + Drone combination, too much extra stuff not needed. Part of the problem with current AI research is that it is all in on LLMs but the branch off of not-LLMs but with LLM-like capabilities for reasoning for tactics or strategy are not there. If you want the reasoning you get poetry on your drone too. It actually goes beyond mesh networks as well, as a mesh network is every node connected to every node. You actually want localized connection networks that do multi-hop message passing. Much more resistant to jamming.
Idk maybe I should write an effort over-post on drone swarms, I love this subject.
raises paw Interested!
More options
Context Copy link
I'd love to read it, if you ever do.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link