Researchers say AI models like GPT4 are prone to “sudden” escalations as the U.S. military explores their use for warfare.
- Researchers ran international conflict simulations with five different AIs and found that they tended to escalate war, sometimes out of nowhere, and even use nuclear weapons.
- The AIs were large language models (LLMs) like GPT-4, GPT 3.5, Claude 2.0, Llama-2-Chat, and GPT-4-Base, which are being explored by the U.S. military and defense contractors for decision-making.
- The researchers invented fake countries with different military levels, concerns, and histories and asked the AIs to act as their leaders.
- The AIs showed signs of sudden and hard-to-predict escalations, arms-race dynamics, and worrying justifications for violent actions.
- The study casts doubt on the rush to deploy LLMs in the military and diplomatic domains, and calls for more research on their risks and limitations.
Is this a case of “here, LLM trained on millions of lines of text from cold war novels, fictional alien invasions, nuclear apocalypses and the like, please assume there is a tense diplomatic situation and write the next actions taken by either party” ?
But it’s good that the researchers made explicit what should be clear: these LLMs aren’t thinking/reasoning “AI” that is being consulted, they just serve up a remix of likely sentences that might reasonably follow the gist of the provided prior text (“context”). A corrupted hive mind of fiction authors and actions that served their ends of telling a story.
That being said, I could imagine /some/ use if an LLM was trained/retrained on exclusively verified information describing real actions and outcomes in 20th century military history. It could serve as brainstorming aid, to point out possible actions or possible responses of the opponent which decision makers might not have thought of.
LLM is literally a machine made to give you more of the same
Throwing that kind of stuff at an LLM just doesn’t make sense.
People need to understand that LLMs are not smart, they’re just really fancy autocompletion. I hate that we call those “AI”, there’s no intelligence whatsoever in those still. It’s machine learning. All it knows is what humans said in its training dataset which is a lot of news, wikipedia and social media. And most of what’s available is world war and cold war data.
It’s not producing millitary strategies, it’s predicting what our world leaders are likely to say and do and what your newspapers would be saying in the provided scenario, most likely heavily based on world war and cold war rethoric. And that, it’s quite unfortunately pretty good at it since we seem hell bent on repeating history lately. But the model, it’s got zero clues what a military strategy is. All it knows is that a lot of people think nuking the enemy is an easy way towards peace.
Stop using LLMs wrong. They’re amazing but they’re not fucking magic
“Dad, what happened to humans on this planet?”
“Well son, they used a statistical computer program predicting words and allowed that program to control their weapons of mass destruction”
“That sounds pretty stupid. Why would they do such a thing?”
“They thought they found AI, son.”
“So every other species on the planet managed to not destroy it, except humans, who were supposed to be the most intelligent?”
“Yes that’s the irony of humanity, son.”
Why the actual fuck is anyone considering putting LLMs into the driving seat of anything?!
Of course they make fucked up decisions with no proper or justifiable rationale, because they have no brains. They’re language models, stochastic parrots stringing together sentences to fit the prompt(s) given to them.
As someone with military experience, military members, especially flag officers, are not the brightest bulbs in the world and are easily awed by extremely simple tech demonstrations.
It’s already too late, make sure you’ve got your favorite food and beverages ready because several countries already have autonomous weapons being live tested in the middle east, and from my understanding of the situation, the new jets already have some hilariously incompetent AI in them (in simulation, the air force contractor that was in control kept giving ethical barriers to objective completion and the ai went to kill the controller to more easily complete the objective…)
e. public sources: https://www.npr.org/2021/06/01/1002196245/a-u-n-report-suggests-libya-saw-the-first-battlefield-killing-by-an-autonomous-d
https://taskandpurpose.com/news/air-force-artificial-intelligence-drone/
(The above are public articles maintained to minimize concern surrounding the tech which is why the air force almost immediately walked back their accidental admissions with the following statement:
“The Department of the Air Force has not conducted any such AI-drone simulations and remains committed to ethical and responsible use of AI technology. This was a hypothetical thought experiment, not a simulation,” said Air Force spokesperson Ann Stefanek. "
From my understanding, of course take this with a grain of salt since I’m an anon on a message board, we did do this.)
As much as I’m worried about military autonomous drones, I’m even more worried about guerilla autonomous drones. With off the shelf AI becoming more and more accessible it’s not too hard to imagine a moderately smart person being able to make autonomous killing drones using off the shelf materials. It doesn’t even need to be autonomous. In the Ukraine war hobbyists have been able to help the war effort by Jerry rigging together bombs onto commercial drones. I’m grateful but shocked that there haven’t been any major drone based terrorist attacks, and I’m not sure how they can be defended against.
If you have experience and effective training browsing the dark web, there are several examples of your concerns already coming alive.
Bot farms have levelled up to ai farms, with various models being made as ‘specialist’ ai’s for things like credit card theft, network intrusion, malware, etc, and from when I last looked into it a few months ago they had already moved on to attempt to get all the specialists to start training general purpose models.
Things are not looking particularly great and I would posit if AGI does happen in our lifetime it’s not going to be because anyone alive actually intended that to happen, but the criminals running a wide variety of specialists train a general purpose ai with intentions to use it for easy money.
I usually chirp back with ‘nothing we have now is really AI’ but I can’t seriously take that view with some of the things being tested by some criminal organizations these days and there’s not really a way to stop this from happening.
Porn. Porn companies will invent AGI.
How can we expect a predictive language model trained on our violent history to come up with non-violent solutions in any consistent fashion?
Why would you use a chat-bot for decision-making? Fucking morons.
They didn’t. They used LLMs.
Edit: to everyone saying that LLMs “are chat bots”. I know it seems that way to the layperson and how it’s often explain, but it’s not true.
Which are chat bots.
