You Look Like a Thing and I Love You Read online

    k k k k  k

  kk  k kkkok

  k kkkk

  k

  kk

  kk k kk

  keokk  k

  k

  k

  Now, it’s not the world’s greatest knock-knock joke. But with this as a starting point, the AI can move on to a second batch of knock-knock jokes, then another. Each time, it adjusts its joke formula to improve its guesses.

  After a few more rounds of guessing and self-adjusting, it has learned more rules. It has learned to employ the occasional question mark at the end of a line. It is beginning to learn to use vowels (o in particular). It’s even making an attempt at using apostrophes.

  noo,

  Lnoc noo

  Kor?

  hnos h nc

  pt’b oa to’

  asutWtnl

  toy nooc

  doc kKe

  w ce

  e

  How well did its rules about generating knock-knock jokes match reality? It still seems to be missing something.

  If it wants to get closer to generating a passable knock-knock joke, it’s going to have to figure out some rules about which order the letters come in. Again, it starts by guessing. The guess that o is always followed by q? Not so great, it turns out. But then it guesses that o is often followed by ck. Gold. It has made some progress. Behold its idea of the perfect joke:

  Whock

  Whock

  Whock

  Whock

  Whock Whock Whock

  Whock Whock

  Whock

  Whock

  It’s not quite a knock-knock joke—it sounds more like some kind of chicken. The AI’s going to need to figure out some more rules.

  It looks at the dataset again. It tries new ways to use “ock,” looking for new combinations that match its example knock-knock jokes better.

  nooc hock hork

  aKoo kWhoc

  Khock

  hors Whnes Whee

  noo oooc

  Koo?

  loo

  noc no’c Wno h orea

  ao no oo k KeockWnork Koos

  Thoe

  nock K ock

  Koc

  hoo

  KKock Knock Thock Whonk

  All this progress happens in just a few minutes. By the time I return with my coffee, the AI has already discovered that starting with “Knock Knock / Who’s There?” fits the existing knock-knock jokes really, really well. It decides to apply this rule always.

  But the rest of the formula takes quite a while to figure out, and the AI frequently devolves into playing a game of telephone with itself—or acting out a robot version of the “Who’s on First?” comedy routine:

  Knock Knock

  Who’s There?

  Iane

  Aatar who?

  Aaane who?

  Aan who?

  Anac who?

  Iobe who?

  Irata who?

  Soon it figures out the rest of the formula, but no actual, you know, words:

  Knock Knock

  Who’s There?

  Bool

  Hane who?

  Scheres are then the there

  It also sometimes gets a bit carried away with the length of the punchline.

  Knock Knock

  Who’s There?

  Cheed

  I?len who?

  Ita gee the lasd emotory oud ane weonges oak anet ane tee soore theis onithit the where your thasg tate tine if thele ter the theung!

  At some point in its training, it learns the joke about the Cow with No Lips, which becomes its Favorite Thing Ever. If you’re not familiar, the joke about the Cow with No Lips goes something like:

  Knock, knock.

  Who’s there?

  A cow with no lips.

  A cow with no lips who?

  A cow with no lips says ooo ooo ooooooo.

  According to the AI, this is comedy gold. So the AI spends a while thinking that the punchline to Every Single Joke is “ooo ooo ooooooo”…

  Knock Knock

  Who’s There?

  Alec

  Anic who?

  Wow co poo the tho tho door to the soout oo ooo oooo ooo oo oo oo oo ooo oo oo oo ooo ooo oooo ooo ooo oo oo ooo ooo ooo ooo ooo oo oo ooo ooo ooo ooo oo oo oo ooo ooo ooo ooo…

  It has discovered a rule, all right. Eventually it (reluctantly, I assume) tones down how often it uses the Cow with No Lips rule to make punchlines. But it still has problems that are remarkably similar to two people telling knock-knock jokes over a very bad phone connection, maybe forgetting who was actually telling the joke:

  Knock Knock

  Who’s There?

  Alther

  Water who?

  Water who who?

  Who who?

  Arton and out on the moor?

  At last it figures out the structure of a knock-knock joke and proceeds to make jokes that, while technically correct, don’t make all that much sense. Many of them are partially plagiarized from jokes in the training dataset.

  Knock Knock

  Who’s There?

  Robin

  Robin who?

  Robin you! Git me and I’m leaving

  Knock Knock

  Who’s There?

  Bet

  Beef who?

  Beef ano with no lips aslly.

  Ireland

  Ireland who?

  Ireland you money, butt.

  And then. It produced. An actual joke. That it had composed entirely on its own, without plagiarizing from the dataset, and that was not only intelligible but also actually… funny?

  Knock Knock

  Who’s There?

  Alec

  Alec who?

  Alec-Knock Knock jokes.

  Did the AI suddenly begin to understand knock-knock jokes and English-language puns? Almost definitely not, given the very small size of the dataset. But the freedom that the AI had—free rein over the entire set of possible characters—allowed it to try new combinations of sounds, one of which ended up actually working. So more of a victory for the infinite monkey theory* than a proof of concept for the next AI-only comedy club.

  The beauty of letting AI make its own rules is that a single approach—here’s the data; you try to figure out how to copy it—works on a lot of different problems. If I had given the joke-telling algorithm another dataset instead of knock-knock jokes, it would have learned to copy that dataset instead.

  It could create new species of birds:

  Yucatan Jungle-Duck

  Boat-billed Sunbird

  Western Prong-billed Flowerpecker

  Black-capped Flufftail

  Iceland Reedhaunter

  Snowy Mourning Heron-Robin

  Or new perfumes:

  Fancy Ten

  Eau de Boffe

  Frogant Flower

  Momite

  Santa for Women

  Or even new recipes.

  BASIC CLAM FROSTING

  main dish, soups

  1 lb chicken

  1 lb pork, cubed

  ½ clove garlic, crushed

  1 cup celery, sliced

  1 head (about ½ cup)

  6 tablespoon electric mixer

  1 teaspoon black pepper

  1 onion—chopped

  3 cup beef broth the owinls for a fruit

  1 freshly crushed half and half; worth water

  With pureed lemon juice and lemon slices in a 3-quart pan.

  Add vegetables, add chicken to sauce, mixing well in onion. Add bay leaf, red pepper, and slowly cover and simmer covered for 3 hours. Add potatoes and carrots to simmering. Heat until sauce boils. Serve with pies.

  If the liced pieces cooked up desserts, and cook over wok.

  Refrigerate up to ½ hour decorated.

  Yield: 6 servings

  JUST LET THE AI FIGURE IT OUT

  Given a set of knock-knock jokes and no further instruction, the AI was able to discover a lot of the rules
that I would have otherwise had to manually program into it. Some of its rules I would never have thought to program in or wouldn’t even have known existed—such as The Cow with No Lips Is the Best Joke.

  This is part of what makes AIs attractive problem solvers, and is particularly handy if the rules are really complicated or just plain mysterious. For example, AI is often used for image recognition, a surprisingly complicated task that’s difficult to do with an ordinary computer program. Although most of us are easily able to identify a cat in a picture, it’s really hard to come up with the rules that define a cat. Do we tell the program that a cat has two eyes, a nose, two ears, and a tail? That also describes a mouse and a giraffe. And what if the cat is curled up or facing away? Even writing down the rules for detecting a single eye is tricky. But an AI can look at tens of thousands of images of cats and come up with rules that correctly identify a cat most of the time.

  Sometimes AI is only a small part of a program while the rest of it is rules-based scripting. Consider a program that lets customers call their banks for account information. The voice-recognition AI matches spoken sounds to options in the help-line menu, but programmer-issued rules govern the list of options the caller can access and the code that identifies the account as belonging to the customer.

  Other programs start out as AI-powered but switch control over to humans if things get tough, an approach called pseudo-AI. Some customer-service chat windows work like this. When you begin a conversation with a bot, if you act too confusing, or if the AI detects that you are getting annoyed, you may suddenly find yourself chatting with a human instead. (A human who unfortunately now has to deal with a confused and/or annoyed customer—maybe a “talk to a human” option would be better for customer and employee.) Today’s self-driving cars work this way, too—the driver has to always be ready to take control if the AI gets flustered.

  AI is also great at strategy games like chess, for which we know how to describe all possible moves but not how to write a formula that tells us what the best next move is. In chess, the sheer number of possible moves and complexity of game play means that even a grandmaster would be unable to come up with hard-and-fast rules governing the best move in any given situation. But an algorithm can play a bunch of practice games against itself—millions of them, more than even the most dedicated grandmaster—to come up with rules that help it win. And since the AI learned without explicit instruction, sometimes its strategies are very unconventional. Sometimes a little too unconventional.

  If you don’t tell AI which moves are valid, it may find and exploit strange loopholes that completely break your game. For 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 a rules-based strategy, built an AI that could evolve its own approach. Surprisingly, the AI suddenly began winning all its games. It turned out that the AI’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 effort would cause it to run out of memory and crash, forfeiting the game.1 Most AI programmers have stories like this—times when their algorithms surprised them by coming up with solutions they hadn’t expected. Sometimes these new solutions are ingenious, and sometimes they’re a problem.

  At its most basic, all AI needs is a goal and a set of data to learn from and it’s off to the races, whether the goal is to copy examples of loan decisions a human made or predict whether a customer will buy a certain sock or maximize the score in a video game or maximize the distance a robot can travel. In every scenario, AI uses trial and error to invent rules that will help it reach its goal.

  SOMETIMES ITS RULES ARE BAD

  Sometimes, an AI’s brilliant problem-solving rules actually rely on mistaken assumptions. For example, some of my weirdest AI experiments have involved Microsoft’s image recognition product, which allows you to submit any image for the AI to tag and caption. Generally, this algorithm gets things right—identifying clouds, subway trains, and even a kid doing some sweet skateboarding tricks. But one day I noticed something odd about its results: it was tagging sheep in pictures that definitely did not contain any sheep. When I investigated further, I discovered that it tended to see sheep in landscapes that had lush green fields—whether or not the sheep were actually there. Why the persistent—and specific—error? Maybe during training this AI had mostly been shown sheep that were in fields of this sort and had failed to realize that the “sheep” caption referred to the animals, not to the grassy landscape. In other words, the AI had been looking at the wrong thing. And sure enough, when I showed it examples of sheep that were not in lush green fields, it tended to get confused. If I showed it pictures of sheep in cars, it would tend to label them as dogs or cats instead. Sheep in living rooms also got labeled as dogs and cats, as did sheep held in people’s arms. And sheep on leashes were identified as dogs. The AI also had similar problems with goats—when they climbed into trees, as they sometimes do, the algorithm thought they were giraffes (and another similar algorithm called them birds).

  Although I couldn’t know for sure, I could guess that the AI had come up with rules like Green Grass = Sheep, and Fur in Cars or Kitchens = Cats. These rules had served it well in training but failed when it encountered the real world and its dizzying variety of sheep-related situations.

  Training errors like these are common with image recognition AIs. But the consequences of these mistakes can be serious. A team at Stanford University once trained an AI to tell the difference between pictures of healthy skin and pictures of skin cancer. After the researchers trained their AI, however, they discovered that they had inadvertently trained a ruler detector instead—many of the tumors in their training data had been photographed next to rulers for scale.2

  HOW TO DETECT A BAD RULE

  It’s often not that easy to tell when AIs make mistakes. Since we don’t write their rules, they come up with their own, and they don’t write them down or explain them the way a human would. Instead, the AIs make complex interdependent adjustments to their own internal structures, turning a generic framework into something that’s fine-tuned for an individual task. It’s like starting with a kitchen full of generic ingredients and ending with cookies. The rules might be stored in the connections between virtual brain cells or in the genes of a virtual organism. The rules might be complex, spread out, and weirdly entangled with one another. Studying an AI’s internal structure can be a lot like studying the brain or an ecosystem—and you don’t need to be a neuroscientist or an ecologist to know how complex those can be.

  Researchers are working on finding out just how AIs make decisions, but in general, it’s hard to discover what an AI’s internal rules actually are. Often it’s merely because the rules are hard to understand, but at other times, especially when it comes to commercial and/or government algorithms, it’s because the algorithm itself is proprietary. So unfortunately, problems often turn up in the algorithm’s results when it’s already in use, sometimes making decisions that can affect lives and potentially cause real harm.

  For example, an AI that was being used to recommend which prisoners would be paroled was found to be making prejudiced decisions, unknowingly copying the racist behaviors it found in its training.3 Even without understanding what bias is, AI can still manage to be biased. After all, many AIs learn by copying humans. The question they’re answering is not “What is the best solution?” but “What would the humans have done?”

  Systematically testing for bias can help catch some of these common problems before they do damage. But another piece of the puzzle is learning to anticipate problems before they occur and designing AIs to avoid them.

  FOUR SIGNS OF AI DOOM

  When people think of AI disaster, they think of AIs refusing orders, deciding that their best interests lie in killing all humans, or creating terminator bots. But all those disaster scenarios assume a level of critical thinking a
nd a humanlike understanding of the world that AIs won’t be capable of for the foreseeable future. As leading machine learning researcher Andrew Ng put it, worrying about an AI takeover is like worrying about overcrowding on Mars.4

  That’s not to say that today’s AIs can’t cause problems. From slightly annoying their programmers all the way to perpetuating prejudices or crashing a self-driving car, today’s AIs are not exactly harmless. But by knowing a little about AI, we can see some of these problems coming.

  Here’s how an AI disaster might actually play out today.

  Let’s say a Silicon Valley startup is offering to save companies time by screening job candidates, identifying the likely top performers by analyzing short video interviews. This could be attractive—companies spend a lot of time and resources interviewing dozens of candidates just to find that one good match. Software never gets tired and never gets hangry, and it doesn’t hold personal grudges. But what are the warning signs that what the company is building is actually an AI disaster?

  Warning sign number 1: The Problem Is Too Hard

  The thing about hiring good people is that it’s really difficult. Even humans have trouble identifying good candidates. Is this candidate genuinely excited to work here or just a good actor? Have we accounted for disability or differences in culture? When you add AI to the mix, it gets even more difficult. It’s nearly impossible for AI to understand the nuances of jokes or tone or cultural references. And what if a candidate makes a reference to the day’s current events? If the AI was trained on data collected the previous year, it won’t have a chance of understanding—and it might punish the candidate for saying something it finds nonsensical. To do the job well, the AI will have to have a huge range of skills and keep track of a large amount of information. If it isn’t capable of doing the job well, we’re in for some kind of failure.