Blog

Meta has had two significant issues brought up this week (hat tip to Futurism):

1. Facebook Is Hosting Huge Numbers of Horrifying AI-Generated Videos of Violent Child Abuse, and Meta Is Barely Even Pretending to Care About Taking Them Down
2. Mother Horrified When Meta’s AI Scours Instagram and Facebook for Invasive Information About Her Young Daughters

Some quick observations:

re: #1: Meta’s PR team sent this note to the Futurism team: “Not all the links you sent violate our policies,” it read. “Can you please avoid suggesting in your reporting that they do?” My main thought isn’t “why doesn’t this violate your policies?” This is horrifying content. But it is not CSAM. Which is what has the attention of government, the public, etc. I’m not too surprised that Meta was caught off guard about both the existence of this content and its prevalence. I also expect that their “Community Standards” are much more reactive. Therefore, since they didn’t anticipate content featuring physical violence against children, they haven’t created a community standard to be violated. I expect that will be fixed “soon”, when the committee assigned to evaluate this issue finally finishes their review. This game of whack-a-mole with humanity’s darkest impulses is troubling, and I wonder if there’s a way to be proactive and not dive into Minority Report territory.

re: #2: This story is troubling in a different way. AI’s ability to analyze a monumental array of data and ACCURATELY detail someone’s personal details is amazing from a technical perspective, and terrifying from a privacy one. As I was reading this article, I came up with the phrase “stalker-tech” to describe it. Now, before we come swinging at Ms. Robins for sharing ANYTHING about her children, she’s not some influencer sharing her intimate details. Most of the information harvested was shared more privately, amongst friends and family. Meta’s AI triangulated amongst a huge spread of information in order to derive deep details about these children. On an aside: I think her response of asking folks to take down these images is, well, hopeless. The information has been harvested. Removing elements of it from one place on the internet doesn’t delete the graph. Meta’s response? “But we missed the mark because the feature never should have prompted the individual with questions like that and we’ve fixed that issue.” I’m curious about how the “issue” has been fixed. Will the prompts not ask similar questions? Is the ability to analyze huge swaths of data been minimized? My guess is that the prompting has been modified, but that the data is still mapped. Therefore, I expect this sort of misuse is still possible, it will just take more steps and bad intent to execute on it.

Humans are a reactive lot. We’re terrible at thinking technology’s negative possibilities through before we run into problems/crisises. We tend towards letting the potential positives blind us to any negatives. I wonder if we have any hope of changing that part of our nature.