When algorithms surprise us



Machine learning algorithms are not like other computer programs. In the usual sort of programming, a human programmer tells the computer exactly what to do. In machine learning, the human programmer merely gives the algorithm the problem to be solved, and through trial-and-error the algorithm has to figure out how to solve it.

This often works really well – machine learning algorithms are widely used for facial recognition, language translation, financial modeling, image recognition, and ad delivery. If you’ve been online today, you’ve probably interacted with a machine learning algorithm.

But it doesn’t always work well. Sometimes the programmer will think the algorithm is doing really well, only to look closer and discover it’s solved an entirely different problem from the one the programmer intended. For example, I looked earlier at an image recognition algorithm that was supposed to recognize sheep but learned to recognize grass instead, and kept labeling empty green fields as containing sheep.


When machine learning algorithms solve problems in unexpected ways, programmers find them, okay yes, annoying sometimes, but often purely delightful.

So delightful, in fact, that in 2018 a group of researchers wrote a fascinating paper that collected dozens of anecdotes that “elicited surprise and wonder from the researchers studying them”. The paper is well worth reading, as are the original references, but here are several of my favorite examples.

Bending the rules to win

First, there’s a long tradition of using simulated creatures to study how different forms of locomotion might have evolved, or to come up with new ways for robots to walk.

Why walk when you can flop? In one example, a simulated robot was supposed to evolve to travel as quickly as possible. But rather than evolve legs, it simply assembled itself into a tall tower, then fell over. Some of these robots even learned to turn their falling motion into a somersault, adding extra distance.


[Image: Robot is simply a tower that falls over.]

Why jump when you can can-can? Another set of simulated robots were supposed to evolve into a form that could jump. But the programmer had originally defined jumping height as the height of the tallest block so – once again – the robots evolved to be very tall. The programmer tried to solve this by defining jumping height as the height of the block that was originally the *lowest*. In response, the robot developed a long skinny leg that it could kick high into the air in a sort of robot can-can. 


[Image: Tall robot flinging a leg into the air instead of jumping]

Hacking the Matrix for superpowers

Potential energy is not the only energy source these simulated robots learned to exploit. It turns out that, like in real life, if an energy source is available, something will evolve to use it.

Floating-point rounding errors as an energy source: In one simulation, robots learned that small rounding errors in the math that calculated forces meant that they got a tiny bit of extra energy with motion. They learned to twitch rapidly, generating lots of free energy that they could harness. The programmer noticed the problem when the robots started swimming extraordinarily fast.

Harvesting energy from crashing into the floor: Another simulation had some problems with its collision detection math that robots learned to use. If they managed to glitch themselves into the floor (they first learned to manipulate time to make this possible), the collision detection would realize they weren’t supposed to be in the floor and would shoot them upward. The robots learned to vibrate rapidly against the floor, colliding repeatedly with it to generate extra energy.


[Image: robot moving by vibrating into the floor]

Clap to fly: In another simulation, jumping bots learned to harness a different collision-detection bug that would propel them high into the air every time they crashed two of their own body parts together. Commercial flight would look a lot different if this worked in real life.

Discovering secret moves: Computer game-playing algorithms are really good at discovering the kind of Matrix glitches that humans usually learn to exploit for speed-running. An algorithm playing the old Atari game Q*bert discovered a previously-unknown bug where it could perform a very specific series of moves at the end of one level and instead of moving to the next level, all the platforms would begin blinking rapidly and the player would start accumulating huge numbers of points. 

A Doom-playing algorithm also figured out a special combination of movements that would stop enemies from firing fireballs – but it only works in the algorithm’s hallucinated dream-version of Doom. Delightfully, you can play the dream-version here


[Image: Q*bert player is accumulating a suspicious number of points, considering that it’s not doing much of anything]

Shooting the moon: In one of the more chilling examples, there was an algorithm that was supposed to figure out how to apply a minimum force to a plane landing on an aircraft carrier. Instead, it discovered that if it applied a *huge* force, it would overflow the program’s memory and would register instead as a very *small* force. The pilot would die but, hey, perfect score.

Destructive problem-solving

Something as apparently benign as a list-sorting algorithm could also solve problems in rather innocently sinister ways.

Well, it’s not unsorted: For example, there was an algorithm that was supposed to sort a list of numbers. Instead, it learned to delete the list, so that it was no longer technically unsorted.

Solving the Kobayashi Maru test: Another algorithm was supposed to minimize the difference between its own answers and the correct answers. It found where the answers were stored and deleted them, so it would get a perfect score.

How to win at tic-tac-toe: In another beautiful example, in 1997 some programmers built algorithms that could play tic-tac-toe remotely against each other on an infinitely large board. One programmer, rather than designing their algorithm’s strategy, let it evolve its own approach. Surprisingly, the algorithm suddenly began winning all its games. It turned out that the algorithm’s strategy was to place its move very, very far away, so that when its opponent’s computer tried to simulate the new greatly-expanded board, the huge gameboard would cause it to run out of memory and crash, forfeiting the game.

In conclusion

When machine learning solves problems, it can come up with solutions that range from clever to downright uncanny. 

Biological evolution works this way, too – as any biologist will tell you, living organisms find the strangest solutions to problems, and the strangest energy sources to exploit. Sometimes I think the surest sign that we’re not living in a computer simulation is that if we were, some microbe would have learned to exploit its flaws.

So as programmers we have to be very very careful that our algorithms are solving the problems that we meant for them to solve, not exploiting shortcuts. If there’s another, easier route toward solving a given problem, machine learning will likely find it. 

Fortunately for us, “kill all humans” is really really hard. If “bake an unbelievably delicious cake” also solves the problem and is easier than “kill all humans”, then machine learning will go with cake.

Mailing list plug

If you enter your email, there will be cake!

f they managed to glitch themselves into the floor (they first learned to manipulate time to make this possible), the collision detection would realize they weren’t supposed to be in the floor and would shoot them upward. The robots learned to vibrate rapidly against the floor, colliding repeatedly with it to generate extra energy.

Bethesda drive






hey everybody, welcome to another amazing installment of Weird Biology and WOWIE ZOWIE do I have an odd one for you today!

this bizarre creature is among the largest of its kind, but bears hardly any resemblance to the rest of the family. (we’re sure this gets mentioned a lot at its family holiday dinners.) it has a real mouthful of a name and the spirit of a cranky old man about to whack you in the shin with his walker.

give it up for…

I’ll just give this image a moment to sink in.

(it’s also called the small-headed softshell turtle, because scientists are a bunch of mean highschoolers.)

seriously, I don’t even really know where to START with this guy. unlike the humble regular earnest hardworking turtle, the Narrow-Headed Softshell Turtle does not have an armored shell (hence the name). instead, its shell is soft and leathery. like a pair of well-broken-in Timblerland boots, except that the boots will not bite you.

oh, he is absolutely going to bite you.

this soft pliable shell cuts down on the turtle’s weight by a huge amount, making them far more agile in the water and faster on land than a conventional everyman turtle (this should make you worried). the flattened shape of the shell also makes them more hydrodynamic, making them faster in the water than you can possibly imagine.

for a turtle, I mean.

this is an important advantage, because the Narrow-Headed Softshell Turtle spends most of its life in the water. they live on the bottoms of sandy rivers across a wide area of central and southern Asia, where they reach sizes best described as fucking huge. adults can reach up to 45 inches (shell length only) and 260 fucking pounds (whole damn turtle). 

their total body length can be over a meter. fuuuuuuuuck. a turtle that size needs a LOT of shoulder room, especially because the adults are a bunch of cranky ginormous chompmonsters. (can’t really blame them, I guess. I’d be irritable too, if my head was that small)

now imagine a cheesed-off 260-pound turtle swimming towards you at Mach Fuck.

Narrow-Headed Softshell Turtles are aggressive, and will attack anything they consider a threat (including humans, fishing boats, and probably also rocks). their primary attack is to just bite the fuck out of whatever is annoying them , but their secondary move is the one to watch out for.

when terminally pissed off, the turtle extends the full length of its surprisingly long neck and delivers a literal cannon headbutt. this attack has been documented as being powerful enough to damage fishing boats. imagine what it would do to your face. (nothing good. if you see this turtle winding up, run.)

the true face of terror.

when left to its own devices, the Narrow-Headed Softshell Turtle spends its time buried at the bottom of the river, waiting for its next meal to happen by. (which it can do almost indefinitely because softshell turtles can breathe underwater, holy shit.) once another animal smaller than itself passes overhead the turtle strikes, mortally wounding the prey with its nightmare bite (no joke, the first strike usually kills instantly. this is a creature capable of taking a chunk out of your leg). it’s a pretty solid gig, if you’re a lonely grumpmonster.

beats pumping gas all day, I guess.

in fact, the Narrow-Headed Softshell Turtle spends so much of its life underwater that we… don’t really know all that much about it. apart from the biting thing, I mean. the turtle has been very clear on that.

we’re not even entirely sure how long they live, though captive turtles have made it more than 70 grouchy, grouchy years. locals in India claim that in the wild individual river bastards can stick around for up to 140 years, which I am inclined to believe because these people fish for a living and they have to remember where the boat-sinking nightmare turtles live.

it’s only common sense.

despite its wide range, the Narrow-Headed Softshell Turtle is now considered Endangered. (note: this is not allowed. what would we replace them with? large cantankerous frogs? big passive-aggressive catfish? I DON’T THINK SO.)

this is primarily due to human hunting, as the turtles are consumed in huge number throughout Asia. (humans will eat anything.)

the government of India has now moved to protect the turtle, restricting trade and moving to conserve the species. we dearly hope this will be enough to save the grumpy frumpy river grandpa.

please stay with us forever, Narrow-Headed Softshell Turtle. we love your tiny tiny face and terrible attitude.

thanks for reading! you can find the rest of the Weird Biology series here.

if you enjoy my work, maybe buy me a coffee to support Weird Biology.


img1- Wikimedia Commons img2- conservationindia.org  img3-zoosrcool.wordpress.com  img4- Joel Sartore  img5- Turtle Survival Alliance   img6- Turtle Survival Alliance img7- The TeCake img8- Joel Sartore

Fantastic! When the apocalypse occurs and I become the morally ambiguous monarch we all deserve, I’m filling my moat with them.

Great post! @gallusrostromegalus , that move would put you firmly into the “evil” gig

You’re only going to get bit/headbutted into next teusday if you’re trying to get into my castle without permission so who’s really the asshole here?

…unless you’re concerned about the welfare of the turtles in which case I’d like to assure you it’s a large and well-circulated moat that mimics thier natural habitat as much as possible.


“independence” is an evil lie

note: this is written in a very angry tone directed at a hypothetical ‘you,’ who is essentially ‘a society run by abled ‘sane’ people.’ but being addressed like that, as if it’s directed at you the reader, might upset some people. if that’s you, here’s your heads up to consider skipping this one. ok, that said:

no one is independent.

we are all weak little cyborgs (thanks @shrimp-bird for giving me that concept!) whose main strategy is adapting the world around us and relying on each other, such that our bodies don’t have to adapt instead.

have you ever worn a jacket? used a road, a car? turned on a tap, a light?

congratulations, your inability to withstand cold, travel long distances in a reasonable timeframe under your own power, find safe drinking water yourself, and see in the dark were just fully accommodated. and you probably didn’t notice, at least in so many words.

and yet, oddly, move the slider bars on those abilities a little and suddenly it becomes exceptional, unacceptable.

for example, add a wheelchair to supplement your accommodation of the very human inability to get as far as we want to as fast as we want to, and oh gosh no, special needs! medical! must provide proof! we will still deny you! but you are very inspiring!

no matter what anyone does or doesn’t do, im still disabled. but with the right accommodations, like those everyone else takes for granted every single day, i could at least be on a level fucking playing field.

but no, since everyone else’s needs are invisibly accommodated and mine are embarrassing and exceptional, you physically lock me out.

you place literal barriers to my entry because you decided that the needs of people like me were simply not worth considering, when you built stairs but no lift or ramp, to pick the absolute most basic example, the most pathetically low standard people still relentlessly fail to meet: let me in the fucking door.

then once you do that, install a goddamn hearing loop and provide captions. make sure your space is well lit but don’t choose fucking fluorescents what is wrong with you those things hurt. make desks and counters and sinks usable from a seated height. don’t make me constantly have to ask other shoppers to get stuff from high shelves for me at the fucking supermarket, it’s humiliating.

these don’t all apply to me, they’re just some obvious examples off the top of my head.

and none of this even touches on the pretty much total lack of accommodation for developmentally disabled people and for crazy people (a term i reclaim given that ‘mental illness’ doesn’t work for me for various reasons), or a whole variety of other ignored disabilities and chronic illnesses, and im not going to right now because frankly im too tired.

and if you want to ask for proof before doing any of this, i would suggest you consider getting a great deal of peanut butter stuck to the roof of your mouth instead, because we don’t owe you shit for maybe stopping for just one second to consider that other people exist and you are ruining our lives.

telling me im worthless because i can’t be ‘independent’ by your arbitrary definitions, when you systematically remove the tools i would need to be able to live the kind of life you take for granted and actively make it impossible to get what i need. ok, sure, that makes sense. fuck me, right?

we are all interdependent. never forget it.





Antiamericanism is good. smug Europeans who denigrate American in order to presume their own nations are not equally complicit in colonialism and imperialist violence are bad.

???? our own nations aren’t complicit in colonialism n imperialist violence, and that’s WHY we look down on america?????


I mean… I mean… America??? as it is now??? is a RESULT of european colonialism??? like its a literal colony of Europe??? with indigenous people


being a minority??? what the fuck are you talking about @ass-th-etic

I also have to add that targeting individuals for hostility–rather than the actual harmful systems–for a cheap shot ego boost is just obnoxious. Also unfortunately par for the course with a certain type of smug person.













This is demonstrating why you absolutely do not pour water on a grease fire.

holy shit

Okaaaay. If any of you actually have a grease fire in the kitchen put the lid on the pan. It will suffocate the flames. Don’t pour water on it, and don’t freak out. Cook safely!

Or throw flour on it to smother it.

/quick safety announcement



Throwing flour into a fire can cause it to combust and make the fire worse because FLOUR/SUGAR IS FLAMMABLE. One cup of flour into a grease fire can have the explosive force of dynamite.

The reason you use baking soda is that it releases carbon dioxide when heated, and CO2 is a fire suppressant.


can we talk about how this is from a tv-show called “do not try this at home” where they tested all sort of stuff you’re not supposed to do, but they only got four episodes because after this experiment they burned the house they were filming in to the ground.

Y’all, remember that Mythbusters episode where they dropped a can of chicken stock in a pot of frying oil just for shits’n’giggles? If you haven’t, I strongly suggest looking up the clip. Grease fires are something you absolutely do not fuck with.

they get abandoned houses now to fuck up actually! they turned a hot water heater and the house plumbing into a coffee machine one time lol.

same guys who made that juice flood episode.