AI may be an existential risk simply because the human mind and society as currently constructed is simply incapable of judiciously using it at scale by undeliberative people and everything turns to slop as its permission structure of incompetence and 1-shot tries effuses into culture and pipelines. - ThreadSky

swiftonsecurity.com • 4 days ago

AI may be an existential risk simply because the human mind and society as currently constructed is simply incapable of judiciously using it at scale by undeliberative people and everything turns to slop as its permission structure of incompetence and 1-shot tries effuses into culture and pipelines.

Reposted from Los Angeles Times

State Bar of California admits it used AI to develop exam questions, triggering new furor

Comments

swiftonsecurity.com•4 days ago

Where, "What do you expect, they couldn't afford to do more than have an AI generate it" is the real mind virus

swiftonsecurity.com•4 days ago

https://bsky.app/profile/ncweaver.skerry-tech.com/post/3lnimk77x7k2g

agtmadcat.bsky.social•4 days ago

Easy enough fix for that: Automatic disbarrment of all partners at a law firm if anything AI-generated goes out the door. After you nuke the first couple of major law firms the problem would disappear pretty quickly.

oxwof.com•4 days ago

It would be a reasonable use case for a limited (non-generative) AI sort of thing to search for citations in a document and then check databases to see if they actually exist

cfit.bsky.social•4 days ago

We've had that for years, including telling you if the cases have any subsequent negative history. Not "AI" though.

meatieocre.bsky.social•4 days ago

So that's why my house collapsed. Who's liable for that?

r3nt.bsky.social•4 days ago

So people are allowed to answer them with AI, too, right? Right?

zendarva.bsky.social•4 days ago

I mean... Most tests use reference texts when being written, and you're generally not allowed to use those same references.

Not sure why this would be different?

sofiaareinaa.bsky.social•4 days ago

And the answer is not “let’s use more AI to solve the problems AI is causing”

eitherorlok.bsky.social•4 days ago

This touches on part of my job, which is to review Generative AI integration requests for my California state-level Department. What the Times story isn't clear on is whether they *even conducted* the mandatory review of the GenAI tools and received approval from the Office of Information Security.

eitherorlok.bsky.social•4 days ago

If the story gave more detail about the timeline in the development of the test, we could see if they tried skirting the (relatively new) rules that have been established by outsourcing it or rushing it to production before the rules took effect.

concoloris.bsky.social•4 days ago

Incompetent Manager As A Service

spooftrooper.bsky.social•3 days ago

Good point. We don't need omnipotent AI to destroy each other with AI.

zendarva.bsky.social•4 days ago

This.... Is probably the best take I've seen.

Not ai at fault, but out shitty mental rigour, and intellectual hygenie.

fulgurbellator.bsky.social•4 days ago

I can’t fucking wait for Reb Masel to get ahold of this article lmao

baghwanb.bsky.social•23 hours ago

I'm thinking that assuming 1-shot may be generous.

nelsonrarebooks.com•4 days ago

Pshhh my AI does all my thinking for me.

nobodyholme.bsky.social•3 days ago

Using an LLM for something is an admission you: do not have competence in the subject, and do not have time to gain that competence.
It’s an existential risk because we as a society are incapable of correctly determining when that’s ok and when it isn’t. I don’t agree that the human mind isn’t.

Comments

Posting Rules

Reply