it’s called “life.” You walk to the store. Maybe you see someone you know on the way. You find a dollar on the ground. The sun shines on your face. You are participating in the world.
So far my main takeaway from the Napoleon biography is a growing conviction that our current social model of keeping kids from any moral or legal responsibility for anything until 18 at a minimum is a huge historical aberration that's actually driving society insane
@bernhard_me@_philschmid@googlehealth This API doesn’t really expose second to second data though right? Moreso just the summary stuff mostly already available on the app?
@JosephEarlyo@AJRuecker@JohnTitor137 Yeah that’s the whole point. A constant awareness that you are creating technology that could potentially be used for lots of harm
@otherside_X42@djcows look it up its pretty cool. one of the good "scientific" applications of AI. granted its not exactly a chatbot application, but I think people need to learn "AI" does not just mean "chatbot"
@clockstiqqun just to inform, this is a summary of the points of Alex Karp's book, not something random they decided to tweet out. Alex Karp is good friends with Thiel
Palantir recently became the military's primary software system for "targeting," and increasingly other aspects of the fight.
Since the scary "targeting" OS used scary "AI," we were never going to have a sane conversation about how this works.
This story is my best shot at a sane conversation about how this works.
First of all, "targeting" isn't shooting.
"Targeting" is the exhaustive, bureaucratic process by which we plan, execute, and *assess* shooting.
How our planes will get the right munitions. Where they’ll refuel. Cloud coverage. The literal illumination of the moon. Matters of physics. Matters of contingency: the AC-130 gunship needs to be in an overwatch position before the ground assault can happen, but the gunship can’t move into place until the air defenses have been taken out. A “scheme of maneuvers” is determined. Only then does the strike proceed.
^That is only a small snippet of the "targeting" process, btw.
As of the start of the conflict with Iran this year, "targeting" was still heavily manual. The glue between disparate processes, even if they featured automation, was: PowerPoint, email, chat, and Excel files. Target lists were relayed in spreadsheets. Sequenced maneuvers sat in Gantt charts in PowerPoint.
That bottleneck deepened a relationship with Palantir that started decades ago. Now, hundreds of AI agents in the "Maven" software system do stuff like: make sure there's nothing in the military's historical intelligence that might disqualify a target before a strike.
Sources believe AI could've prevented the U.S. strike on a school in Iran, which killed more than 100 children, as it was based on "outdated data."
Palantir has been called a lot of things. Evil, dystopian.
I've been a tech reporter for almost 10 years, and among my travels, nothing is more dystopian than dysfunctional, manual tech. It literally makes people want to die. It literally causes death. The Pentagon has wanted war AI since Vietnam, for good reason.
This is the definitive story of how Palantir built it.
https://t.co/dO3cQi7Gdm