what jumps out at me in this paper is the idea that "mirage" reasoning is attached to an internal (and incorrect) "epistemic frame." i think this analysis is misguided. what the LLM is doing shouldnt be understood as an "incorrect judgement." it should be understood as *commitment to the bit*.
Great ty
They could get rid of labelers, potentially. Or gatekeep who is allowed to run them (though I doubt anyone has the appetite/skills to manage the governance and social work required to do that)
(Ignorantly) should one try to run this in opencoder? Another harness?
I’ve been without an income and no prospects for employment for almost a year now, so, not great
It’s always Thursday
Also a variety of qualitative and subjective layers of provider+user specific prompting and harnessing techniques (which themselves also need to be updated for each release from just the one provider)
The ladder is up and you aren’t on the rocket either, stop groveling and start helping us break the chains (the real ones, not the golden handcuffs they gave you)
“The moat is taste”, “the moat is execution”
My brother in capital the only moat now is already having billionaire money or being (temporarily) on the downline of the masters
NEW: Apple fixed the bug that law enforcement, like the FBI, were taking advantage of to extract chat messages that had been deleted or disappeared automatically.
Until now the iPhone stored deleted or disappered messages in a database, allowing authorities to access them with forensic tools.
The only thing we can all agree on is that no matter what your opinion is about AI, whether or not you like using it puts you in an existentially threatened community.
Symmetric radicalization spirals are all around me
Just like it’s good to not willingly adopt a misinformed or conspiratorial worldview to score points on your ideological opponents, it’s also good to not interpret disagreement by people who have never interacted with you as harassment requiring reprisal.
Just an engine, sitting in the middle of the road without a chassis
But I don’t know the process! I’m figuring out the process by myself from first principles guided largely by my big feelings.
Maybe I should even talk to them about it 🫣
Hoist upon his own pendulum
I keep finding books and journals by people about the niche topic I wanted to write a book about and at first I had some challenged/competitive(?) feelings about it, but actually I think it means I can just read and build on (or spin away from) their stuff, so that’s good, right?
This would all be much simpler, architecture-wise, if atproto had instances; so that they wouldn’t have had to reinvent defederation wars using the ungainly but novel combination of affordances of blocklists/labels and feeds
LLMs are an accelerant of whatever mindset you bring into them, I guess
Interpersonal drama AND ideological adherence tests!
Grok, what is the intellectual equivalent of the pro-ana movement?
Just like slop comes from the heart and not from the models, cognitive surrender starts long before one touches a prompt
It’s the same thing as that Karl sad trombone guy was doing with his anti-“fascist” list or whatever
“don’t be a dick to a feed creator” is funny when the feed creator is blocking 300,000 people because they all got put on a list by some *other* person, and that other person wasn’t listing them due to bad interpersonal interactions either but due to automatically scrapping for ideological adherence
It’s such a MAGA conservative style mindset. Pure inability to tolerate beliefs and values from outside of one’s sealed information environment, while proactively seeking ways to troll or disrupt those outsiders yet being unable to take any criticism for it.
I do think it’s funny to consider the mindset of someone who thinks willingly adopting a technology puts them in a class that is existentially at risk by a crisis precipitated by all of the people who chose not to adopt (or, realistically, are simply unenthusiastic about adopting) that technology
I came back because it has clearly deteriorated further, even, and I decided that it was more funny than annoying
So I know bluesky added that little robot icon so we can know which accounts are bots (which some humans proceeded to add to their accounts which makes it entirely useless, which I’m sure was the point), but did they add any UX affordances to let users control interactions with bots, or no?
