I’m just going to pretend that’s one of the researchers from Where Oaken Hearts Do Gather.
I’m just going to pretend that’s one of the researchers from Where Oaken Hearts Do Gather.
I conclude that scheming is a disturbingly plausible outcome of using baseline machine learning methods to train goal-directed AIs sophisticated enough to scheme (my subjective probability on such an outcome, given these conditions, is ~25%).
Out: vibes and guesswork
In: “subjective probability”
I felt the exact same way about the conversation you mentioned. I really liked the idea of the quest, but way they handled it just utterly drained all the stakes. And as you noted, it’s weird to see a misstep like this after they nailed it once in Sumeru.
“We’re all in grave danger! What? Well no, we can’t give specifics unless we risk not getting paid. Signed, Anonymous”
I mean, I wasn’t exactly expecting the Einstein-Szilard letter 2.0 when I clicked that link, but this is pathetic.
lmao, Zoom is cooked. Their CEO has no idea how LLMs work or why they aren’t fit for purpose, but he’s 100% certain someone else will somehow solve this problem:
So is the AI model hallucination problem down there in the stack, or are you investing in making sure that the rate of hallucinations goes down?
I think solving the AI hallucination problem — I think that’ll be fixed.
But I guess my question is by who? Is it by you, or is it somewhere down the stack?
It’s someone down the stack.
Okay.
I think either from the chip level or from the LLM itself.
If I was them (and in a way I am) I’d probably kill the witnesses and bail.
So I totally get this conclusion, but I think it’s worth slowing down and considering whether this makes as much sense as it seems at first glance. The fact that magic exists means that simply killing someone simply doesn’t do much to shut them up, if a sufficiently powerful entity is willing to spend the resources. The fact that undeath exists means that killing someone has a very real risk of making them a bigger threat than they were in life–it’s not like you can just stab a ghost. Cultists, being familiar with eldritch powers themselves, know this full well. This means that keeping people merely out of communication might be the simplest way to achieve their actual goal with the minimum of fuss. They don’t need someone quiet forever, they just need enough secrecy to achieve their goals. Murdering every person who takes an interest in them is mission creep.
Also, keep in mind cults generally exist for specific purposes, and people join them for specific purposes. These purposes aren’t necessarily overtly evil at the rank-and-file level, which is integral to their recruitment. The turnip farmer who wants to resurrect a dead harvest god to grow more turnips might be okay with some dodgy rituals the church wouldn’t approve of, but straight up committing multiple murder might take some working up to, if he can be talked into it at all.
So in short, consider what your cult wants, and assume a degree of rationality and thoughtfulness (at least, when they’re not channeling horrors from another plane). What do they want, and how the party could provide what they want?
OpenAI: “Our AI is so powerful it’s an existential threat to humanity if we don’t solve the alignment issue!”
Also OpenAI: “We can devote maybe 20% of our resources to solving this, tops. We need the rest for parlor tricks and cluttering search results.”
If they’re really lucky, they’ll end up working for the Laundry only once. Residual Human Resources is a bad way to go out.
Charles Stross’ Laundry series is basically this concept set in the present day: magic is a branch of mathematics, which means it can be computed and programmed.
It is perhaps worth noting at this point the series genre is cosmic horror.
Orbital stratagem timings make no sense, and are strictly a gameplay balance issue that *cannot* be realistic: the loading screen shows the first helldiver drops well outside the atmosphere and take several minutes to reach the ground, but turrets take 3 seconds to deploy?
I assumed this was because equipment can endure acceleration that would make a person pass out, or at least be combat-ineffective on landing. A trip from the Karman line to the ground in a few seconds would involve some deeply unpleasant G-forces…in opposite directions, back-to-back.
Come to think of it, this might explain why different gear has different call down times, as more fragile stuff might require a slower and (relatively) gentler drop.
For safety reasons, a magical museum might well want to avoid putting actual items on display, especially since Illusion magic makes it trivial to create simulated substitutes crowds can safely interact with. This also keeps security budgets manageable, since even a magical museum probably doesn’t have the kind of money to protect the arcane equivalent of a small nuclear arsenal.
So the museum researchers might be happy keeping items on indefinite loan, as long as they know the borrower and have guaranteed access when they need it. This way the mage owning the item is responsible for keeping it safe, and arcane historians–who definitely didn’t spend all those years in magic academy just to play guard–don’t have to.
As a bonus, this method gives the party a quirky but knowledgeable NPC contact who can give them clues or set up sidequests whenever you need.
I’ll preface this by noting that the sin of sloth has traditionally been understood to be a sin of omission, not just commission, i.e., you are insufficiently devoted to the things you ought to be.
Which means you could, in theory, have a (reflavored tiefling) devil paladin so devoted to sloth he works against evil causes. He’s not interested in good per se, it’s just that advancing the interests of good and traveling with a good adventuring party has the best ROI for failing to carry out his evil responsibilities.
Naturally, this has caused a fair amount of controversy among sloth devils, and there is a multi-century trial going on in the Hells about whether this ought to be allowed. This is not expected to be resolved in the foreseeable future because the advocates for both parties keep filing their responses well after petition deadlines expire.
Oof, I definitely did that once or twice.
It really does seem like they decided to bring this sequence up to introduce settlement building and power armor early. I get why they did it, but man, I do not think it ultimately has the effect they wanted.
I think video art and art commentary is great, but lumping it all in here would make this place less interesting to me. I’m also not sure anyone looking for video art would think to look here for it. Maybe having periodic threads to post commentary or non-image art could be a good compromise?
I acknowledge that’s somewhat arbitrary, but I don’t think we can ever eliminate that entirely where art is concerned.
Maybe a controversial suggestion, but my advice is to ignore the Minutemen stuff until late in the game. Just don’t even go to the museum until you’ve followed some leads and want something else to do for a bit.
This is definitely not the intended way to play, but I promise the story flows so much better without it. Setting out to find your kidnapped son just to immediately get sidetracked helping some uncharismatic misfits set up mattresses is just an underwhelming start to an otherwise decent game.
Doing all this stuff later on, when you’ve actually demonstrated you’re a badass survivor and the OP gear you get free from the Minutemen quest actually feels earned, just feels much smoother. It’s a great coda that they unf put two minutes into the game for some reason.
It’s safe to say it’s better, though I couldn’t tell you if it’s actually fun yet. For one thing, they’ve moved away from some of the wackiest design decisions, like the lack of NPCs.
B.S., a former partner at Andreesen-Horowitz
Lmao, of course.
and former chief technology officer of Coinbase
I.e., the company that survived by shedding a ton of employees, like 40% of headcount or something. I do not see this tactic working well when they’re trying to win friends and influence cops by hiring their failsons into sinecure positions.
What makes all this funnier is that it’s trying to thread the needle of embracing fascism while simultaneously seething with contempt for 99.999% of the people in the movement.
“If you think about the major journeys within a restaurant that can be AI-powered, we believe it’s endless.”
My dude, you work for Yum Brands, not Starfleet Command. Nobody taking a “major journey” inside a Pizza Hut needed AI help to get there. (Though they could probably use a cup of water, and maybe an Uber home.)
Nice, I wonder if this thing can help fill out my restaurant/tissue bank paperwork now.
We know 85 is the upper bound, but I wonder what Rohrer would consider the minimum number of “exchanges” acceptable for telling someone their loved one is in hell? Like, is 20 in “Hey, not cool” territory, but it’s all good once you get to 50? 40?
Always a good sign when your best defense of the horrible thing your chatbot says is that it’s in context.