Peter Kovalsky

I've been pitching Palmer's excellent Inventing the Renaissance to friends as, essentially, "Baby's First Historiography."

This essay does an unsurprisingly spectacular job of more clearly articulating that pitch.

But also, like... even if you entirely set aside trans issues, isn't that just kind of the nature of *activism*?

Like, a big part of an activist's job description is to know a lot about their thing and to read to filth people in power who make decisions about the thing while knowing less about it.

Wrapping up an onboarding conversation with a new client about work style and communication cadence: "I'm not too worried about it. You're Dutch, I'm autistic... we've got this!"

Okay, look, I'm not saying that AI is *exactly* the same as porn, but:
- real ethical concerns about production
- rampant misogyny
- can *sometimes* be used appropriately as long as you remember it doesn't reflect reality
- creeping into every product/service
- techbros use it all day every day

Ask me sometime about the history of predictive text in general. It was wild long before AI got involved!

So much of this boils down to "stop asking the trends machine for facts."

Even if you buy what the AI hawkers are hawking, all the AI can ever really give you are trends and averages. Trends and averages can be useful! But trends aren't *facts.*

Hello former gifted kids. You didn't "peak" in high school, you discovered what you could accomplish when your basic needs were met and you were allowed to pursue your interests.

We could build that into society, if we wanted to.

A repugnant recent experience was sitting on hold with the US Patent and Trademark Office and being told that their agents were "busy assisting other customers." The notion of being a *customer* of one's government is just so gross.

Maybe in the Infomocracy/Terra Ignota worlds, but even then.

Just finished We Will Rise Again, which I enjoyed a lot even though there were several very jarring moments in artists' statements where they referred to "during covid" as if that were in the past. "During covid" is right now and for the foreseeable future!

I keep seeing posts about how to detect AI-generated text. It's always em-dashes or "it's not x, it's y" or random low-incidence words, etc. These things are all crude heuristics.

The real way to detect AI writing isn't by looking inside the text -- it's to read more (human-written) books.

Bonus wrinkle: fewer and fewer places are offering curbside pickup, which means there are fewer and fewer options for getting groceries without exposing ourselves to airborne pathogens (and, you know, The Public™).

Question for my fellow #autistic professionals: have you passed on applying for a job you're qualified to do because it asks for "a proven ability to [x]," and your inner Descartes immediately popped up to be like "I've done [x] dozens of times, but that's merely *evidence*, not *proof*"?

Just me?

It's not that we need to touch grass, it's that we need them to only be able to touch their families through panes of glass.

Which as broadly an anti-carceralist is a disorienting position to find myself supporting, but here we are.

I'm reading "We Will Rise Again," an anthology of resistance fiction curated by @older.bsky.social and others, and I find myself hoping for a similar volume of "it's hard but necessary work" stories about processing individual and collective guilt without narcissistically centering the guilty. 4/4🧵

I think the time for that work to start is now, if it hasn't already. If we wait to start until the "crisis" has passed, the tools won't be ready before the rationalization fully calcifies. These are seeds that must be planted in freshly tilled soil, iron that must be struck while hot, etc. 3/4🧵

someone is out there who's going to gather (or is already gathering) some international historians and experts to put together a We Done Goofed starter pack for the US to use in making amends, both internally and externally, assuming it survives the current *gestures vaguely*. Importantly, ... 2/4🧵

Not that places like Germany and South Africa don't still have their own problems, but to my knowledge they're among the only countries that have done any kind of formal, intentional reckoning with their nation-level "missteps." I'm extremely not qualified to do this, but I really hope that... 1/4🧵

Just finished @goingmedieval.bsky.social's excellent "The Once and Future Sex" and felt throughout like it'd make an excellent companion piece to @adapalmer.bsky.social's also-excellent "Inventing the Renaissance."

If you've read and liked one, consider picking up the other!

How would you feel about being addressed as a customer by the judge hearing your case, or the professor teaching your class?

The USPTO may provide a service, but it's not a business. I may be a patron, but I'm not a customer.

Spent ~90 minutes on hold with the USPTO. The most frustrating part was hearing "our specialists are busy assisting other customers."

It feels gross to be addressed as a customer by the government.

"Customer" implies a distance that just isn't apt for institutions we're constituents of.

Immigration is good, actually
Fascism is bad
I thought we learned this over the past century
and really wish that we had

What It Takes To Love Your Job And who gets to do it

I'm wondering if there's quantitative work that's been done on the qualitative issue I discuss in this extremely-relevant essay: medium.com/brain-labs/w....

It seems to me that work can be a source of meaning only to the extent that it gets you beyond survival. One can't love mere life support.

When I was 16ish, my stepfather taught an intro-level lit class at a major university. One day I came home to find the parents laughcrying. Turns out an undergrad's essay about The Cherry Orchard had contained the phrase "The Russian aristocracy simply did not understand the value of the dollar."

The strong form is that AI must not, without providing thorough and verifiable citations, produce *any* output that could be reasonably expected to cause a human seeing it to form a belief.

AI output can be a *component* of human thinking, but should never be its terminus. 3/3🧵

The weak form is that an AI must not produce any output that a non-expert viewer would perceive to be authentic (photos of real people; videos of real events; quotes from real, human-written books; etc.).

AI-generated content should always be immediately identifiable as such without forensics. 2/3🧵

I've been thinking about appropriate guardrails for genAI tools (assuming we can't yeet them into the sun). Most of the ones floating out there are content-specific: "don't teach people how to make bombs," etc. What I'm proposing here is content-agnostic, and comes in a weak and a strong form. 1/3🧵

That is: what does accountability look like in a world without evidence? What happens if we can't "check the tape"? How do we speak truth to power in a post-truth world?

And is it any surprise that the biggest pushers of this tech are aligned with an administration so willing to lie to us? 5/5🧵

What happens when, next time, the DHS releases its own authentic-seeming footage of the victim threatening the ICE agent? What happens when they assert that the actual bystander footage has been doctored?

Forensics exist, but they're slower than the court of public opinion or the media cycle. 4/5🧵

But we're not on firm ground in that world. We're rapidly approaching a place, if we're not there already, where generative AI tools can create convincingly realistic photos and videos. That's merely annoying when it comes to bears on trampolines. It's *disastrous* when it comes to *evidence*. 3/5🧵

The problem has 2 main parts. The first part is that the administration has demonstrated a willingness to make brazen, provably false assertions.

The second part is that those assertions are *provably* false only because we live in a world where evidence exists: we can point to the video. 2/5🧵