Lesley Carhart Mar 8
Please, please stop supporting gen AI companies. It’s not worth it. https://futurism.com/artificial-intelligence/pentagon-ai-claude-bombing-elementary-school
Pentagon Refuses to Say If AI Was Used to Select Elementary School as Bombing Target

Did the US military use Anthropic's Claude to select targets in its weekend operations in Iran, with devastating results?

Futurism
Show threadLesley Carhart 
Phew, that spun people up on BlueSky
Mar 8 at 11:32amWeb
Show threadWendy NatherMar 8
@hacks4pancakes You’re my hero.
Show threadIan Campbell 🏴Mar 8
@wendynather @hacks4pancakes seconded
Show threadKim ScheinbergMar 8
@neurovagrant @wendynather @hacks4pancakes
Me three
Show threadJohn MaxwellMar 8
@hacks4pancakes Thank you.
Show threadMomoMar 8

@hacks4pancakes
I wonder if that decision to use AI in targetting was based on "Yes, use AI, it's perfect and can't do mistakes!!" or rather "We know that AI makes wrong decisions and combining this with live ammunition costs innocent lives. But it's not our civilians who get blown up, so we don't care. Acceptable is good enough!"-Irresponsibity.

Or in simple terms: Are they ignorantly stupid or just cruel?

Show threadGeorge BMar 8

@momo @hacks4pancakes

That one's really hard because they have shown themselves to be both so many times

Show threadJohn MaxwellMar 8
@momo @hacks4pancakes Both.
Show threadMacCruiskeenMar 8
@momo @hacks4pancakes The leadership is swayed by the AI $$$. And they're too stupid to conider anything else. At work we've been testing a Claude-based system to remediate mssing alt-text in ebooks, and the results would not make me confident you could use it for targeting. In fact, the opposite of that.
Show threadThe FrogMar 8

@momo @hacks4pancakes

Third option:
They're cruel and murderous on purpose (Israel is committing genocide in Gaza and the US supports it after all) and initially planned to use AI as plausible denial for PR reasons, but don't because it doesn't make it look any better. Worse, in fact.

Show threadWhichOne'sPink 🇫🇮Mar 9
@momo @hacks4pancakes The latter.
They'll be using this to test all kinds of diabolical fuckery.
Show threadJuffo-WupMar 8

@hacks4pancakes

Congrats for the second post in that screen shot, I think the point you made there is not stated often enough.

I really hate how LLM took over the general term "AI", that has been used in CS research for dozens of different approaches.

Show threadLesley Carhart Mar 8
@mndflayr I drink and I know something
Show threadellteeMar 8

@hacks4pancakes
Its the opposite for me. I drink and I stop knowing things. 😋

@mndflayr

Show thread☣️ akaasjager ☣️Mar 8
@hacks4pancakes I don’t, and that probably has brought me in the permanent state that Terry Pratchett once called “knurd”. -5/10 can’t recommend
Show threadXCondEMar 8
@hacks4pancakes non-sequitur: that glass is sweating buckets. You must still be in Singapore.
Show threadSuperstars of the ApocalypseMar 8

@hacks4pancakes

To be fair, it's really easy to spin people up on Bluesky.

I don't know of any truly left-leaning people who support AI.

Show threadSuperstars of the ApocalypseMar 8

@hacks4pancakes

We just need to give better training to the LLMs and make them wear body cams

Show threadEric LawtonMar 8

@CosmickTrigger

Thank you.

The whole "AI safety" field in a nutshell.

https://en.wikipedia.org/wiki/AI_safety

@hacks4pancakes @AnnaAnthro

AI safety - Wikipedia

Show threadbigiainMar 9
@CosmickTrigger Of course domestic LLMs, OpenAI/Anthropic/Gemeni/etc, have qualified immunity. We _strongly_ recommend against using Deepseek...
@hacks4pancakes
Show threadRaven667Mar 8
@hacks4pancakes and the smug "military targeters would probably have made the same mistake" makes me think of Chief Wright (Stonekettle Station) who did targeting during the first Gulf War and talked about sleepless nights double-checking intelligence to make sure they didn't kill civilians, and called off attacks when they couldn't be sure, who would have been personally devastated if they learned a missile they launched had killed school-kids. The military structure at the time allowed him to care about such things, but Secretary Kegseth doesn't seem to care or have a soul as far as I can tell.
Show threadDźwiedziuMar 8
@hacks4pancakes
I wish that would be an energy source…
Show threadInterestingMooMar 8
@hacks4pancakes on the other side of things, does it provide them less liability if Claude makes the targets? Or more? Possibly shifts liabilities away from admin to AI devs. I mean it shouldn't but I can see them arguing that. Administration's love scapegoats.
Show threadgeolawMar 8
@hacks4pancakes https://imgs.xkcd.com/comics/the_important_field.png