Mike the Mad Biologist

While the AI safety cultists are worried about the singularity, what we should be concerned about is how AI can be used to make existing problems worse (boldface mine):

An explosive and newly partially unsealed court filing alleges that the US Department of Homeland Security built up dossiers on ICE observers—including their photos, license plate numbers, and other information—and stored them in a system called the Investigative Case Management database, which was built by Palantir.

Palantir did not immediately respond to requests for comment…

Plaintiffs allege in the filing, which is heavily based on documents produced by the government and depositions of DHS agents, that agents had a practice of intimidating protesters and observers. They did this, the filing says, by doing drive-bys of the observers’ homes and, in the case of at least one observer, requiring that she be referred to secondary inspection any time she tries to cross the border until at least January 2027…

According to the legal filing, that afternoon, J.C. called Williams on her cell phone, identifying himself only as “the Department of Homeland Security.” According to Williams, he told her she “should encourage” her spouse, Pantos, “not to do anything like that in the future because people who are doing things like that might get added to a domestic terrorism watchlist.” J.C. testified in a deposition that he warned Williams she would be arrested if she “continue[d] to do this.” Government attorneys asserted in a court filing that he denied referencing a “domestic terrorism watchlist.”

But you might think, well, this is just a database and it doesn’t involve AI. Nope:

Palantir’s role extends beyond field enforcement. ICE is using generative AI tools supplied by the company to assist with software development, database queries, and system diagnostics inside its Investigative Case Management System, a customized version of Palantir’s Gotham platform. DHS filings again describe the models as off-the-shelf systems rather than agency-trained AI.

Per FedScoop, “ICE added 25 AI use cases since its disclosure last summer, including to process tips, review mobile device data relevant to investigations, confirm identities of individuals via biometric data and detect intentional misidentification. Of the newly added uses at ICE, three are products from Palantir, which has been a notable — and at times controversial — technology partner for the U.S. government under the Trump administration.”

This is first and foremost a data privacy and civil liberties problem, but AI makes it worse, in that DHS minions can use and abuse this information far more easily. There are a whole slew of existing problems, from hacking* to deepfakes that various AI-based tools make worse, but they are longstanding problems–data privacy was a problem before 2022 (give or take), but it’s just easier to acquire and misuse data now.

Essentially, those that purport to govern us haven’t moved on a whole bunch of issues, and now those problems have become magnified by AI.

*From what I’ve read, most of the hacks are cribbed from hacks designed by people, and aren’t particularly novel in and of themselves, and are just applied by the various AI systems.