Deep learning, deep insights, deep artificial minds — the list goes on and on. But with unprecedented promise comes some unprecedented peril.
Around the end of each year major dictionaries declare their “word of the year.” Last year, for instance, the most looked-up word at Merriam-Webster.com was “justice.” Well, even though it’s early, I’m ready to declare the word of the year for 2019.
The word is “deep.”
Why? Because recent advances in the speed and scope of digitization, connectivity, big data and artificial intelligence are now taking us “deep” into places and into powers that we’ve never experienced before — and that governments have never had to regulate before. I’m talking about
- deep learning,
- deep insights,
- deep surveillance,
- deep facial recognition,
- deep voice recognition,
- deep automation and
- deep artificial minds.
..Which is why it may not be an accident that one of the biggest hit songs today is “Shallow,” from the movie “A Star Is Born.” The main refrain, sung by Lady Gaga and Bradley Cooper, is: “I’m off the deep end, watch as I dive in. … We’re far from the shallow now.”
.. We sure are. But the lifeguard is still on the beach and — here’s what’s really scary — he doesn’t know how to swim! More about that later. For now, how did we get so deep down where the sharks live?
The short answer: Technology moves up in steps, and each step, each new platform, is usually biased toward a new set of capabilities. Around the year 2000 we took a huge step up that was biased toward connectivity, because of the explosion of fiber-optic cable, wireless and satellites.
Suddenly connectivity became so fast, cheap, easy for you and ubiquitous that it felt like you could touch someone whom you could never touch before and that you could be touched by someone who could never touch you before.
Around 2007, we took another big step up. The iPhone, sensors, digitization, big data, the internet of things, artificial intelligence and cloud computing melded together and created a new platform that was biased toward abstracting complexity at a speed, scope and scale we’d never experienced before.
So many complex things became simplified. Complexity became so fast, free, easy to use and invisible that soon with one touch on Uber’s app you could page a taxi, direct a taxi, pay a taxi, rate a taxi driver and be rated by a taxi driver.
Bret: Anyone who survives a half-dozen bankruptcies and goes on to win the presidency should never be written off.
Gail: Sigh. Good point.
Bret: Trump is a master of inventing new dramas to make us forget the old ones. And if unemployment and growth figures remain good a year from now, he’ll still have a powerful argument for a second term.
Bret: I’m not too worried. Capitalism survived the transition from horse-and-buggy to the Model T. It survived the transition from an agricultural economy to a manufacturing economy to a service-based one. And it survived the creative destruction of countless other forms of employment. Where, for instance, are the typesetters these days?
Gail: Well, they’re not creating hot new social media sites.
Bret: Now the question everyone is asking is what will happen to all those truck and cab and Uber drivers — a total of three million professional drivers — once driverless cars become ubiquitous. There’s no doubt the transition will be painful for some of them, and policymakers need to be sensitive on that point. But if history is any guide, things will work out. Many of those drivers will find work in industries that currently don’t exist. Just ask yourself, where was the mobile apps economy at the turn of the century? Where was the internet economy in 1990, or the personal computing industry in 1975?
Gail: I still don’t see the truck drivers working on mobile apps. And if you’re worried about the left’s solutions, I don’t see a whole lot of candidates running around talking about the state taking over the means of production.
Bret: Just wait an election cycle or two.
Gail: But if we’re moving to an economy in which trucks are automated, robots do all the warehouse work and some kind of artificial intelligence is taking orders at the restaurant, we’ll need a government that can create a whole lot of useful public service employment to make up the difference.
Bret: Heaven forfend.
Gail: And underwrite free college education for everybody who needs it.
Gail: And assure lower-middle-class people decent housing.
Bret: My soul is dying.
Gail: All of which would have to be paid for by large taxes on the very rich.
Bret: Now it’s dead.
.. Bret: I’m all for universities figuring out ways to become more affordable for those who need and deserve it, but making college free for everybody makes it bad for everybody. We would wreck a university system that’s still the envy of the world.
.. Bret: As for affordable housing, I’d sooner trust the invisible hand of the market than the heavy hand of the state. Large taxes on the very rich won’t raise the kind of income you need, and sooner rather than later those taxes will land on the decidedly less rich. And A.O.C. should start mastering her facts rather than getting into Twitter wars with fact checkers.
Gail: Hehehe. Knew I’d get you with A.O.C. That’s what people love about her.
.. Bret: I was with you until you mentioned taxes. Purely theoretical question for you (and our readers) for our next conversation: If Congress would agree to cut the top marginal rate to 33 percent in exchange for a pledge by Trump not to run again, would you take it? I’m sure we’ll be hearing from readers on the comments page.
American men do have genuine reasons for anxiety. The traditional jobs that many men have filled are disappearing, thanks to automation and outsourcing. The jobs that remain require, in most cases, higher education, which is increasingly difficult for non-affluent families to afford. We should indeed tremble for the future of both men and women in our country unless we address that problem, and related problems of declining health and well-being for working-class men.
.. Three emotions, all infused by fear, play a role in today’s misogyny. The most obvious is anger — at women making demands, speaking up, in general standing in the way of unearned male privilege. Women were once good mothers and good wives, props and supports for male ambition, the idea goes –but here they are asserting themselves in the workplace. Here they are daring to speak about their histories of sexual abuse at the hands of powerful men. It’s okay for women to charge strangers with rape, especially if the rapist is of inferior social status. But to dare to accuse the powerful is to assail a bastion of privilege to which men still cling.
.. Coupled with anger is envy. All over the world, women are seeing unprecedented success in higher education, holding a majority of university seats. In our nation many universities quietly practice affirmative action for males with inferior scores, to achieve a “gender balance” that is sometimes dictated by commitment to male sports teams, given Title IX’s mandate of proportional funding.
.. But men still feel that women are taking “their” places in college classes, in professional schools.
.. Envy, propelled by fear, can be even more toxic than anger, because it involves the thought that other people enjoy the good things of life which the envier can’t hope to attain through hard work and emulation. Envy is the emotion of Aaron Burr in Lin-Manuel Miranda’s “Hamilton”
.. And then, beneath the hysteria, lurks a more primitive emotion: disgust at women’s animal bodies.
.. In the United States, we observe this dynamic in racism, in homophobia and even in revulsion toward the bodies of people who are aging. But in every culture male disgust targets women, as emblems of bodily nature, symbolic animals by contrast to males, almost angels with pure minds.
.. Disgust for women’s bodily fluids is fully compatible with sexual desire. Indeed, it often singles out women seen as promiscuous, the repositories of many men’s fluids.
.. as with the apparent defamation of Renate Dolphin in Kavanaugh’s infamous yearbook, men often crow with pride over intercourse with a woman imagined as sluttish and at the same time defame and marginalize her.
.. Disgust is often more deeply buried than envy and anger, but it compounds and intensifies the other negative emotions.
.. Our president seems to be especially gripped by disgust: for women’s menstrual fluids, their bathroom breaks, the blood imagined streaming from their surgical incisions, even their flesh, if they are more than stick-thin.
I think one could propose a whole list of unhelpful ways of talking about current developments in machine learning. For example:
- Data is the new oil
- Google and China (or Facebook, or Amazon, or BAT) have all the data
- AI will take all the jobs
- And, of course, saying AI itself.
More useful things to talk about, perhaps, might be:
- Enabling technology layers
- Relational databases.
.. Before relational databases appeared in the late 1970s, if you wanted your database to show you, say, ‘all customers who bought this product and live in this city’, that would generally need a custom engineering project. Databases were not built with structure such that any arbitrary cross-referenced query was an easy, routine thing to do. If you wanted to ask a question, someone would have to build it. Databases were record-keeping systems; relational databases turned them into business intelligence systems.
This changed what databases could be used for in important ways, and so created new use cases and new billion dollar companies. Relational databases gave us Oracle, but they also gave us SAP, and SAP and its peers gave us global just-in-time supply chains – they gave us Apple and Starbucks.
.. with each wave of automation, we imagine we’re creating something anthropomorphic or something with general intelligence. In the 1920s and 30s we imagined steel men walking around factories holding hammers, and in the 1950s we imagined humanoid robots walking around the kitchen doing the housework. We didn’t get robot servants – we got washing machines... machine learning lets us solve classes of problem that computers could not usefully address before, but each of those problems will require a different implementation, and different data, a different route to market, and often a different company... Machine learning is not going to create HAL 9000 (at least, very few people in the field think that it will do so any time soon), but it’s also not useful to call it ‘just statistics’... this might be rather like talking about SQL in 1980 – how do you get from explaining table joins to thinking about Salesforce.com? It’s all very well to say ‘this lets you ask these new kinds of questions‘, but it isn’t always very obvious what questions.
- .. Machine learning may well deliver better results for questions you’re already asking about data you already
- .. Machine learning lets you ask new questions of the data you already have. For example, a lawyer doing discovery might search for ‘angry’ emails, or ‘anxious’ or anomalous threads or clusters of documents, as well as doing keyword searches,
- .. machine learning opens up new data types to analysis – computers could not really read audio, images or video before and now, increasingly, that will be possible.
.. Within this, I find imaging much the most exciting. Computers have been able to process text and numbers for as long as we’ve had computers, but images (and video) have been mostly opaque.
.. Now they’ll be able to ‘see’ in the same sense as they can ‘read’. This means that image sensors (and microphones) become a whole new input mechanism – less a ‘camera’ than a new, powerful and flexible sensor that generates a stream of (potentially) machine-readable data. All sorts of things will turn out to be computer vision problems that don’t look like computer vision problems today.
.. I met a company recently that supplies seats to the car industry, which has put a neural network on a cheap DSP chip with a cheap smartphone image sensor, to detect whether there’s a wrinkle in the fabric (we should expect all sorts of similar uses for machine learning in very small, cheap widgets, doing just one thing, as described here). It’s not useful to describe this as ‘artificial intelligence’: it’s automation of a task that could not previously be automated. A person had to look.
.. one of my colleagues suggested that machine learning will be able to do anything you could train a dog to do
.. Ng has suggested that ML will be able to do anything you could do in less than one second.
.. I prefer the metaphor that this gives you infinite interns, or, perhaps, infinite ten year olds.
.. Five years ago, if you gave a computer a pile of photos, it couldn’t do much more than sort them by size. A ten year old could sort them into men and women, a fifteen year old into cool and uncool and an intern could say ‘this one’s really interesting’. Today, with ML, the computer will match the ten year old and perhaps the fifteen year old. It might never get to the intern. But what would you do if you had a million fifteen year olds to look at your data? What calls would you listen to, what images would you look at, and what file transfers or credit card payments would you inspect?
.. machine learning doesn’t have to match experts or decades of experience or judgement. We’re not automating experts. Rather, we’re asking ‘listen to all the phone calls and find the angry ones’. ‘Read all the emails and find the anxious ones’. ‘Look at a hundred thousand photos and find the cool (or at least weird) people’.
.. this is what automation always does;
- Excel didn’t give us artificial accountants,
- Photoshop and Indesign didn’t give us artificial graphic designers and indeed
- steam engines didn’t give us artificial horses. ..
Rather, we automated one discrete task, at massive scale.
.. Where this metaphor breaks down (as all metaphors do) is in the sense that in some fields, machine learning can not just find things we can already recognize, but find things that humans can’t recognize, or find levels of pattern, inference or implication that no ten year old (or 50 year old) would recognize.
.. This is best seen Deepmind’s AlphaGo. AlphaGo doesn’t play Go the way the chess computers played chess – by analysing every possible tree of moves in sequence. Rather, it was given the rules and a board and left to try to work out strategies by itself, playing more games against itself than a human could do in many lifetimes. That is, this not so much a thousand interns as one intern that’s very very fast, and you give your intern 10 million images and they come back and say ‘it’s a funny thing, but when I looked at the third million images, this pattern really started coming out’.
.. what fields are narrow enough that we can tell an ML system the rules (or give it a score), but deep enough that looking at all of the data, as no human could ever do, might bring out new results?