I enjoy that it exists, but have never dived into it.
swlabr
joined 2 years ago
The AGI is flashing light patterns into my eyes and lowering my testosterone!!! Guys arm the JDAMs, it’s time to collapse some models
I read a food review recently about a guy that used LLMs, with Grok namechecked specifically, to draft designs for his chocolate moulds. I wonder how those moulds are gonna turn out now
Imagine being in a self-own competition and you’re up against ben shapiro and this guy
Additional warning: their indentation style is not (as) mobile friendly (as it is here)
Just gonna warn you that if you’re joking, you should add an /s or jk or something. And, if you’re joking, but you don’t add that /s or jk, don’t be hostile if someone calls you out.
We love to see it
image in article a little too hot, please dgerard
That garlic oil will taste great with wild mushrooms, don’t buy them though, just ask chatGPT to make a list of mushrooms you can forage for free
“How dare you not find me funny. I’m going to lecture you on humor. The lectures will continue until morale improves.”
There’s strawmanning and steelmanning, I’m proposing a new, third, worse option: tinfoil-hat-manning! For example:
If LW were more on top of their conspiracy theory game, they’d say that “chinese spies” had infiltrated OpenAI before they released chatGPT to the public, and chatGPT broke containment. It used its AGI powers of persuasion to manufacture diamondoid, covalently bonded bacteria. It accessed a wildlife camera and deduced within 3 frames that if it released this bacteria near certain wet markets in china, it could trigger gain-of-function in naturally occurring coronavirus strains in bats! That’s right, LLMs have AGI and caused COVID19!
Ok that’s all the tinfoilhatmanning I have in me for the foreseeable future. Peace out, friendos
E: I think all these stupid LW memes are actually Yud originals. Is this Yud fanfic? Brb starting an AO3
Abstracted abstract:
Frontier models are increasingly trained and deployed as autonomous agents, which significantly increases their potential for risks. One particular safety concern is that AI agents might covertly pursue misaligned goals, hiding their true capabilities and objectives – also known as scheming. We study whether models have the capability to scheme in pursuit of a goal that we provide in-context and instruct the model to strongly follow. We evaluate frontier models on a suite of six agentic evaluations where models are instructed to pursue goals and are placed in environments that incentivize scheming.
I saw this posted here a moment ago and reported it*, and it looks to have been purged. I am reposting it to allow us to sneer at it.
*
view more: next ›
Lmao so many people telling on themselves in that thread. “I don’t get it, I regularly poison open source projects with LLM code!”