i generally like Dr. Fatima, so i was curious about this video, but it was pretty disappointing. I have several more thoughts but i wanted to keep this reasonably short.

The first section about "back end harms" is the best part. Unfortunately, section 2, the "Front End Harms" section names valid problems but falls flat when it comes to solutions. She rolls out a lot of lib tropes about "education" and gestures at companies self-regulating the sycophancy of their own models despite evidence (a massive amount of it if we consider corporate "self-regulation" more broadly) to the contrary. Remember "media literacy" discourse about social media misinformation? it went nowhere, because it's not a problem that can be solved with education, it's imperative to actually learn lessons from history and bring this technology under political control. You cannot do this effectively when your government is 3 monopolistic corporations in a trenchcoat.

She says that anthropic are "better than the competition" which is trivially true but extremely credulous. If faced with the choice, I would prefer that Claude beats out Grok and chatGPT but this is ultimately a marginal difference due to the nature of the industry and ultimately of capitalism itself.

Section 3 is fine for what it is, but it's really about the psychology of persuasion and not AI as such. Some of the discourse on this site would benefit from the reminder that moral absolutism isn't very persuasive, but this section is way too long, we can just dispense with the moralizing and "harm reduction" anyways, because just like plastic recycling, personal reduction in AI use for harm reduction reasons is a fake solution to a systemic problem.

The harms of AI are intimately linked with the nature of monopoly & platform capital. You cannot defeat this enemy if you cannot actually describe it properly, and no amount of leveraging NIMBYism to defeat your local datacenter project will fix it.

Maybe Dr. Fatima isn't the right person to be delivering that message, or simply doesn't view the problem through a materialist lens. It's not enough to be against AI, you need to be for something. This is the disease of liberal technocratic managerialism and it manifests in myriad ways including AI critique. We need to move beyond it.