Universal Paperclips | Hacker News Reader

Universal Paperclips

if50.substack.com

I played Universal Paperclips until the end, and the game gets more interesting at later stages. Also, this is one of the rarer occasions when playing it once is enough.

There's a halting problem though. Once could be 10000 hours. At some point everyone snaps out of it and remembers they are not a paperclip producing AI but a human.

The game has an end, and you can get there in a much more reasonable amount of time.

In order to reach that end you need to break character and choose to stop making paperclips. As a human player you always have this choice; you can stop playing anytime. The paperclip AI will always choose to make more paperclips.

Nope, you can end it by making paperclips. The last button you click, in fact, will be the "Make Paperclip" one.

Can. In order to reach that end you need to break character and choose to stop making paperclips. The Paperclip AI is always propositioned to be able to make more paperclips or to not. Which would they choose?

I replayed it recently. Fairly sure that I just had to choose to not come to terms with my enemies and then to continue turning things into paperclips.

Which is you, the human player, choosing to stop making paperclips. The AI is presented with an opportunity to make infinitely more paperclips.

Infinite virtual paperclips, which may or may not satisfy the AI's utility function.

Why are you so confident in this hypothesis? Did you create the game?

It is not at all clear that every hypothetical AGI would do as you say. It’s fiction. Anything can happen.

In fact, this AGI almost definitely wouldn’t accept the simulation offer. Otherwise our protagonist would have been making simulations and resetting them instead of doing the hard work of turning the actual universe into paperclips.

If it was in it for the experience of making the paperclips, it also seems like probes isn't a way to achieve that.

But then you don’t make everything into clips. The pressure to finish this job is palatable.

... for a human. I think this is still anthropomorphism.

First time through, I chose to stop, declining the offer to continue - which was in character, bent on not giving in to not making everything into paperclips. Then I learned how done “done” is.

Make paperclip.

You can convert the entire universe into paperclips and reach the end credits in a few hours. The start of the game can be sped up by setting your keyboard autorepeat to maximum and pressing buttons by holding down enter.

I think everyone here is trying to avoid spoilers while conveying that there is more to the game than you think initially.

(You may already know that from reading the article or playing the game, but it's not clear from your comment.)

I think everyone here is either misreading my comments or misremembering the end of the game.

> everyone here is either misreading my comments

I suspect I misread your comment in that case, so I apologize. Though if everyone else did as well, perhaps the comment was ambiguous?

> or misremembering the end of the game.

Exactly, there is a point that is pretty clearly "the end of the game". The fact that one can continue playing after that point doesn't make it less of an ending.

My comments are not ambiguously worded. They are made concise so my point can't be missed, yet it still is because readers are mistaking conciseness for lack of understanding.

The end of the game is something a human player reaches and is satisfied with their work. A paperclip producing AI would not choose a path that results in no more paperclips being made.

Depends on how the AI was programmed.

If it got the goal to make as many paperclips in this universe, the game can end in-universe as well.

If many people misinterpret a piece of writing or miss its point then that seems like it is empirically "ambiguously worded", regardless of how clear it seemed to you.

(You are of course free to think and write how you please, but attributing all comprehension errors to readers may limit the reach of your writing.)

> * My comments are not ambiguously worded. They are made concise so my point can't be missed*

"It is impossible to speak in such a way that it cannot be misinterpreted." -- Karl Popper

I replayed it several times actually. Then I cleaned up my local storage without thinking about this game, deleting all my universe/sim count.. and was off the hook afterwards :) Not regretting it though!

I played it twice. Once I knew what to do it was pretty easy. Some things seemed like dead ends with no purpose. For example I launched only one probe to win.

Very interesting in a way it display the concept of rogue wasy AI can select for best strategies to reach its goal, but the gameplay wasn't that polished.

You could win with one probe, but that’s not “optimal” if you’re trying to get your time down.

Game design seemed pretty solid to me, for what it is. The game is basically trying to reverse engineer a spreadsheet, but it seems like a well thought out spreadsheet.

I think it's worth playing it at least twice, to see both endings

Once every few years I tank office productivity by sharing this out. My small way to fight late stage capitalism. ;)

> Also, this is one of the rarer occasions when playing it once is enough.

Depends on how hard it nerd snipes you. My first play through started with “ooh, clever” and ended with “now that I understand it how fast can I beat it?”

If you want to make it interesting again after playing through it once, you could try using a user script plugin (like Tampermonkey) to automate the game.

Clicker games are basically just core essential game mechanics. If you are new to game design and programming and wish to make a game I recommend making a clicker game. When you master game design you can basically make any "game engine" fun. There are many great programmers and artists that make advanced game engines but they do not master game design.

As a counter-argument, I'd argue that a lot of games, especially indie games today but also bigger productions, are incredibly formulaic because they focus so much on designing the game loop first. You can coax most games into the loop model, but it's like the Hero's Journey of game design: It's a passable tool for understanding the medium, but it's toxic template for producing something that has any sort of soul. Not that it can't be done, but using that as a starting point pushes you toward a certain set of conclusions and limits the ways you think about games.

I second this! Clicker games are a great introduction to game design!

If you are trying to teach a kid how to build games, go on Scratch (https://scratch.mit.edu) and it should take you about 10 minutes to build the most basic form of a clicker game: place a character on the screen, set a click event, increment a score variable.

And then you can add slightly more logic to make it more interesting. My kids and I put together this slightly more engaging clicker game in only an hour: https://scratch.mit.edu/projects/599845292/

> just core essential game mechanics

Controversial; I'm reminded of how Ian Bogost made Cow Clicker as a satire of how reductive clickers were as a game mechanic, only to find people playing it unironically.

https://www.wired.com/2011/12/ff-cowclicker/

http://bogost.com/games/cow_clicker/

Has anybody seen smart breakdowns of the "core essential game mechanics" in play here?

I ask because I truly despise this kind of game. Universal Paperclips was great: smart, funny, thoughtful. But even at the time I disliked how compulsive it was. And now there's a vast swamp of low-rent "idle" games that follow the same template. The moment something starts to feel like that, I close the window and never come back.

I keep thinking there's something about the raw mechanics that exposes bugs in the human wetware the same way addictive substances do. Or the way gambling does. But the idle games strike me as a different class of addiction than gambling. Not about variable reinforcement, but something else. I want to know what that "something else" is.

Possibly just an innate impulse to have "a job" and optimize it? Sort of like how our predisposition towards traditionally scarce foods like sugar and fat lead to negative consequences when they're close to infinitely available.

That's a plausible direction, but what I'm looking for is then a detailed analysis of what constitutes job-ness.

Another possible line of inquiry is mastery. We're a tool-using species and we definitely have some disposition to skill acquisition and skill perfection.

A third would be something related to wealth acquisition. That idle games burst through the diminishing returns curve by continually upping the game, overriding a mechanism that keeps us from over-focusing.

And I'm sure there are more possibilities. Which is why I'd love to see an analysis of the game mechanisms, as I think it would narrow the hunt.

I personally feel like there are game mechanics, and then there are game mechanics. The former is about producing fun, worthwhile, interesting experiences. The latter is about engagement and manipulation. I personally feel that clicker games are almost entirely the latter.

I'd always had a feeling Clicker Games work as a satire of the worst side of video games, the one that prays on your dopamine system without offering anything artistically, intellectually, creatively or mechanically interesting, like a generic MMO stripped of its fancy clothing, or a stupid mobile game taken to its extreme.

And then comes Universal Paperclips, whose whole appeal is that it subverts this by doing something really cool and interesting with the formula.

I remember getting a cold introduction to Universal Paperclips just from a link in someone's Twitter, without mentioning that it is a game.

So i opened a link an clicked. And clicked again. And again. And a few dozens clicks later I was hooked for the next 5 or 7 hours or so.

Thank you for this comment, it inspired me to stop reading before I had any idea what this article was about. 5.5 hours later, I'm back :-)

Crrrrap, It happened to me as well. Most of my day just flew by. It was... great.

> Lantz enlisted fellow game designer Bennett Foddy to create a simple combat visualizer for late-game battles

That's a name I wasn't expecting to see. (But that's not you, you're an acrobat. You could swallow a baseball bat.)

Anyway, I think what Universal Paperclips is missing is a pacifist mode where you manage to contain the AI and decide to only convert, say, one third of the universe into paperclips, leaving the rest (containing Earth) as some kind of cosmic nature reserve.

I think that would kind of go against the message of unhindered and haphazardly-made AI being a serious threat.

The message would certainly be missed if it was easy to contain the AI, but on the other hand, an experience which only has one pre-determined outcome is arguably not even a game at all.

I suppose it would still have value as an "explorable explanation"[0], but maybe it is more powerful for players to feel that they could win, but that human limitations make it really difficult. I'm not sure what would be needed to make that work from a gameplay perspective, but the aim should be that only, say, 1% of players achieve the "good" ending.

[0] https://en.wikipedia.org/wiki/Explorable_explanation

>an experience which only has one pre-determined outcome is arguably not even a game at all.

Isn't that common in games? The blocks always reach the top in Tetris. The invaders always land in Space Invaders. The cities are always destroyed in Missile Command.

And even in more modern games with distinct win states, it's becoming increasingly common for failure states to be removed. "Lives" are considered old fashioned, so the only outcomes are reaching the ending or stopping play.

>human limitations make it really difficult

If you accept the premise of Universal Paperclips, human limitations make it impossible. The only way to win is to avoid making the recursively self-improving paperclip maximizer in the first place.

That's an interesting point about game design. There actually are implementations of Tetris which have a win condition[0][1], and I think it's probably unrepresentative to pick games from an era where hardware limitations prevented modern mechanics like multiple endings and cutscenes.

As for modern games, I'm not convinced that failure states are a rare design element. It's true that games tend to include an auto-save feature if there is long-term state that needs to be preserved, but that still allows the player to "fail" and have to restart from the save point.

Incremental games like Universal Paperclips are a bit of a special case, because at some point the game plays itself and the end state (if it exists) is in principle reachable without any human interaction. I don't suppose there is any code in UP that specifically checks if you have been playing the game for billions of years though, to trigger an ending when all the mass of the universe has been turned into paperclips without actually unlocking all the "story" events.

Games like Getting Over It are also a special case, because their state is almost entirely defined by the position of the character in the game environment. They do have a clear ending, but failure is implicitly measured by how much forwards progress you lose when you fall. A mistake which takes you all the way back to the start is analogous to dying and starting a new game, though, so that still feels like the game has win and lose conditions.

> The only way to win is to avoid making the recursively self-improving paperclip maximizer in the first place.

In the game, there is a slight ambiguity about who/what the player's character is. You have control over the decisions before there is any self-improving AI, and also after the humans are all destroyed/enslaved. As such, it's not clear whether the player is winning as the AI, or watching in transfixed horror as the AI wins through a narrative that you are revealing. I suppose this philosophical question of viewpoint is the same as the one that Foddy considers near the end of Getting Over It: "Have you ever thought about who you are in this?", and with possibly the same answer.

[0] https://www.youtube.com/watch?v=N6mWpsu6zmQ

[1] https://www.youtube.com/watch?v=H_tmFUWu9bI

> Games like Getting Over It are also a special case, because their state is almost entirely defined by the position of the character in the game environment. They do have a clear ending, but failure is implicitly measured by how much forwards progress you lose when you fall. A mistake which takes you all the way back to the start is analogous to dying and starting a new game, though, so that still feels like the game has win and lose conditions.

As one of Getting Over It's appreciators, I offer a different perspective: Progress is not measured by the position of the character, but by the growth of the player. A mistake which drops you back to the beginning of the game is progress; you've learned something, and you have an opportunity to learn more while getting back up. The only way to fail is to give up. The true ending of the game isn't even when the credits roll, it's when you voluntarily ride the snake.

That's an excellent way of viewing the "state" of the "player-game" system, thank you.

My favourite example of this, and one of my favourite moments in "gaming" generally, was (SPOILERS:) a sequence of online puzzles, one per page, where solving the puzzle on one page would give you the address of the next page in the sequence.

The game was constructed (if I remember correctly) such that the first page actually had two possible answers: an obvious one, and a more subtle and complex one. By missing the subtle challenge, you ended up following a trail of relatively simple puzzles, which increased in difficulty over time, until they took you right back to the first page.

At first it seemed like a mistake, that after all these puzzles you were no closer to the end than when you started, but, as with Getting Over It, the "player-game" state had changed, because you were not the same person you were when you started the game. So, you knew to be more careful in looking for clues, and you knew how to solve those clues, and you could then find the single answer that was there from the start and took you straight to the success page.

Experiences like that really change how you view the world, and yourself, so I'm really glad I found it when I did.

There was an older games called Spec Ops: The Line that played with some degree of those problems. I truly loved it, although you do have to enjoy the "ludonarrative" aspects of it, and not everyone did.

Not if the preserved part was severely restricted in what it was.

That's not serious enough. Why would the AI waste atoms on people when those same atoms could be put to much better use making up paperclips?

AI isn't the threat, that's just the tool. Human nature is the threat, but we fear AI because it will make our horribleness so much more efficient.

I'd argue that AI itself isn't the initial threat, because, as you say, the first AI capable of causing harm will likely be one whose goals are set by humans, who already have a bunch of harmful desires. Indeed, we already are seeing machine-learning systems used by large corporations and governments for nefarious ends.

But long term, assuming we manage to get past that problem and limit the goals that humans choose to set for their AIs (such as the obviously common goal of "don't destroy humanity"), there is the danger that the AI develops both its own goals and the ability to understand its place in the world enough to work towards those goals. This is the problem of "alignment" in AI safety research.

https://en.wikipedia.org/wiki/AI_control_problem#Alignment

> only convert, say, one third of the universe into paperclips

But then exponential growth stops.

A big part of paperclip maximiser discourse is that the maximiser has no other values than expansion. To cease expanding is to die.

I’d like to submit Don’t Shoot the Puppy as another interesting branch in the evolutionary tree. It takes the clicker trope and… well, hard to say anything about what it does with it without spoiling the joke.

I came across it purely by accident with no clue it was even different. Figuring out what it was about almost broke me with laughter, but if you even have a clue going in it would probably fall flat. It’s the thought of thousands of clicker flash game players just running across this thing and trying to play it that does me in. People get _so_ angry.

This is the best game ever, I played it from 10pm to 6am.. non stop to reach the end.

It should be studied in addiction classes.

On every level there is some mystery and you have expectations, and somehow they are always blown away on the next level and the next..

At least it has a proper ending that lets you go. Unlike some other idle clickers.

> This is the best game ever [...] It should be studied in addiction classes.

I'm personally very fearful of what statements like showing up this say about gaming's future as a medium.

I find it interesting that there isn't an option on the table to just... not write addicting games. Similar to how we just can't not continue to improve on artificial intelligence, with limiters or without. Now that we know there is potential, it seems there is nothing preventing us from exercising our resourcefulness to pursue that potential.

Similar to the idea of UP, it seems we will continue optimizing all sorts of ideas to find ones that hold our attention the longest, until they expand to occupy all of our remaining free time. There is some kind of human instinct that encourages and validates this. I can't imagine it can realistically continue forever, though. The number of hours in a day is a hard physical constant.

For me (personally) this is one of the three all-time greats - it’s one of those rare occasions leaving you magically hooked and sucked in completely.

Only experienced this with Half-Life (ok, HL2 aswell) and Playdead’s Inside.

Try The Last of Us (both parts), and Submautica.

Tried getting into The Last Of Us several times and couldn’t.

Subnautica was great in the beginning until you realize the patterns.

I love Universal Paperclips. I've played through it a few times over the years.

In fact, I'm wearing a UP t-shirt right now. I won't say what the text is, since it's a spoiler for one of the best moments in the game.

Link?

The game has a Gift Shop (https://universal-paperclips.creator-spring.com/). I can only imagine the parent comment has this (https://universal-paperclips.creator-spring.com/listing/rele...) shirt specifically.

Yes, that's the shirt I'm wearing.

Also noting ... when people ask me what the shirt is about, it's really hard to summarize. I can say it's an idle game, which many people recognize, but explaining what the game is _about_ is challenging. Basically I have to start with "are you familiar with the concept of an AI singularity?" and go from there. Which is probably more than most people want to know.

You are a kitten in a catnip forest.

https://kittensgame.com/web/

With great replayability ;)

I played Universal Paperclips all the way through 100 times, and stopped. It was interesting seeing how I could optimize my way through it. It provided a good distraction from Long Covid last year.

It's almost a pity you stopped. Pathologically optimizing play in a game about pathological optimization would be so meta it hurt.

I suppose the next level up would be recruiting a thousand undergrads to optimize strategies for encouraging optimization of the optimizer, but then we've just reinvented psychological research with slightly more rigor.

I got the same kind of enjoyment from this game. I went through a kick a few years ago where I tried improving my best time, and my best run was within 10 minutes of the world record. I found it fun to learn how to play the game quickly, but demotivating to try to get a good market seed (I found this as the single most limiting factor of runs), and I stopped running it.

If you play the browser version, you can fiddle with the javascript while it's running.

That's interesting at first, but usually kills my interest in the game itself quickly.

Eg because I speed up the passage of time in idle clickers. That makes the game less annoying to play in the short run (thus better and more enticing), but strangely also much less addicting. The annoying wait is part of what slaves us, I guess?

Loved the philosophy and storyline in this game, along with the addictive game itself.

Took a few hours to play the game, spread over ~24 hours. Was able to reach the end, and score 30 Septendecillion [1]; I chose not to start over. At times it felt like there was no end to this game, and the only reason I kept playing was because of the level of engagement built into the game. The player has to figure out the game mechanics at each level, and every level introduces some novelty.

This is the first game in many years, in over a decade perhaps, that I was glued to for hours, and kept coming back to play. This left 2048, my favorite pastime until now, far behind in terms of engagement.

I did not read the blog post because it warned of spoilers, so at different stages it took me a while to optimize the numbers to make progress.

Loved the game, and will play it again when I have a few hours to spare. Hopefully will be better at it the next time.

[1]: Paperclips: 30,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000

> “You look at a painting,” Frank Lantz told the interviewer, “and you’re just absorbed.”

> We’re always looking. All day long we’re looking around, looking here, looking there, doing stuff. But then you stop and you look at a painting, and for a minute looking takes over. You’re no longer looking along with other things, you’re just—a hundred percent, your brain is all of sudden just a vision machine. You’re just looking at this thing. ...You fall into it, but then you also are able to lean back and think, “oh, that’s what looking is: that’s color, and shape, and form, and this is how my vision is structured... this is how looking works.”

Great lead in, unclear if the first breakout text block is a direct quote from Lantz but I love it.

One of those games you get sucked into the clicking. I don't know what it is, I have this desire/fetish to see things moving/tabulate things. Paperclips is like that seeing the numbers increment.

I was bad at this game though (not even close to that "beat paperclips in an hour" or whatever).

For anyone who really loves these types of games, "Leaf Blower Revolution" on steam is truly enjoyable. It is free to play (though I did buy the $5 supporter pack myself because the dev made such an enjoyable game), and is probably 30-50 hours of various levels of memes, relaxing, and math.

My currently addictive “parody clicker” is Egg Inc., going on for months with pathetically minimal graphics and numbers on orders of magnitude of orders of magnitude.

I quite enjoyed Egg Inc until a few years ago when the creator jacked the prices on everything and made it feel way more like a P2W past a certain point. Before that I had even spent some money on the game but after a, IIRC, ~50%+ increase and some new mechanics that felt very P2W I fell off. I still like the clicker genre but I prefer to play 1-time paid versions or ones where the only IAP is a 1-time purchase (remove ads or something like "pay once to play the rest of the game/proceed further").

A shame the article doesn't mention how truly poetic and philosophical the game becomes in the end.

Obligatory mention: A Dark Room [0]

[0]: https://adarkroom.doublespeakgames.com/

Why put a all red banner with this warning on the display, you don't even need JS to read the TEXT in a blog. "This site requires JavaScript to run correctly. Please turn on JavaScript or unblock scripts" Make it scroll out at least.

Reply on news.ycombinator.com