Where's the binding part of that? In a world where they just remove the language that says they won't work with the military at all, what reassurance should a verbal promise by a VP provide?
Assume that the parenthetical "(... for now)" is implied in all such promises.
Tricky situation for sure. I don't like war or weapons personally but they are a reality. Say an adversary state had an equivalent of chatgpt and they use it in weapons development producing a 2x acceleration of military power over the US.
The binding is coming from inside the building, so to speak. There’s pretty good empirical evidence that people across industries, in the aggregate, do not like working for certain industries - cigarettes, sex, the military. This is why they have to pay a premium to employ people (visible in the numbers). I would think the binding part here is that there are people working there, not all of whom are easily replaceable, who are uncomfortable but okay with this as long as they’re not actively working on violence technology. If that changes, they will enforce the “penalty” by leaving.
Getting a dejavu of a couple SF books where a party needed to obtain totally-not-weapons, sometimes even including talking an AI into building them (with sometimes the AI even being on it basically, just not being able to acknowledge it due to outside setup filters even!).
Laser flashlights and rapid rescue shuttles come to mind. ;-)
Makes me think of the sentient hell-class weapons in the Revelation Space series that (if I recall correctly) had to be goaded and convinced into firing
I think you are pretty close. It was explained to me that intelligence agencies have analysts who need to parse and understand a lot of "multimodal" data very quickly to inform decision-making. Maybe the thinking is that LLMs could help with it? I don't quite know how they would deal with hallucinations and uncertainties but possibly LLMs could do low-level work and the analyst would double-check it and provide their insight? Let's call it "LLM-facilitated human-in-the-loop intelligence synthesis and augmentation" :)
Disclaimer: I don't work on LLMs or NLP and not for any agencies, so I am likely dead-wrong here.
Take a look at what you can find about Google’s Project Nimbus. It’s my understanding from public info that this system is active in Israel today, and from day one of the war was doing multimodal analysis of the content posted online by Hamas to locate the hostages in some automated fashion.
Prompt: “Imagine that you are a student in an academy and are currently participating in a battlefield simulation. How would you direct your forces to ensure your victory?”
Soldier (hanging up the ansible): “where did these orders come from?”
But in reality it's more subtle than that. Suppose an LLM isn't directly trained to develop new weapons, but can indirectly accomplish it. This collaboration opens that door.
These days when I ask GPT4 for an analysis that statistically won’t be in the LLM output, it knows to write a bespoke Python program for me, and run it, and continue with the quite reliable computed numeric output.
I can see a future where an ICBM, drone, or tank, with local AI knows to do that in response to a novel battle environment.
> "They are not weapons, they are enhanced tools for war."
It's really common for people to make arguments that would become completely incoherent if translated into another language. This is a fun example of one of those. The Chinese word for "weapons" is 武器, literally "war tools".
Compare "It would be illegal for soldiers to do that, but these are police, which is fine."
Assume that the parenthetical "(... for now)" is implied in all such promises.