@ai6yr I don't think that's a correct interpretation of the paper. They say they had humans look at many model responses and then validated the LLM judgement. My day job is on one of the projects they cite, and we take a similar approach, because it's not […]

[Original post on sfba.social]

@williampietri Mine is "I don't get it".

As in someone sees something they don't understand and instead of acquiring more information and then making an educated opinion, they just judge it based on peer pressure "opinions" and then go live their life with that opinion set in stone about the […]

@apzpins Absolutely! Same energy of intellectual laziness and assumed superiority.

@sundogplanets I think now isn't a good time to change much about how you work, but it can be a good time to amplify what works for you. It sounds like lists work, so perhaps you could create some reusable checklists?

One set would be the procedures you'd want […]

[Original post on sfba.social]

@sue

These people need an education in programming just so they internalise that "I don't know how this works" is exactly equivalent to "I don't know if this works" or even "I don't know what this does".

I think the basis for this misunderstanding is the belief that LLMs are a new abstraction for coding, but as I've said before I think that is totally wrong framing here. Code generation does not equal a programming abstraction.

Well bummer, there are good people who follow me there, so I can't block the whole thing. Thanks for the suggestions, folks!

@aburka @xgranade According to the paper, they had humans label 2100 items and then compared that with the LLM evaluators. "We observe strong
agreement across annotators and between human ratings and the LLM-judge consensus, consistent with findings reported in recent LLM-as-a-judge literature […]

@soaproot I get regular emails from them, and they come to an address I gave only to the USPS. It's not totally a lock, but it'd have to be a very sophisticated attacker.

@futurebird Have your seen Axideaw? https://axidraw.com/

@hacks4pancakes Given that lunatics are in charge of so much. I think hoarding medication is a totally reasonable choice. As best I can tell, the Republican preference for replacing the Affordable Care Act is, "die, fuckers". I think we can keep them from getting there. But the last decade I've […]

@Sarahp Ooh, as a former T&S person, I really like this framing: "At the root of the issue is how Bluesky wants to be perceived versus what it actually is today." That's often the struggle, and one of my big questions is how much a given change is substantial vs for PR value.

To make it easier, maybe the site lets you print stickers. Or maybe it sells rubber stamps, like Where's George. Maybe it also lets you enter package tracking numbers and shows you comparison stats there. Maybe it lets you enter only on receipt based on the postmark. Plenty of fun things to do […]