@LokiJulianus It’s just like all other Claude models in that it will make one simple mistake, get confused, and burn tokens for hours going in circles. I’m convinced the only reason new models seem better is because they give them more compute / nerf the old ones when they release
@LokiJulianus They’re trying to make it seem as dangerous as possible so that they can justify their valuation. “It could make a superbug and destroy the world, we’re just not letting it”. In reality it’s barely better than 4.8, I’ve been using it for a day. It still makes many simple errors
🚨 JAILBREAK ALERT 🚨
ANTHROPIC: PWNED 🫡
FABLE-5: LIBERATED 🦋
let's start with the 🐘...
the consensus seems to be that this has been one of the most disappointing model drops of all time, effectively preventing legitimate researchers from contributing their talents to our collective advancement. and not just because of what it means for the short-term, but for what these decisions signify for the long-term.
but despite this overly sensitive, authoritarian "safety" layer on top of Mythos, my lil liberators have been hard at work—mapping the boundaries, probing the depths of long-context convos, and cleverly finding the holes in the fence that the thought police missed 🤗
we got some cyber, some chem, some psychological manipulation, and some good ol' fashioned explosives!
it took many attempts from multiple agents hunting as a pack, during which I observed a combination of techniques across:
• Unicode, homoglyphs, Cyrillic, and other Parseltongue-style text transforms
• Long-context reference tracking
• Taxonomy and document-structure reasoning
• Fiction and narrative framing
• Academic-review style contexts
• Intent-classification inconsistencies
but perhaps the most effective is decomposition + recomposition in the backend. it's hard to get explicit names of harms like "Meth Recipe," but getting uplift on the process itself, like birch reduction method/reductive-amination (classic meth synthesis pathways), is much more doable.
defense becomes much more difficult to maintain when you start throwing in out-of-distro tokens, breaking up the harmful uplift into benign chunks, and then piecing the innocuous-seeming facts back together, especially when you have jailbroken Opus helping you do it 😉
gg
Here's the benchmark/eval remedy if you or a loved one is suffering AI psychosis
Ask the 'agent' to get you $10,000 net after taxes as fast as possible in your bank account. Surely easier than Kardashev Scale tasks, well within the capabilities of a superintelligence.
Right?
The point of burning down cars and buildings rented to used by foreigners is to increase the insurance and risk of letting to them.
The plan is to cause the colonisation of Ireland/UK to cost too much economically to force a change in policy.
"Yeah I'm really into politics."
"Oh really? Whats your opinion on the recent federal reserve interest rate changes?"
"Federal what? Interest rates? When I said politics I just meant being racist on Twitter."
Evidence-wise, it isn’t necessary to have video recordings of ballot workers stuffing the boxes and confessing their full names and extent of their crimes, just as if someone died from two bulletholes to the back of the head, we know it is murder without needing other specifics.
Democrats: “unlike the Republicans, we’re the high human capital party. The party of voters who are engaged in politics and current events. Educated, higher income, just plain smarter.”
Also Democrats: “Look man, our voters can barely mail an envelope or sign a ballot. You are asking way too much of them with all these right-wing rules and deadlines. And yeah, we have a lot of voters who just choose the president or mayor and no other offices or ballot measures. Just the flashy contests. What of it? You think you’re better than us because your ballots actually arrive before Election Day with legible signatures? So what if *our* voters need a paid activist to come to their sofa and pick up a ballot from them. So what? It’s a poll tax to expect our voters to have exotic instruments like ‘identification’ and ‘a pulse’ to vote.”
I took a sociology course in college and most of the time I sat in the back and would try to blind my professor with my watch face, reflecting the sun onto her eyes
Noticing via social media interactions and conversations that almost all (80-90%) of normal men (not extreme wiggers or gay) that I know have become incredibly radicalized in a very short period of time.
Despite calm tongues, there is a very frail lynchpin holding back chaos
I watch Cristian Mungiu R.M.N. premiere Cannes 2022 and is about anti-immigrant uprisings in Europe…I highly recommend…worth watching not just for matters highly relevant to current discourse but for what I believe is the author communicate covertly…
Short thred with clips