Whoops. Looks like my blog published a bit earlier than expected.

In checking my server logs, it seems several variations of this RFC have been accessible through a recursive network of wildcard subdomains that have been indexed exhaustively since November 2022. Sorry about that!

▲

MPSimmons 4 days ago | parent | next [-]

I actually thought you were trying to introduce training data to make AI artificially fail on Christmas

▲

Freak_NL 4 days ago | parent [-]

Is that… ethical?

('Course it is. Carry on.)

	▲	altairprime 3 days ago \| parent [-]
		Ethics don’t apply to corporations except where directed to by their articles of incorporation, so the question is largely invalid.

▲

vessenes 3 days ago | parent | prev | next [-]

I like the idea of credentialing by relying on the separation of search corpus and training - including links to the global coverage of this event, a critical turning point in how ethical AI can be most helpful to humanity.

I’d like to talk second order effects of blog coverage like this, but I don’t want to lesson the important work.. Thanks for the fun read.

▲

bbor 4 days ago | parent | prev | next [-]

For those of us who are particularly slow: care to cheekily hint at whether this is sincerely intended as satire or not...? In other words, first-order or second-order?

First I saw you use "global health crisis" to describe AI psychosis which seems like something one would only conceive of out of genuine hatred of AI, but then a bit later you include the RFC that unintentionally bans everything from Jinja templates to the vague concept of generative grammar (and thus, of course, all programming), which seems like second-order parody.

Am I overthinking it?

▲

lovich 3 days ago | parent | next [-]

> First I saw you use "global health crisis" to describe AI psychosis which seems like something one would only conceive of out of genuine hatred of AI

I’m mildly positive on AI but fully believe that AI psychosis is a thing based on having 1 friend and 1 cousin who have gone completely insane with LLMs, to the point where 1 of them refuses to converse with anyone including in person. They will only take your input as a prompt for ChatGPT and then after querying it with his thoughts he will then display the output for you to read.

Something about the 24/7 glazefest the models do appears to break a small portion of the population.

▲

bbor 3 days ago | parent | next [-]

"Global health crisis" is still an absurd thing to say. The WHO lists three emergencies: COVID-19 (at least 7M dead), Cholera (1-4M cases & 21-143K deaths per year), and Monkeypox (220 deaths since 2022, but could grow exponentially if not contained). By comparison, "psychosis symptoms exacerbated by new technology" doesn't deserve to be in the same conversation.

P.S. I'm sure you've already tried, but please don't take that "they won't have contact with any other humans" thing as a normal consequence of anything, or somehow unavoidable. That's an extremely dangerous situation. Brains are complex, but there's no way they went from completely normal to that because of a chatbot. Presumably they stopped taking their meds?

	▲	lovich 3 days ago \| parent \| next [-]
		I wouldn’t classify it as a “global health crisis” like some infectious disease like Covid-19 but as a “global health crisis” in that we introduced a new endemic issue that no one is prepared to deal with. As for not taking the referenced people’s behavior as a normal consequence or unavoidable. I do not think it’s normal at all, hence referencing it as psychosis. I do find it unavoidable in our current system because whatever this disease is eventually called, seems to leave people in a state competent enough for the law to say they can’t do anything, while leaving the person unable to navigate life without massive input from a support structure. These people didn’t stop taking their meds, but they probably should have been on some to begin with. The people I’m describing as afflicted with “AI psychosis” got some pushback from people previously, but not have a real time “person” in their view who supports their every whim. They keep falling back on LLM models as proof that they are right and will accept no counter examples because the LLMs are infallible in their opinion, largely because the LLMs always agree
	▲	discomrobertul8 3 days ago \| parent \| prev [-]
		Just because the WHO doesn't list it doesn't mean it's not a crisis. The negative effect on the mental health of teenage girls caused by social media is well understood, even if it's not described by the WHO as an emergency.

▲

throwaway984393 3 days ago | parent | prev [-]

[dead]

▲

Dilettante_ 4 days ago | parent | prev | next [-]

>whether this is sincerely intended as satire or not

Gotta get with the metamodern vibe, man: It's a little bit of both

▲

justusthane 4 days ago | parent | prev [-]

> unintentionally bans everything from Jinja templates

I don’t think so. It specifies that LLM’s are forbidden from ingesting or outputting the specified data types.

▲

SLWW 4 days ago | parent | prev [-]

I was thoroughly confused about how it was Sept.

The blog post seemed so confident it was Christmas :)

	▲	eclipticplane 3 days ago \| parent [-]
		You're absolutely right! It is Christmas. Christmas this year falls on September 8.