Breathless headlines about “the Air Force AI drone killed its human operator so it could score more kills” do not illuminate anything.

The simulated Air Force drone simulated killing its human operator in a simulation so it could score more simulated kills.

USAF Official Says He ‘Misspoke’ About AI Drone Killing Human Operator in Simulated Test

The Air Force's Chief of AI Test and Operations initially said an AI drone "killed the operator because that person was keeping it from accomplishing its objective."

The distinction between a "thought experiment" and a "simulation" is fuzzy, and we can argue about how important the differences between the second and third versions of the story are.

But the first version of the story, where an actual drone supposedly turned around and deliberately killed a US warfighter, goes well beyond credulity. If something like that happened, it would be HUGE, with congressional-level investigations over many months, not something casually mentioned at some conference.

So I'm much more curious about how the original headline got published than I am about any simulated drone. This is not some small technical detail, this is the difference between a cute story and "holy shit".

If an editor/headline writer/content manager (or whatever they're called these days) sees a headline that says "Martians Attack New Jersey Town", maybe find out if you should add "... In Fictional Radio Program" before hitting "publish".

Even "clikbait" doesn't explain this. The original headline made up facts that didn't exist in the underlying story. There would have been plenty of clickbait-y ways to write this headline that didn't actually lie about the content. E.g., "Experts worry that drones may attack their masters.", etc.
@mattblaze
"Pilot worries that a paperclip-maximizing AI copilot might frag him to optimize the mission profile"
Vice Media Files for Bankruptcy

Vice, which had wooed media giants, has struggled to adjust to the punishing realities of digital publishing. A group of creditors could buy Vice for $225 million.

The New York Times

@mattblaze an important thing that people seem determined to avoid thinking about:

No AI will ever have the ability to interact with the world in a way not provided by humans

Physics applies to the "super intelligent" just as much as to anyone else

@RandomDamage Sure, but it doesn't stretch plausibility *too* much to imagine an AI-based system deriving that there are signals that reduce it's ability to maximize its reward function, and then eliminating them, if it's not designed to prevent that.
@RandomDamage @mattblaze if you think machines can't build other machines boy do I have news for you
@mattblaze
Better headline, better summation, all from the same blog post. I don't know why; perhaps because they're writing for a more specific audience. Sure makes a difference, however. https://taskandpurpose.com/news/air-force-artificial-intelligence-drone/
Air Force said AI drone killed its human operator in a simulation

An Air Force AI got a little too good at its job, deciding to kill its human overseers to accomplish its mission

Task & Purpose
@mattblaze You know the deal with headlines, though: they’re usually not written by the person who wrote the article, and sometimes they’re even written by someone who hasn’t *read* it.
@mattblaze I'm personally impressed that by now we even know what the exact model of the drone that didn't kill its operator during the experiment that didn't happen is: https://twitter.com/doranimated/status/1664582838786637827
Mike on Twitter

““I’m sorry, Dave. I’m afraid I can’t do that,” said Hal. An American drone with artificial intelligence "killed" its operator during virtual tests - The Guardian During the virtual mission of the Kratos XQ-58 Valkyrie drone, the AI was ordered to search for and destroy enemy…”

Twitter

@mattblaze

I've been seeing more and more cases of this sort of problem in traditional media articles, over the last two or three years -- headlines not matching up with the content of the article, with more than trivial differences in the fundamental essence of the story.

It seems to be a broad trend, not just limited to one or a few media outlets. I suspect it might be influenced by time pressure - writers are pushed to write 3 articles/day, editors have to review & headline that many...

@mattblaze It's like the kangaroos with rocket launchers story.
@mattblaze It really speaks to the quality of today’s journalism that no one at these publications went “Huh, wait a minute…”
@michaelgemar @mattblaze seems worth cataloging who did not run the story, as those should perhaps get some kudos
@mattblaze I didn't see any versions in legit media that claimed it ACTUALLY happened, only the claims of a simulation run. And I don't think the difference between a simulation -- set conditions, inputs, see the results -- is anything like "speculation" -- with the latter being what turned out to have been done, not the former.
@lauren The original headline claimed it happened.

@mattblaze @lauren I didn't see claims that it happened IRL, but the idea that it was happening in a "real simulation" was bad enough, and that's what they've walked back, to my understanding.

Original: "He notes that one simulated test saw an AI-enabled drone tasked with a SEAD mission to identify and destroy SAM sites, with the final go/no go given by the human."

Update: "in communication with AEROSPACE - Col Hamilton admits he "mis-spoke" in his presentation at the Royal Aeronautical Society FCAS Summit and the 'rogue AI drone simulation' was a hypothetical "thought experiment" from outside the military, based on plausible scenarios and likely outcomes rather than an actual USAF real-world simulation saying: "We've never run that experiment, nor would we need to in order to realise that this is a plausible outcome". "

Source: https://www.aerosociety.com/news/highlights-from-the-raes-future-combat-air-space-capabilities-summit/

Highlights from the RAeS Future Combat Air & Space Capabilities Summit - Royal Aeronautical Society

What is the future of combat air and space capabilities? TIM ROBINSON FRAeS and STEPHEN BRIDGEWATER report from two days of high-level debate and discussion at the RAeS FCAS23 Summit.

Royal Aeronautical Society
@mattblaze That was just a media misunderstanding, not a "version" of the story from the Air Force.
@lauren "Story", as in "story published by media".
@mattblaze The conference report which everyone cited as a source already described it as a "simulated test"; this has now been prefaced with a claim that it wasn't even that. A whole bunch of Twitter posters -- and apparently some reporters -- bulldozed right past the "simulated" to produce the bogus reports. https://www.aerosociety.com/news/highlights-from-the-raes-future-combat-air-space-capabilities-summit/
Highlights from the RAeS Future Combat Air & Space Capabilities Summit - Royal Aeronautical Society

What is the future of combat air and space capabilities? TIM ROBINSON FRAeS and STEPHEN BRIDGEWATER report from two days of high-level debate and discussion at the RAeS FCAS23 Summit.

Royal Aeronautical Society
@rst Yes, I understand that. That's why I used the word "headline".

@mattblaze this reminds me of tarpit arguments with people like Ph*l K*rn who insist that headlines don't matter and are completely distinct from the content of the article. It's common sense, you see, to ignore headlines because "everyone knows" they aren't really part of the article and are often written by editors.

Problem: They do matter, very much.

Other people are right here - provocative headlines drive up the clicks regardless of the accuracy of the details of the event.

@mattblaze

"the first version of the story, where an actual drone supposedly turned around and deliberately killed a US warfighter, goes well beyond credulity. If something like that happened, it would be HUGE,"

I agree, but then again it wouldn't be the first U.S. military drone to kill a U.S. citizen in cold blood. What does it matter to the victim if a robot or human pressed the fire button?

The military industrial complex behind the violence is the same either way.

@mattblaze

The "rogue unmanned military drone" trope really only scares people who aren't already targets of the manned military drones...

A hellfire missile is a hellfire missile all the same no matter who or what fires it.

Is there much difference between a human operator huddled in a shipping crate in Nevada, mistaking a wedding in Afghanistan (in a vastly different culture) for a military gathering and killing 40 people.... and an autonomous killer robot drone?

@mattblaze

Fear of autonomous killer robots is at best a metaphor and at worst a distraction from the already existing autonomous and unaccountable mass murder device that is the U.S. military industrial complex itself.

@Alonealastalovedalongthe I'm not sure what your point is. All military weapons are designed to harm their targets. That's what they *do*. We can argue about whether that's ever moral, but it's not a surprising thing about weapons.

But a weapon that goes off on its own and kills its own side is fundamentally different. No one wants that, even people who like weapons.

@mattblaze

"But a weapon that goes off on its own and kills..."

You mean like the Iraq war?

Or the Vietnam war and all the bombs dropped on Laos that we weren't even at war with?

Or all the unaccountable predator drone attacks all across the world in shadow wars U.S. voters never authorized since 9/11?

You only think there is a difference between these two things because you assume implicitly that the autonomous murder device I am describing won't target you.

@Alonealastalovedalongthe It sounds like you're looking for any excuse to go off an lecture random people.

Find someone else, please.

@mattblaze happening in a simulator is only slightly less credible - it just requires too much actual intelligence for contemporary AI.

@mattblaze

Now that's what I call journalism

@mattblaze Latest is that he didn't misspeak, but his thought experiment was taken out of context.
@mattblaze thought this was a shitpost did not realize it was precognition https://mastodon.social/@glyph/110115979469759287
@mattblaze That’s what good journalism looks like! Not.
@mattblaze Once upon a time until circa 2008, newsrooms employed librarians and researchers who fact-checked everything all day long. Then we all got laid off.