When Algorithms Surprise Us

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!

Tags

*thinking* math

ACADEMIC PHRASE BANK MASTERPOST: CONNECTING WORDS FOR ESSAY WRITING

Addition

To begin with,

In the first place,

Firstly,

The first reason

Additionally

Furthermore,

Another reason why

Secondly, Thirdly,

Next,

Pursuing this further,

Also

Lastly, Finally

In the same way,

Comparison

Similarly,

In the same way,

Likewise,

As with,

Equally,

Contrasting

On the same contrary,

However,

Nevertheless,

On the other hand,

Even so

Alternatively

At the same time

Otherwise

Instead

Conversely

Result

Hence

Therefore

Accordingly

Consequently

Thus

As a result

In consequence

For this reason

For this purpose

Time

Meanwhile

Presently

At last

Finally

Immediately

Thereafter

At that time

Eventually

Currently

Subsequently

In the meantime

Importance

Importantly

Especially

Above all

With attention to

Example

For example

For instance

That is

Such as

As revealed by

Illustrated by

Specifically

In particular

For one thing

This can be seen by

An instance of this

Literary

Clarifies

Conveys

Depicts

Demonstrates

Determines

Displays

Emphasizes

Establishes

Explains

Exemplifies

Highlights

Illustrates

Indicates

Potrays

Represents

Shows

Signifies

Suggests

Beginnings/Causes/Effects

Affects

Generates

Ignites

Impacts

Imposes

Influences

Initiates

Introduces

Involves

Launches

Leads to

Presents

Promotes

Prompts

Results in

Summary

In conclusion,

To sum it all up,

To summarize,

In the final analysis

You can see why …

Finally,

To wrap it all up,

Therefore,

In summary,

In short,

In brief,

Tags

words essays

62K notes View Post

swirlspill-study

6 years ago

this is a masterpost dedicated to those who want to organize themselves but are too lazy to start aka me everyday

01. organizing your schoolwork

there are so many ways to organize your papers nowadays, so it’s pretty important to find what works best for you

I personally use binders and sheet protectors for my handouts that I get at school

pros: binders keep your papers secure, you can use dividers for further organization

cons: that loud noise from opening the binder rings, can be heavy, the binder rings can mess up your papers, the plastic pockets on the inside of the binder often tears

as mentioned, to combat the tearing of papers, I use sheet protectors

if you’re not a fan of binders, a nice alternative is expandable file folders

see this video by @studyign/sareena for some more info!

pros: not as heavy as a binder, there is some organization already with the extra pockets

cons: depending on the brand, might be flimsy, the plastic pockets can tear

I currently use one of these to store all of my homework, and so far it’s worked out for me quite nicely! I would definitely recommend them

if you don’t get that many papers in class but still want to hold onto them, folders are always a nice choice

pros: are lightweight, capable of holding a lot of papers

cons: can tear, papers can fall out (unless you have the kind w/ brads)

I use both folders and binders bc sometimes I’m too lazy to open the rings of the binder oops

also, if you want to store more, it can help if you buy the folders that contain brads

psa: if you don’t keep binder clips or paper clips with you, then you should bc they’re extremely useful and will make your life easier when you’re sorting out papers

for more tips, see this masterpost by @tbhstudying/seo!

02. organizing your supplies

some of you might have a lot of stationary and pens lying around

for those that don’t, it’s totally fine and you might have it better off bc you don’t have to keep track of that many things ahahaha

it’s always nice to have a pen case or pouch to hold your supplies, it can really help you keep your supplies together

if you’re willing to make the investment, try the kipling 100 pens pencil case

here is a cheaper alternative

this website is also really nice, their products are super cute and they have a large variety of pen cases to choose from

when you’re organizing your supplies, it’s important to know what you need the most on a daily basis - it’s not helping you if you keep a complete set of staedtler triplus fineliners in your pen case when in reality you only use two colors every day

ask yourself: what do I need? will it be okay if I leave it behind?

if you choose to leave some supplies at home, be sure that you have a place to leave them hehe

03. organizing your desk space

okay so if you’re like me, you probably have a lot of papers and things that stay on your desk, such as stationary, extra paper, textbooks, etc.

to organize some of my pens/pencils, I recommend using a jar or cup of some sort to keep on your desk so that it’s within reach

for those extra packs of lined/graphed paper you might have, you can store it on your desk in a magazine file or in your drawer

going back to the magazine file: 11/10 would recommend

for me, I have a small desk, so I can’t afford to have everything being piled on my desk - the magazine files are really nice for organizing

be sure that when you’re working, you only have the things you need on your desk! anything that isn’t needed should be stored as neatly as possible

you can find different desk organizers and containers in places like target and amazon!

I hope this was helpful to you guys! I personally still struggle with organization, but over the course of my 2015-16 school year, I’m starting to find ways to keep my things and my life together. if you have any questions, feel free to drop an ask in my inbox ^-^

- soko ◟(๑•͈ᴗ•͈)◞

Tags

resources organizing

7K notes View Post

swirlspill-study

6 years ago

Medical School Resources! (and other human biology,physiology,biochemistry-related resources)

Hi Everyone!

Update: I am now officially done with my second year! I know i’ve been MIA on here for a while now - but that’s only because I was drowning in textbooks and assignments! I will be writing a whole other post on what my second year in medical school was like - so watch out for that :)

I, for one, can not just rely on one method of learning. Meaning, I’ll jump from videos, to textbooks, to flashcards. In this post I’m going to list some of my holy grail youtube channels that have helped saved me.

1) Handwritten Tutorials

https://www.youtube.com/user/harpinmartin

Every video in this channel is short, but not so much that you feel like you’re missing out on information. Definitely one to save as a favourite!

2) Armando Hasudungan

https://www.youtube.com/user/armandohasudungan

The best thing about this channel is the fact that there are over 300 videos, covering a wide range of core topics in endocrinology, neurology, physiology and pharmacology. Another pro is the presentation of topics (otherwise considered snooze-worthy) in an artistic manner!

3) Speed Pharmacology

https://www.youtube.com/channel/UC-i2EBYXH6-GAglvuDIaufQ

Raise your hand if you’ve ever fallen asleep trying to read about the mechanism of action of opioids, their side effects and contraindications. I know I have. Fret not, for this youtube channel will introduce you to a world where pharmacology is actually interesting.

4) Wendy Riggs

https://www.youtube.com/user/wendogg1

Wendy Riggs is a very down-to-earth professor in Northern California, and she covers a wide range of topics in Anatomy, Physiology and General Biology.

5) Anatomy Zone

https://www.youtube.com/user/TheAnatomyZone

A better way to learn anatomy is to supplement your textbook information with videos from this channel. The explanations and visuals provided are absolute gold.

I hope you all find these channels as helpful as I did!

Tags

reference

7K notes View Post

swirlspill-study

4 years ago

general language learning resources

dictionaries:

wordreference - has spanish, french, italian, portuguese, catalan, german, swedish, dutch, russian, polish, romanian, czech, greek, turkish, chinese, japanese, korean, & arabic

reverso translation - has arabic, chinese, dutch, french, german, hebrew, italian, japanese, polish, portuguese, romanian, russian, spanish & turkish

bab.la - has spanish, arabic, chinese, czech, danish, dutch, finnish, french, german, greek, hindi, hungarian, indonesian, italian, japanese, korean, norwegian, polish, portuguese, romanian, russian, swedish, swahili, thai, turkish, vietnamese, & esperanto

digital dictionaries of south asia - has dictionaries for assamese, baluchi, bengali, divehi, hindi, kashmiri, khowar, lushai, malayalam, marathi, nepali, oriya, pali, panjabi, pashto, persian, prakrit, rajasthani, sanskrit, sindhi, sinhala, tamil, telugu & urdu

resources for learning words in context:

reverso context - has arabic, chinese (in beta), dutch, french, german, hebrew, italian, japanese, polish, portuguese, romanian, russian, spanish & turkish (in beta)

linguee - has german, spanish, portuguese, french, italian, russian, japanese, chinese, polish, dutch, swedish, danish, finnish, greek, czech, romanian, hungarian, slovak, bulgarian, slovene, lithuanian, latvian, maltese, & estonian

for learning different writing systems

omniglot - an encyclopedia with literally any language you could think of including ancient languages

scripts - an app for learning other writing systems with a limited amount for free (you can do 5 minutes a day for free) - has the ASL alphabet, Russian cyrillic, devanagari, Japanese kana, Chinese hanzi, & Korean hangul

Wikipedia is also helpful for learning different writing systems honestly!

pronunciation

forvo - a pronunciation dictionary with MANY languages (literally an underrated resource i use it all the time)

a really helpful video by luca lampariello with tips on how to get better pronunciation in any language

ipachart.com - an interactive chart with almost every sound!! literally such an amazing resource for learning the IPA (however does not include tones)

another interactive IPA chart (this one does have tones)

language tutoring

italki - there’s many websites for language tutoring but i think italki has the most languages (i have a referral link & if you use it we can both get $10 toward tutoring lol) - they say they support 130 languages!

there’s also preply and verbling which are also good but there aren’t as many options for languages - preply has 27 and verbling has 43

(obviously these are not free but if you have the money i think tutoring is a great way to learn a language!)

getting corrections/input from native speakers

hellotalk - an app for language exchanges with native speakers & they also have functions where you can put up a piece of writing and ask for corrections - honestly this app is great

tandem - language exchange app but unlike hellotalk you can choose multiple languages (although i think hellotalk is a little bit better)

LangCorrect - supports 170 languages!

HiNative - supports 113 languages!

Lang-8 - supports 90 languages!

verb conjugation

verbix - supports a ton of languages

Reverso conjugation - only has english, french, spanish, german, italian, portuguese, hebrew russian, arabic, & japanese

apps

duolingo - obviously everybody knows about duolingo but i’m still going to put it here - i will say i think duolingo is a lot more useful for languages that use the latin alphabet than languages with another writing system however they do have a lot of languages and add more all the time - currently they have 19 languages but you can see what languages they’re going to add on the incubator

memrise - great for vocab! personally i prefer the app to the desktop website

drops - you can only do 5 minutes a day for free but i still recommend it because it’s fun and has 42 languages!

LingoDeer - specifically geared towards asian languages - includes korean, japanese, chinese & vietnamese (as well as spanish, french, german, portuguese and russian), however only a limited amount is available for free

busuu - has arabic, chinese, french, german, italian, japanese, polish, portuguese, spanish, russian, spanish, & turkish,

Mondly - has 33 languages including spanish, french, german, italian, russian, japanese, korean, chinese, turkish, arabic, persian, hebrew, portuguese (both brazilian & european), catalan, latin, dutch, swedish, norwegian, danish, finnish, latvian, lithuanian, greek, romanian, afrikaans, croatian, polish, bulgarian, czech, slovak, hungarian, ukrainian, vietnamese, hindi, bengali, urdu, indonesian, tagalog & thai

misc

a video by the polyglot Lýdia Machová about how different polyglots learn languages - this video is great especially if you don’t know where to start in terms of self study

LangFocus - a youtube channel of this guy who talks about different languages which is always a good place to start to understand how a specific language works also his videos are fun

Polyglot: How I Learn Languages by Kató Lomb - this book is great and available online completely for free!

Fluent Forever by Gabriel Wyner (on pdfdrive) - another great book about language learning

Anki - a flashcard app (free on desktop for any system & free on android mobile - not free on ios mobile) that specifically uses spaced repetition to help you learn vocabulary, it’s got a slightly ugly design but it’s beloved by many language learners & is honestly so helpful

YouTube - literally utilize youtube it is so good.

Easy Languages - a youtube channel with several languages (basically they go around asking people on the street stuff so the language in the videos is really natural) & they also have breakaway channels for german, french, spanish, polish, italian, greek, turkish, russian, catalan & english

there’s also the LanguagePod101 youtube channels (e.g. FrenchPod101, JapanesePod101, HebrewPod101) which are super great for listening practice & language lessons as well as learning writing systems!

Tags

reference resources languages

5K notes View Post

swirlspill-study

7 years ago