Anthropic has pushed AI forward dramatically over the past two years. It's currently the crown jewel of US AI tech.
The Feds don't like @DarioAmodei because he won't do all their bidding. And so, we've now entering the Soviet-style propaganda portion of the program with the White House feeding every reporter it can find with laughable claims like Dario is unreachable at a wellness retreat. Come on.
I'd hoped the US would not be self-defeating on AI, since it's kinda one of the last hopes the US has versus China. But here we are . . . . already
Walking a funny line of trying to blame Anthropic’s marketing, but also saying their model is in fact too powerful, but also promising they want to return Fable for wide release ASAP so please don’t blow up AI stocks
@daniel_mac8 Anthropic stated that they would want to make Fable and 'Mythos-class' models subscription-available as soon as compute allows. Hopefully that does not take too long.
@logangraham Please pay attention to the high rate of false positives. I asked it to review a paper I'm writing on the EU Chips Act 1.0 and it flagged it and routed the request to Opus 4.8. Just mentioning "cyber" also triggers this.
Many are also reporting false pos. for simple bio/med q's
@Mononofu My this, maybe. Fable mistakenly refuses you half the time. Add that to the secret “sabotage” and it’s super shady. And I love Anthropic. https://t.co/VDLUV8od88
@TobinSouth Fable guardrails are extremely trigger happy (including possible memory bugs) and the model card language implies silent sabotage.
it is extremely frustrating and makes you paranoid about being randomly rejected and being routed to Opus.
This is NOT good user experience.
@TobinSouth Fable guardrails are extremely trigger happy (including possible memory bugs) and the model card language implies silent sabotage.
it is extremely frustrating and makes you paranoid about being randomly rejected and being routed to Opus.
This is NOT good user experience.
@logangraham Please pay attention to the high rate of false positives. I asked it to review a paper I'm writing on the EU Chips Act 1.0 and it flagged it and routed the request to Opus 4.8. Just mentioning "cyber" also triggers this.
Many are also reporting false pos. for simple bio/med q's
Degrading performance on ML research *without telling the user* is shockingly hostile and a terrible look. That could silently damage all sorts of work, including some of my own. Also the type of thing that could raise the eyebrows of antitrust enforcers worldwide.
@soilupdates@Aella_Girl Yeah, the key feature of nationally representative surveys is that *you know the selection probability of every unit* and can make adjustments for non-response (self-selection) accordingly