The real reason why states are pouring billions into AI isn't to generate silly emoji, help people rewrite their emails, or make a web app without learning how to code.
It's for autonomous drone warfare.
Invading a neighboring nation doesn't seem too appealing if it means sending tens/hundreds of thousands of your young people (a resource that's getting scarcer and scarcer most everywhere in the world) to the meat grinder. Millions of drones though...
> Not sure if many noticed, but the English put out by Chinese teams is often perfect these days
how are these two sentences connected?
I have learnt English for over 10 years, and so have most Chinese students in the past 20 years. This is a part of the Chinese education system. For every Chinese student in top universities, there's a 99.99% chance that he/she is very good at English.
how is this connected to "Online propaganda is insanely powerful too"
I've lived in china for nearly a decade. My family is Chinese. Just because some Chinese have great English, does not mean that all Chinese have great English, and you'd often see releases with fragmented or otherwise poor English. Now with LLMs, everyone can correct the English in their posts no matter their proficiency level.
How does this relate to propaganda? Propaganda is often posted to look like organic and natural conversations online rather than government bulletins. It's a lot easier to do that if you are perfectly proficient in the language. LLM tech enables this.
This is not particular to the Chinese by the way, all the major players are doing this. Like I said, my family is Chinese, and my intent is not to convey a negative meaning.
Online propaganda since 2014ish led the US to its current state, which is incredible. It turns out you can destroy a country with nothing more than words.
I think that the autonomous drone warfare is more of a near-term milestone than the goal. The goal itself seems to be AGI and the advances in the basic research and engineering that it brings.
Surely it'd be a lot more cost-effective to train up people to fly drones remotely, or is the issue more that they can't secure the communications to control them remotely?
>Invading a neighboring nation doesn't seem too appealing if it means sending tens/hundreds of thousands of your young people (a resource that's getting scarcer and scarcer most everywhere in the world) to the meat grinder. Millions of drones though...
Sorry but this is alarmist bullshit. What keeps countries from going to war is not the cost in soldiers; it's the cost of retaliation on its home soil. Indeed, countries that don't anticipate the enemy to meaningfully retaliate on its civilians are quite trigger-happy to engage in various small wars around the world.
Bro ai via programming with people is already good enough to fight wars with drones in the air. You don’t even need LLMs or machine learning black boxes.
Basically the reason is because there’s so little noise in the air. If it’s black and metal and flying and doesn’t match the iff signal than simple “if then” logic can verify it’s a bogey.
This is partly true, Automated object detection already exists. If LLMs haven't fundamentally shifted Ukraine war yet then it's plausible it has limited value in progressing drone warfare.
I disagree you can simply automate it via IFF though... basically because IFF is easy grounds for counter drone attacks. Establishing IFF just makes your enemies job easier unless you have heavy anti drone tech which no country does yet.
Automating "this person holding a gun in a given GPS grid is hopefully not holding a stick and is hopefully not a friendly/civilian... and calculate odds they should be executed" is not something easily done by a drone. Especially when there's electronic warfare and drones don't have the luxury of spending lots of time looking.
In Ukraine for example people wear 50 different uniforms because they buy them on the internet or come from random countries. And they carry similar AK rifles. What's an LLM going to do? Kill random people with ~50% accuracy using a very finite supply of drones? Or just have humans fly them from the rear.
Building 1000's of drones capable of taking out armoured vehicles from disposable commodity parts is more important than a 4% improved detection rate of hotdogs.
It's for autonomous drone warfare.
Invading a neighboring nation doesn't seem too appealing if it means sending tens/hundreds of thousands of your young people (a resource that's getting scarcer and scarcer most everywhere in the world) to the meat grinder. Millions of drones though...