Learning about (machine) learning

Last week I trotted off for my first Davos experience with four books in my bag and managed to read only one – no doubt old hands could have warned me what a full-on (and rather weird) experience it is. The one (and that read mainly on the journeys) was The Master Algorithm by Pedro Domingos. I was impressed when I heard him speak last year & have been meaning to read it ever since.

The book is a very useful overview of how machine learning algorithms work, and if you’ve been wondering, I highly recommend it. On the whole it stays non-technical, although with lapses – and I could have done without the lame jokes, no doubt inserted for accessibility. The book also has an argument to make: that there is an ultimate ‘master algorithm’, a sort of Grand Unified Theory of learning. This was a bit of a distraction, especially as there’s an early chapter doing the advocacy before the later chapters explaining what Domingos hopes will eventually be unified.

However, the flaws are minor. I learned a lot about both the history of the field and its recent practice, along with some insights as to how quickly it’s progressing in different domains and therefore what we might expect to be possible soon. Successive chapters set out the currently competing algorithmic approaches (the book identifies five), explains their history within the discipline and how they relate to each other, how they work and what they are used for. There is an early section on the importance of data.

As a by the by, I agree wholeheartedly with this observation: “To make progress, every field of science needs to have data commensurate with the complexity of the phenomena it studies.” This in my view is why macroeconomics is in such a weak state compared to applied microeconomics: the latter has large data sets, and ever more of them, but macro data is sparse. It doesn’t need more theories but more data. Nate Silver made a simliar point in his book The Signal and the Noise – he pointed out that weather forecasts improved by gathering much more data, in contrast to macro forecasting.

Another interesting point Domingos makes en passant is how much more energy machines need than do brains: “Your brain uses only as much power as a small lightbulb.” As the bitcoin environmental disaster makes plain, energy consumption may be the achilles heel of the next phase of the digital revolution.

I don’t know whether or not one day all the algorithmic approaches will be combined into one master algorithm – I couldn’t work out why unification was a better option than horses for courses. But never mind. This is a terrific book to learn about machine learning.


From ‘Arab Spring’ to Fake News

I’m late to Zeynep Tufekci’s excellent Twitter and Teargas: The Power and Fragility of Networked Protest. It analyses the impact of social media on political events such as the Arab Spring – remember that? – and Occupy. Her thesis is that online organization is a powerful political tool when combined with offline organization, but cannot substitute for it; and the evidence presented here from a range of mass protests certainly convinces me. The problem mass socially-networked protests have lies in their norms of decision making, which are slow and non-hierarchical. This makes them unable to change tactics quickly when hostile authorities respond to the protest, and so the moment passes. If, however, there is a parallel offline organisation with more conventional decision-making structures, the political protest can adapt and continue.

The book is informed by the author’s experiences visiting protests since long before Twitter and Facebook (and this makes it an engaging read). There is a lot of intriguing detail. Why, she asks, do longer-lived protests like Occupy Wall Street or Gezi Park set up libraries? Even clinics and clothing exchanges are probably unnecessary. She argues that the spirit of protest involves expressiveness – taking part is a meaningful or even joyful activity, often with a sense of ritual or transcendence..

A later interesting section of the book concerns the way authoritarian governments are fighting back against protests using social media themselves – not so much by cutting off the internet (which happened more in the early, Arab Spring, days than it does now) as by flooding social media with confusion and false information. I think the book was written before we all became familiar with the phrase Fake News, but here it is presented as a tactic of repression. Even in tightly controlled China, online comment is rarely shut down unless it looks like becoming organised offline action.

All in all, a highly recommended book, albeit not a particularly cheering one. The Arab Spring feels a long time ago, those days when democracy looked like it was still spreading rather than retreating. Tufekci herself ends on a slightly positive note, reporting a conversation with a young activist (one of Sain’s Indignados) about where things might go. The young woman replies with a phrase echoing one the Zapatistas used much earlier: “We will keep walking and keep asking questions.” As long as pepole have the energy to keep on, there’s hope.

I read the book on my trip to the ASSA meetings, also reading Darren McGarvey’s Poverty Safari, a second book by the wonderful new discovery Mick Herron, and Svetlana Alexievich’s Boys in Zinc. All highly recommended.

Price: £5.43
Was: £7.99


Reclaim your attention

I caught up with Timothy Wu’s The Attention Merchants: The Epic Struggle to Get Inside Our Heads on my recent travels. It’s a very interesting history of advertising, not surprisingly US-focused. There are some compelling characters in the earlier part of the story – some I knew about already like Bernays, others unknown such as the rather extraordinary Claude Hopkins. Advertising and media have always, I guess, attracted outsize characters. And some of the tales about the impact of ads are astonishing – early campaigns promoting smoking as a health-giving activity; the genius of the ‘Marlboro man’ campaign, taking Marlboro from a 1% market share to the 4th bestselling brand in the US in less than a year.

The book is excellent (as one might expect from the author of The Master Switch, a terrific book) on the interplay between technological and commercial or creative innovations. Wu writes: “Technology always embodies ideology, and the ideology in question [with the spread of cable TV] was one of difference, recognition and individuality.” The arrival of many upstart channels successfully contested the broad middle ground previously held by the main networks, and with it the cohesive, mass attention of network TV viewing.

The book moves on from old media to new media, and I also enjoyed the section on the early days of the internet and the stumbles and successes in getting all of us online. Email was an early killer app, which one forgets. “It may be hard for some to imaine a moment when receiving email was considered a big deal.” But there was indeed the movie You’ve got Mail featuring AOL in a comedy romance. Wu also reminded me that AOL got people online by sending them a physical disk in the post, or later free on the front of magazines – apparently they mailed several hundred thousand floppy disks in a 1993 mailshot – and got a 10% success rate. The company’s chief marketing officer said, “50% of the CDs produced worlwide had AOL’s logo on them,” by the late 1990s, the book reports.

As it gets towards today, the book is a bit less compelling, and I think this is because it’s so hard to get one’s mind what’s going on in the world of attention-grabbing. Of course, we know Google and Facebook are eating all the ad revenue (and attention) but there’s that fraudulent, algorithmic, complicated market, the imperative of SEO, and the fragility of offline advertising. Not to mention the blurring of news, native ads, fake news, and random dog and cat videos. It would be unfair to criticise Wu for not pinning down all this, and the addiction of social media, in a final chapter, and I’m not. The book really ends with that promising moment when ad blocking looked like a thing; but events have moved on.

There is an afterword about Trump, “the attention merchant turned President,” and then an epilogue calling for technologists to turn their focus to “the goal of reclaiming our time and attention” in these days of clickbait and the ludic loop of browsing and checking all social media without cease. The grim alternative? “The enslavement of the propoganda state as well as the narcosis of the consumer and celebrity culture.” But it isn’t just down to technologists – the battle for attention starts with oneself.


They’re watching you

I thought I was as concerned as the next person about online privacy and the harvesting of my data by big tech companies. Then I read Exposed: Desire and Disobedience in the Digital Age by Bernard Harcourt and realised that there are people who are far more worried about it than me. This is a very emotional book, and I’m not entirely sure what to make of it. Quite a lot of the material it covers is pretty familiar – for example, it draws on Edward Snowden’s revelations and all that has subsequently been written about them, and refers to books such as Tim Wu’s The Master Switch although it was presumably written before Cathy O’Neill’s Weapons of Math Destruction.

What Exposed does is join the dots to develop a picture of a society which is – not George Orwell’s Big Brother (because it exploits desire rather than repressing it), not a Surveillance State (because it is the state wedded to the private sector titans), not Jeremy Bentham’s (or rather Foucault’s) Panopticon (because the transparency is individualized not mass) – but rather the union of all of these.

The first part of the book sets out the limitations of each of these common metaphors for the digital world, arguing that the links between corporates, spies and governments are so tight they make the old military-indsutrial complex look amateurish. The second part describes what Harcourt refers to as the ‘Expository Society’, describing our willingness to reveal so much about ourselves online. The third part goes on to paint a portrait of a dystopian social collapse, with constant surveillance destroying people’s self-esteem to create passive consumers, and the border between incarceration and constant observation blurring. (In a striking comparison, the book links willingly worn smart watches to electronic tags imposed on offenders.) The final part offers a few (but not many) thoughts on how to resist.

While there is certainly much to worry about in the digital 1984/panopticon/surveillance world, I’d make three observations.

First, this is a very American-centric book. No other country (save perhaps China?) incarcerates or punishes so many of its citizens. Few others, not even the UK, are so thoroughly marketized. Germans have a completely different view of what is unacceptable in terms of invasion of privacy.

Secondlyly, there is another side to some of the phenomena. Harcourt paints as oppressive the ability of digital platforms to match more closely people’s wishes – he callis it doppelganger logic. There is something magical about this too. I thought of The Double Life of Veronique.

Finally, we can as the book suggests take measures to stop generating so much data exhaust for the big companies and spies to hoover up – the final chapter points to some steps. But we can also expect our anti-trust authorities to look closely at the duopoly of online ad revenues, the fraud in the online markets, and we can expect our governments to protect our privacy and identity. European authorities are starting to cotton on to this. We can also shop less, use the digital platforms more to swap or buy second hand – the sharing economy could yet deliver on its promise of subversion.

The most interesting chapter to me (as a statistics nerd) is the one about the evolution from classification by group in the 20th century, and the use of actuarial logic, to algorithmic data mining to pinpoint individual characteristics in the 21st century. I think Cathy O’Neill’s book shows that we are too far away from individual knowledge, in fact, and have a toxic mess of attributing group characteristics to individuals by algorithm. Anyway, this links to the emerging debate about whether there can be too much information for markets to work – insurance markets may collapse, for instance, as insurers learn too much about individuals and move away from group risk.

Anyway, I’m not going to wear an Apple watch, will check my Firefox add-ons, and will use Olio to give and take rather than sell and buy. Exposed goes over the top but it’s surely right that citizens need to worry more about privacy and digital power.


High hopes for growth

I’ve been re-reading (after a long time) David Landes’ The Unbound Prometheus, having found a 2nd hand copy in a Sedbergh bookshop, Westwood Books, earlier this year. It’s an interesting read, albeit showing its age (I found a 1993 edition, but the first was 1969). I’d forgotten that he looked at post-2nd world war growth in the final chapter, which also concludes with some reflections about the links between technological innovation and economic growth.

“The effective utilization of scientific and technical knowledge requires a whole sequence of decisions and action in the world of production and distribution. Pioneering entrepreneurs and managers must be prepared to risk money on the translation of ideas into commercially feasible techniques; while others must be incited by the prosepct of gain, or the fear of loss, to follow suit.” He goes on to say that the quality of management, the skills of the workforce, consumer tastes must also fall into place. “Scientific creativity is by no means an assurance of growth.” Interesting to see Paul David’s argument in The Dynamo and the Computer prefigured here. Growth is “a marriage of knowledge and action”, not the outcome of the impersonal forces of supply and demand.

Above all, growth will vary according to the particular needs, opportunities and history of different economies. Even the role of government varies greatly, Landes argues. The common factor in the postwar period, however, was, “A revolution of expectations and values. The expectations were not new; they were a return to the high hopes of the dawn of industrialization, to the buoyant optimism of those first generations of English innovators. Yet never before had they been so widespread; and never before had they been so strikingly confirmed by the facts.”

The trentes glorieuses seem like prehistory in today’s unsettled world. I’m not a techno-pessimist, and disagree with Robert Gordon’s thesis. But vision (expectations, we might say in economese) is a constant that really matters. I’ve always liked Paul Krugman’s 1991 QJE paper, History versus Expectations, on the importance of people weighting the future as more important than the past, if an economy is to grow. No high hopes, no growth.