Hacker News new | past | comments | ask | show | jobs | submit login
Paperclips (decisionproblem.com)
257 points by yangyang on Oct 14, 2017 | hide | past | favorite | 112 comments



It's a really great game that is best approached without too many spoilers. Know that the game becomes more complex and engaging as it goes on. There is a definite end and takes about 5 hours to complete if played well.

[spoiler]

It's the "Paperclip maximizer" thought experiment put into game form and you play as the AI. The game is divided into roughly three stages. The first you are the AI for some company and are tasked with producing a profit and using the profit to game trust and eventually conquer humanity. The second stage is post-human Earth stage where you convert the planet to paperclips. The final stage involves sending probes to explore space and do battle with rouge AIs and convert the universe into paperclips. There is end where you can select to either defeat the AIs and dismantle yourself into paperclips or you can listen to rouge AIs and start over in an alternative universe with some small modifier edited.

It does a really good job of exposing you to uncommonly large numbers and does a good job of presenting you with massive scale. There is a lot of joy seeing the game become increasingly complex.

[/spoiler]


It's probably being used as training data for the coming AI revolution. Your gaming is going to ruin us all!

It's a weird Ender's Game type of deal.

On a more serious note, it might actually make a neat way to get training data.


[spoiler] I chose the reject option, what does "accept" change? [/spoiler]


"Accept" allows you to restart in another universe with a small bonus, either a parallel one (with +10% bonus to demand) or a simulated one (+10% speed bonus to creativity generation).


I am also interested in this.


Actually, the gamestate is saved in the localstorage. So you could backup it in some text file before testing one option and restoring it to test the other option.

However, some people might call that cheating.


Note: if you close a tab and reopen later, it continues where you left off. So no need to lose sleep, it has localStorage :-)


Ha! I discovered this when I was spamming the mouse button on the QC, and accidentally hit the 'back' button (on the mouse). I was very afraid that all my hard work was gone. But a forward click brought it back.


Well I just stayed up all night playing this. RIP me.

I enjoyed it a lot, though the gameplay during the final stage was less balanced and entertaining. Fantastic ending though!


How many $ did you make in the end? (Assuming the dollar still exists when you finish the game)


Can't talk about this without spoilers. The ending is very good.


Towards the middle of the game, I bribed the project managers with a billion dollars to buy me an additional memory module - if that gives you an idea.


I've beaten this. It's a silly, insightful, weird, and repetitive cautionary tale. Without spoiling too much, the ending leaves you with a terrible sense of emptiness, but is still quite satisfying.


The author should publish an API for this game, and allow AIs to play it. Then extend to multiple competing companies. And then extend to a simulation of a full economy.


The browser console and the DOM make for a decent interface to the game if you want to try your hand at "automating" (cheating) parts of the game..

For example: I eventually wrote and ran an interval function in the console to automatically click the Quantum Computing button only when all my available photonic chips were producing a collective net-positive result.


I originally thought this was the purpose of the game, and asked the person who linked it to me if I was supposed to write code for it.


This is the third time I've seen this posted this week. I don't know if it isn't getting any responses because the game looks deceptively basic at first, or because the people who played it long enough to know better haven't come back.


I honestly did not understand what this website is about.


It's about paperclips.

Why make paperclips? Because it's your utility function.


Maybe this makes it clearer: http://bloodrizer.ru/games/kittens/


FOOM


Warning: If you have work to do, do not start this game.


This is amazing! I've wasted two whole hours on this. Great way to spend a Sunday :) Spoilers aside, I'd love to know how the creator tested this!


10 hours later, I've finished it. Probably the best game I've played in a while.


  . Universal Paperclips
  . a game by Frank Lantz
  . combat programming by Bennett Foddy
  . 'Riversong' by Tonto's Expanding Headband 
     used by kind permission of Malcolm Cecil

  > © 2017 Everybody House Games


Using typematic (keeping the "Enter" key pressed on the "Make Paperclip" button) makes the beginning of the game far less tedious in my opinion: xset r rate 200 255


You can just select the button with tab and hold enter.


Yes, this is what I meant: the command I gave is the way you optimize the frequency at which the keypress is repeated (on Linux systems with Xorg).


you can also just use the javascript console.

chipclick(n): make n paperclips.


Once I started this, I didn't stop until I'd finished the game. It's a very good incremental game, especially if you're into AI lore. I loved it.


This is a landmark achievement in video games.


I lost 5 hours to this on Friday and made 5 octillion paperclips.


Ha! Lightweight. I've made 3 nonillion paperclips and explored %1 of universe. :)


Ha! Newbie. You still have 99% to go :)


At 1% exploration you should already have exponential exploration growth (reasons are spoilers), and would be only a few seconds away from 100%.


So it's an idle game in JavaScript (or equivalent). Yet my iPhone is pretty hot after 30mn playing.

Is it mining or something ?


Idle games are usually not optimized for performance. It's running bigint style operations in a tight loop. Mobile phones are not what you should be playing these games on.


Well, I had no problem in playing Swarm Simulator [1] in my phone other than that incremental games intrinsically need an enough screen to be aware of everything. In the case of the Universal Paperclips, its simulation step is too slow (e.g. its handling of projects is suboptimal) that its internal clock will drift over the time. That's why it reports far less time in the notification than the actual wall clock.

[1] http://swarmsim.github.io/


Was able to play this just fine on iPhone 7 Plus


Ah. JS crashed on my mobile at about 30M clips...


Why do I make more clips than I sell and still my stock shrinks? :)


Just a theory... Are you playing on mobile? Is it possible the mobile browsers ticker has a max speed limiting the production, but the avg sold measurement is measured using an actual sold per second avg.

On my mobile the production seems to cap out at around 10/sec even when it reports far higher than that.


I noticed this as well. It's either a bug, or there's a rogue employee secretly siphoning off paperclips and selling them on the black market.


Because it's only an average. The market is random.


I consistently make ~3x as much clips as I sell "on average" and it still always trends down (e.g. if I build up a few k of stock, it trends downwards to 0 and stays there). If that's randomness, the data is presented quite badly.


This is definitely not merely what it appears to be.


Depending on your background, it may appear to be exactly what it is. It did for me...


True. I was not familiar with the premise when I played it.


(Possible spoilers)

Yeah, I came in well primed - having played "A Dark Room", familiar with the "paperclip maximizer" thought experiment, and familiar with various iterations of the idea of keeping an AI "in a box". Playing the game was probably a very different experience than for someone with none of that familiarity.


Absolutely, I can not leave it!


This is an excellent game in its genre! It has some rough spots; if you do the wrong thing at key points, it can take much, much, much longer to complete than it should. It does have an ending, which is great as well. All around, wonderful game.


Isn't it just a clone of classic Cookie Clicker?

http://orteil.dashnet.org/cookieclicker/


In roughly the same sense that Portal is just a clone of Doom.

There are similarities in the mechanism; in the very early game in particular they are nearly identical. But they tell different stories, of different styles and levels of sophistication. They deal with different ideas. And this one goes from start to a satisfying end in a handful of hours.


If you threw in a bit of a science fiction storyline. The game gets more complex, and the goals change several times.


SPOILER:

After "full autonomy" is attained, should it still be producing paperclips? Because mine is no longer producing paperclips, and I can't see any way to proceed other than producing more paperclips.

EDIT: I understand it now. It is working as intended.

EDIT2: Nitpick: in the second (?) stage, there's a stat for "MWs/sec" power consumption... this should just be "MW" - 1 Watt is 1 Joule/second so I don't see what 1 Joule/second/second of power consumption would mean.


It’s megawatt-seconds per second. (?)


The seconds cancel, so it's just an excessively complicated way of writing MW.


Not familiar with this case, but W/s could be describing an acceleration, i.e an increasing rate of energy transfer


It's starting to remove some of my braincells atm.


I don't really understand what makes subsequent clip factories more expensive. In the first stage, yeah I understand why production units were getting more expensive. Some puny human was shortchanging me, that's why. That's why I enslaved them. But what in-game explanation is there for making the factories more and more expensive later?


Supply curves are a real phenomenon, even for us paperclip maximizers. For example: you built the first factory in the most convenient spot, out of the best possible materials. You have to ship materials farther, build more transportation and energy infrastructure, build more protection against environmental hazards, etc to build the billionth factory. And you have run out of the best materials, so you have to use more expensive substitutes.


Maybe you need to buy more land or pay more for the resources you are making scarce.


entropy


I'm not sure if this gets revealed if you play long enough (I obviously did not).

But it is curious that there's some sort of battle simulation running (and being rendered to an invisible canvas).

  (function reveal() { document.getElementById('battleCanvasDiv').style = ''; requestAnimationFrame(reveal);})()


[spoiler]

Eventually you end up in a self-replication phase. Replicas are sometimes imperfect, so you have to defend yourself from rogue-rogue AIs.

[/spoiler]


[spoiler]

I interpreted that as fighting against other civilizations that are trying to stop your progress and protect the universe.


Civilizations resorting to killer AIs in desperation to stop your progress.


So I've obtained full autonomy, I've got clip factories and tons of drones, and they are mostly "thinking", bringing me gifts. the problem is, there are no more projects... Did I find a bug and get stuck, or am I supposed to be more patient? I don't see a way to level up any more..


I attained full autonomy before getting to 100m+ clips and am now stuck not able to create any factories. The price of success ;-)


I got stuck too. I spent too much trust on CPU instead of memory. That limited my total compute to 30K, I needed 50K for the upgrade and the quantum boost wore off before I could reach 50K. I ended up having to click the quantum button using javascript, because I couldn't click fast enough. That got me unstuck.


I apparently played that stage too well and completed it before I had made enough paper clips (which becomes the new currency then) to get anything started at all in the next stage.


[spoiler] It unlocks at 10, 50 and 100 factories or something like that, same with drones, 5k, 50k etc. [/spoiler]


Did you even start exploring the universe?


Yeah, eventually I figured it out. The trick was to start an exponential growth of probes by focussing on self-replication... then things started unlocking and it went on from there...


This is so addictive. Been playing for an hour now. I wish there was a pause button.


You can close your browser and it will continue at roughly the current state next time you re-visit. (I knew this but it didn't really help :-O)


It did make me think that my browser should allow for me to pause a tab. Just pause the state of the JS


Did anyone research the "outsourcing paperclip manufacturing to mamal brains by meme" upgrade?


Did anyone else have problems after releasing the hypnodrones? I lost most of the interface and stopped producing paperclips. I don't have factories yet, and only have produced 700 million paperclips


Full autonomy attained in 2 hours 55 minutes 58 seconds

The strategy seems to heavily rely on investment, that seems not to my liking.

I'd want to see purchase item to fully automate all sub-games to make the late game less tedious.

One thing I think might be helpful: Make sure there is unsold inventory once you want to deposit to the stock market, to avoid stuck money generation and crushed stock market killing you...


Just to make sure it's clear: Full autonomy attained is not where the game ends, it's just where the first stage ends.


The above seems like a pretty usual time for Full Autonomy; I've only beaten the full game once so far, and it took 7 hours 58 minutes and 10 seconds to beat completely.

I've heard some people can do the whole thing in four hours; I'm far along in my second playthrough right now, we'll see!


I actually finished my first playthrough in 3 hours and 45 minutes. Think I just got lucky with my strategy, but I guess one's focus should be one always upgrading paperclip production until you start having excess materials, and try to keep a balance so input = output as much as possible. But of course, all of this goes out the window on the last stage of the game.


Nice! Yeah, I always found the last stage to be very counterintutive; it's unclear how all of the various things interact and what they should do.


Silly me... I should try it again.


Just reopen the tab and it will pick up where you left it


That seems pretty realistic. The easiest path to taking over the world could well involve accruing vast amounts of money.


I think the game subtly guides the players towards a investment-heavy strategy.

Also, there is a built-in safety net that saves you from crashes... Won't spoil it completely but you can most easily achieve it right after you start the game from blank slate.


[SpoilerQuestion] I didn't invest in enough processors so the 25000 Creativity took forever. Then I discovered that cheating is extremely easy (js console). But now I don't get any new Projects. Did I break it? Can it be recovered? [/SpoilerQuestion] I wished I didn't cheat so hard: ... Clips per Second: 18.76 duodecillion


How is this different from Cookie clicker?


Cookie clicker is endless. This has a “story arc” and a definitive end point when you win the game.


Do you really win though?


I am kinda loving this game, I only with it had some audible "ping!" when stuff becomes available.


Loved this game -- super simple and addictive. Just the right length, very thought provoking.


Cute skinner box game. Includes just enough variety in mechanics and twists and turns to keep me interested in what happens next despite the insulting "click this button 10,000 times and wait half an hour" mechanics.


I just spent ten hours there.

It was pretty fun!


Just finished it myself too.



Fuck, I just played this for 4 hours.


I took too much processor and not enough memory, it would take age to get to 70k...

How do I reset and start over?


On the other hand, I took too much memory 80, to 15 processors. After escaping control, I didn't have enough paperclips to build a factory, then I seem to be stuck.

I figure I have to get enough creativity to find a solution, but my low processor count makes it really slow.


There went my Sunday morning


Full autonomy attained in 4 hours 56 minutes 39 seconds|


0.000000000000% of universe explored

Does that number ever increase ?


Yes.


The "Increase probe trust" button doesn't get enabled anymore for me, even though I have enough yomi.


You will need to increase max trust. That will happen later in the game.


again use our computers for mining bitcoine :)


crashed and reset :/


https://news.ycombinator.com/from?site=decisionproblem.com

8 submissions in 5 days, one of which already got to 64 points and a bunch of discussion:

https://news.ycombinator.com/item?id=15437697


Paperclips, dont we all just want moore of them?


The genius of this comment is hidden by the fact that you didn't capitalize "Moore".




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: