Welcome to the First AI War
April 4: The first AI war … World Central Kitchen fires back … Slowing deforestation … Electoral College victory …
An Israeli artificial-intelligence system is responsible for targeting as many as 37,000 people as members of Hamas and is directly responsible for the deaths of thousands of Palestinians, most of them women and children, according to a stunning new report by Israelis and Palestinians at +972 magazine and Local Call.
According to the report, humanity is now fighting the first large-scale, wholesale AI war.
Israel has denied some of the reporting, but confirmed some of it, too, and has left significant wiggle room on most of it. Here are some of the highlights from your first Fucking News dispatch on the first AI war:
The AI system, called Lavender, identified about 37,000 people as Hamas militants; specifically, people to be killed. (Which is weird because, before the war, Israeli humans estimated that Hamas had no more than 30,000 people.)
The Israeli Defense Forces (IDF) usually strikes Lavender’s targets at night — when they’re home with their families — because that’s when and where it’s easiest to find them. (Which is weird, because Israel has said it takes measures to avoid killing non-combatants, including infant non-combatants.)
The AI system that alerts the IDF that Lavender’s targets have arrived home to their families is called “Where’s Daddy?”
Another AI system that identifies structural targets is called “The Gospel.”
When choosing munitions to bomb Lavender’s targets — once Where’s Daddy? flagged them as arriving home — the IDF preferred to use dumb bombs, which can cause far more casualties by taking down entire buildings, because they’re cheaper than precision missiles.
The IDF set caps of acceptable collateral damage for specific types of targets. To attack a junior Hamas target, it’s okay to kill as many as 20 civilians. The number of acceptable civilian dead goes up with the rank of the target. The IDF several times approved more than 100 civilian deaths for strikes aimed at a single Hamas commander.
Lavender’s targets were verified by human personnel who didn’t assess why Lavender chose them, or the data it used, and typically did nothing more than spend about 20 seconds checking that the target was male.
The lax checks have been kept in place despite a Lavender error rate of about 10%, according to an estimate by the IDF, which has an error rate of self-serving confirmation bias.
Lavender sometimes puts targets on the list who have no connection to Hamas other than joining Hamas on Lavender’s target list. On some occasions, homes were bombed after Where’s Daddy? wrongly said the target was there, and no one verified it.
Before Oct. 7, the IDF used a meticulous, human process to identify its Hamas targets, limiting the list to military operatives senior enough to justify targeting them at their home. Only in these cases were civilian deaths deemed acceptable under the proportionality principles of international law regarding exactly who countries can kill in the name of war.
Only a few dozen people were on this list.
But after Oct. 7, the IDF expanded the list. Exponentially. It now included people regardless of rank or military significance.
In fact, the target list included so many people that Israel no longer had the bandwidth to authorize individual attacks anymore, let alone confirm where the targets were. So they outsourced it all to the AI; to Lavender and Where’s Daddy?
In the first two weeks after Oct. 7, humans checked the AI’s results. Once the sample set hit 90% accuracy, they stopped checking and started treating the AI findings as orders for the IDF to carry out.
About 15,000 people were killed in the first six weeks of the war. About two-thirds of all the dead are said to be women and children.
One unintended consequence of the IDF’s use of AI is that Israel’s critics have just been handed a massive stockpile of ammunition to destroy Israel’s claim that it protects civilians. And it’s this claim that provides the legal underpinning for Pres. Joe Biden to keep supplying weapons to Israel’s AI.
World Central Kitchen Strikes Back (with Words)
We don’t know whether Monday’s airstrike that killed seven World Central Kitchen workers was the result of AI. We don’t know what the U.S. and other Israeli allies knew about Israel’s machinery of death, or when.
But we probably will eventually.
And as I mentioned yesterday in your humble TFN, there’s a difference between U.S. demands that the investigation of Monday’s airstrike be transparent, and what World Central Kitchen is now demanding: Namely, that it be independent.
In a statement this morning, World Central Kitchen said, “We have asked the governments of Australia, Canada, the United States of America, Poland, and the United Kingdom to join us in demanding an independent, third-party investigation into these attacks, including whether they were carried out intentionally or otherwise violated international law.”
It’s worth noting that we haven’t really wrapped our heads around what “intentionally” means when we’re talking about AI, but programming AI to violate international law by abandoning proportionality certainly violates international law.
New York City Sides with AI in War on Workers
New York City Mayor Eric Adams has decided AI is so helpful, that he’s gonna keep using it even though it’s advising companies to illegally punish workers.
The Markup first reported that the city’s AI chatbot was spewing evil, illegal, bullshit in response to questions from small-business employers who just want to know exactly how much evil and bullshit is legal.
Weirdly enough, the Microsoft AI’s answers always seem to be wrong in a way that makes life easier for people with money. Among the wrong-but-boxx-friendly answers the chatbot gave about what businesses can do, the AP reports:
Businesses who want to put out their trash in black bags rather than the required bins can go right ahead, because they’re the real heroes.
Bosses can go right ahead and fire pesky workers who insist on jamming up the gears of our heroic capitalist system by:
Complaining about sexual harassment
Keeping their dreadlocks
Keeping a fetus in their uterus.
Adams on Tuesday acknowledged the errors, and assured residents he’s cool with them.
“Only those who are fearful sit down and say, ‘Oh, it is not working the way we want, now we have to run away from it all together.’ I don’t live that way.” (In fact, Adams does live that way when we’re talking about fear of, say, crime.)
TFN Human Identifies Correct Target
AI is not the enemy. Not in New York. Not in Gaza.
AI is a tool. Its use is not an AI problem, it’s a problem of the humans using it and how they use it.
If anything, AI holds the potential to unleash a glorious new phase of human existence. Here’s what I, a human, mean by that.
Right now, we live in fear of what AI can do. Despite Mayor Eric Adams’s claim, though, what we really fear is not AI but what humans will do to other humans with AI. Because AI can dehumanize us when humanity isn’t baked into its algorithms.
Which reduces us to commodities. Or collateral damage.
Either way, at some point, AI will be capable of carrying out just about any task humans can now do. And the fear is that human life will then be rendered worthless.
But what if the real problem all along has been what we thought gave us worth in the first place?
As I’ve argued, one reason people are so down on Pres. Joe Biden despite a roaring economy is that human life is increasingly being defined in terms of its economic value. The human economy is leaving humanity behind.
Our cities and towns are planned and run to maximize productivity and the extraction of money from people by corporations. Our educational system is increasingly geared toward cranking out economic units rather than lovers of art and appreciators of skill and creativity. Even our unions are fighting for jobs rather than a four-day workweek or other equitable distributions of the fruits of automation’s productivity.
But maybe if humans no longer have economic value, we’ll be forced to choose a different metric for valuing human life, and human existence.
Maybe if AI does everything required to fuel humanity’s economic engines, we can choose a different purpose — or purposes — for human beings to be.
But first we have to reject the idea that human value derives from its economic value.
Something Resembling Good News
First, the bad news, because we’re The Fucking News and that’s how we roll.
Last year, the Earth lost about 14,000 square miles of old-growth forests. That’s forests that have been unfucked-with by humans, also known as primary forests, as opposed to secondary or “fucked-with” forests.
That’s an area somewhere between the size of Maryland and Switzerland. The climate impact of that lost forest is the equivalent of about half the annual U.S. consumption of fossil fuels.
(Meaning the billionaires currently holding our money who dream about changing the world should just buy primary forests and not fuck with them.)
The good news, according to the folks at Global Forest Watch, is that the rate of primary-forest deforestation actually fell by 9% in 2022. Most of that decrease came in Brazil and Colombia, Global Forest Watch said. And why is that? POLITICS! That thing people say they hate!
Brazil is still one of the top deforesting countries, along with Indonesia, the Democratic Republic of Congo, and Bolivia, where deforesting last year rose a defucking 27%.
But new Brazilian President Luiz Inacio Lula da Silva came into office promising environmentalism and managed to scale back deforestation by 36% in just a year. With more to come, hopefully.
In Colombia, President Gustavo Petro baked climate concerns into his peace talks with armed militants who live in his country’s jungles. Deforestation shrank by 49% just last year. Thanks, politics!
Campaign Watch
PRESIDENT One of the 538 Electoral College votes up for grabs in the 2024 presidential election could be decided well before Election Day. And Democrats fighting for that vote won an important but possibly temporary victory last night.
Bear with me.
Nebraska has three congressional districts. State law says each of those districts can choose one of the state’s electors — unlike most states, where whoever wins the state vote wins all the state’s electors. Because of this, Democrats have been counting on getting one single, precious Electoral College vote out of Omaha’s district.
Naturally, the party of states’ rights has been pushing Nebraska to do things differently. D4FRFP1 Donald Trump has been pressuring the Republican-controlled legislature to go winner-takes-all. And Gov. Jim Pillen (R-NE) is on board.
Last night, just a day after Trump’s push, the winner-takes-all legislators mounted another push — and failed. The vote was on a technicality — whether the winner-takes-all measure was relevant enough to attach to another bill up for a vote. But not enough Republicans agreed to redefine “relevant” in a way that would make Trump happy.
And late last night, Daniel Nichanian of Bolts magazine, an elections essential, Tweeted that the bill’s sponsor is admitting defeat. The sponsor could turn out to be wrong, of course, but as of last night they said that “Winner Takes All isn’t moving in 2024.”
That’s one down, 269 to go.
Recommended reading
CHINA AND THE U.S. The Wall Street Journal has a really interesting look at the evolving relationship between the U.S. and China. Treasury Secretary Janet Yellen — and her own evolution — is the focus. An important but subtextual theme, for me, anyway, is that even the biggest of wigs in economic policy do not know for sure what to do.
TCB
Okay, newsfuckers, TFN is live on Instagram. That’s right, we’re finally joining the 21st century. Mid-2016, to be exact!
Substack is pretty subpar about generating shareable images, but I did what I could with my recent story on The Family and the Kennedy Space Center. It’s about an event college kids are going to THIS WEEKEND, so if you’re on Instagram, please consider robustly sharing my sad attempt at ‘gramming this story.
Sharing my work on social media — I’m also on Twitter and Facebook and Bluesky and Mastodon, and now Spoutible, too! — is one of supporting it.
You can also wrap TFN in plastic and toss it in your neighbor’s driveway every morning. Or you can support my work directly by becoming a monthly or annual subscriber. Thank you.
Now go get ‘em, kids…!
D4FRFP = Disgraced, quadicted, fraudster, rapist, former President.
AI—how it's programmed to learn, how it's being used—humans are still the problem. This is similar to blaming something on "the computer"—people program it, use it, and people are the problem. I didn't think I'd see some of the SciFi I read as a kid become a reality so quickly.
Nebraska—as much as they try to shove more red areas into our purple haven, we sometimes go blue for prez. And the Unicam managed to keep the potential for a blue dot in the red sea alive by doing nothing (again).