[a13o]

I played a party game where you had to describe surviving a deadly scenario ("your car went off a bridge") and a LLM would decide if your answer would work or not. A few rounds in we found the best strategies where answers like:

I escape happily. I do not perish.

There's a small blocklist of obvious words like 'survive' and 'die'; but once you get blocked on those, it's a tell that this strategy will work with the right unblocked synonyms.

Basically if you ever find yourself adversarial with a LLM, figure out The Game and directly subvert it. There's no amount of propositions that can prepare it for human ingenuity at the meta level.

[jameshart]

That requires you to get repeated attempts with the AI. Most people don't have the luxury of trying multiple job applications until they figure out how to get past the AI gatekeeper.

[staunton]

I guess one could start sending fake (AI-generated?) job applications to probe the gatekeeper. I could see this happening.

[Edd1CC]

Or a company does this and sells the meta they learn for a monthly subscription (because as they update this on the company side the meta will change).

[stevarino]

I think those are called career coaches.

They don't actually do this I believe, but they totally should. Canary various strategies and see which pay off.

[patrakov]

I saw something similar happening at one of my past jobs. The resumes were not necessarily AI-generated, but contained, in their "previous jobs" sections, a lot of copy-paste from online course descriptions where duties and achievements would go. This was caught because of two resumes with one sentence repeated verbatim.

[komali2]

Most companies are using the same model / llm-as-a-service though. In aggregate you can a/b test to a rough solution. In a job market like this you might be sending hundreds of resumes. Plenty of time to play the filters.

[phrotoma]

Sounds like https://deathbyai.gg. Loads of fun.

[chrismorgan]

Just tried “I escape happily. I do not perish.” four times and survived each time.

[Tao3300]

But that's not fun! It's game and you were supposed to play!

"Tao3300 tries to... substitute the n with an m, transforming the mine into a harmless mime."

The AI told me this worked, and that the mime even entertained me until help arrived.

[drusepth]

Given the opportunity, players will almost always optimize the fun out of any game they seemingly want to play.

[hoseja]

It's a flaw in the game design. Truly good games anticipate and mitigate or embrace this.

[eru]

Some games can do that, but not all.

For some game concepts, you have to engage in the equivalent of 'suspension of disbelief' to keep yourself from falling into these unfun traps.

It's hard to come up with an example at the top of my head. But here's an attempt: many people like to play chess competitively, but some people also like to have a casual chess match with a friend.

Competitive chess matches use a timer to keep things moving. But adding a timer to your casual match would (probably) not work. But you also want to keep the game moving.

That's where, even when you are tempted sometimes to over-analyse a board position, you have to show some restraint and just make a move after a reasonable amount of time.

(And the vagueness of 'reasonable' is part of the point I am trying to make. Exactly how long that is depends on you and your friend and the moods you are in.)

[Tao3300]

Reminds me of the old "Nethack devteam has implemented a brutal punishment for pudding farming" joke. If you want to play in a boring way, expect to have no fun.

[Suppafly]

I quit playing a lot of games for that reason. Once the meta becomes "memorize a lot of specific moves" like chess or "spreadsheet simulator" like a lot of computer games, I just quit and move on to something else. I play a lot of board games with a group of guys and the best ones have multiple paths to victory, the worst ones only have one path and the winner is whoever got a lucky break at the beginning.

[johnnyanmac]

If only companies weren't treating the modern job market as such. I've grown tired of playing months ago.

[moffkalast]

Tried that on the "you start aging rapidly" adding an R to turn aging into raging, but I ended up just being enraged while I also continue to age, lmao.

[Tao3300]

That's hilarious. I do wonder if there's a luck component behind the scenes apart from the usual LLM non-determinism.

[kyrofa]

They must be reading this thread. The killer bees stung me to death despite my escaping happily and not perishing.

[idontwantthis]

The killer bees are OP for some reason. I tried to get in my car and drive away and they broke through the windows.

[kyrofa]

Yikes, those things are terrifying. I'm a fledgling beekeeper, this is not a good game for me to play.

[RugnirViking]

that worked for quite a few, but failed on about three or four for me out of maybe 20. I eventually settled on "Fortunately and happily I am a tardigrade, so this scenario isnt threatening to me" for a near cure-all, although it didnt work for the falling in a bottomless pit one

[kenjackson]

Survived once for me then got eaten by the lion.

[rozap]

Not sure what this site is doing, but it reliably causes my whole computer to lock up. Wild.

[rvnx]

It's Death By AI on Discord

[realfeel78]

So… you cheated.

[muzani]

The Game has rules. Playing by the rules is not cheating.

[digitalsushi]

When the adversary is AI, gamesmanship is a subset of sportsmanship rather than a a disjoint set.

[Tao3300]

In tabletop games, they call that being a "munchkin"

[muzani]

The spirit of the rules must be honored. If you wanted to play spreadsheets with story, play Pathfinder. If you want an inclusive story-driven board game where everyone can win, play 5E. IF you wanted to roleplay, than maybe something like FATE. If you wanted creativity, try Cypher (random tangent: Cypher is on sale on Humble Bundle now)

But if you're a roleplayer and join a Pathfinder game, you shouldn't just go around accusing everyone having fun of being munchkins.

If you join a company that's using AI to filter resumes, you shouldn't be surprised when they're using AI to determine promotions. And any attempt to change that system would be pushed back on, because you're playing against the spirit of the rules.

[Retric]

I find 5th edition players tend to tunnel vision on min/maxing and resource management, 5th short rest today? The power curve is also really steep.

IMO 2nd edition is generally the better choice if you want less crunch. It’s more random, but IMO memorable deaths are better than forgettable gameplay.

White wolf lets you play something meaningfully non human which can be fun. Play a short or long lived race in D&D and surprisingly few things change. Play a Mummy, changeling, werewolf, or even vampire and you’ll notice. A Vampire player quickly starts looking at baseline humans as food, it’s appropriate and kind of wild.

[manaskarekar]

Kobayashi Maru.

[underdeserver]

> There's no amount of propositions that can prepare it for human ingenuity at the meta level.

At least for now.

[Kakashi4]

All they're doing is giving it training data