Tag Archives: technology

The gorilla problem

Posted on by

I was so keen to read Stuart Russell’s new book on AI, Human Compatible: AI and the Problem of Control, that I ordered it three times over. I’m not disappointed (although I returend two copies). It’s a very interesting and thoughtful book, and has some important implications for welfare economics – an area of the discipline in great need of being revisited, after years – decades – of a general lack of interest.

The book’s theme is how to engineer AI that can be guaranteed to serve human interests, rather than taking control and serving the specific interests programmed into its objective functions and rewards. The control problem is much-debated in the AI literature, in various forms. AI systems aim to achieve a specified objective given what they perceive from data inputs including through sensors. How to control them is becoming an urgent challenge – as the book points out, by 2008 there were more objects than people connected to the internet, giving AI systems ever more extensive access, input and output, to the real world. The potential of AI is their scope to communicate – machines can do better than any number n of humans because they access information that isn’t kept in n separate brains and communicated imperfectly between them. Humans have to spend a ton of time in meetings, machines don’t.

Russell argues that the AI community has been too slow to face up to the probability that machines as currently designed will gain control over humans – keep us at best as pets and at worst create a hostile environment for us, driving us slowly extinct, as we have gorillas (hence the gorilla problem). Some of the solutions proposed by those recognising the problem have been bizarre, such as ‘neural lace’ that permanently connects the human cortex to machines. As the book comments: “If humans need brain surgery merely to survive the threat posed by their own technology, perhaps we’ve made a mistake somewhere along the line.”

He proposes instead three principles to be adopted by the AI community:

  • the machine’s only objective is to maximise the realisation of human preferences
  • the machine is initially uncertain about what those preferences are
  • the ultimate source of information about human preferences is human behaviour.

He notes that AI systems embed uncertainty except in the objective they are set to maximise. The utility function and cost/reward/loss function are assumed to be perfectly known. This is an approach shared of course with economics. There is a great need to study planning and decision making with partial and uncertain information about preferences, Russell argues. There are also difficult social welfare questions. It’s one thing to think about an AI system deciding for an individual but what about groups? Utilitarianism has some well-known issues, much chewed over in social choice theory. But here we are asking AI systems to (implicitly) aggregate over individuals and make interpersonal comparisons. As I noted in my inaugural lecture, we’ve created AI systems that are homo economicus on steroids and it’s far from obvious this is a good idea. In a forthcoming paper with some of my favourite computer scientists, we look at the implications of the use of AI in public decisions for social choice and politics. The principles also require being able to teach machines (and ourselves) a lot about the links between human behaviour, the decision environment and underlying preferences. I’ll need to think about it some more, but these principles seem a good foundation for developing AI systems that serve human purposes in a fundamental way, rather than in a short-term instrumental one.

Anyway, Human Compatible is a terrific book. It doesn’t need any technical knowledge and indeed has appendices that are good explainers of some of the technical stuff.  I also like it that the book is rather optimistic, even about the geopolitical AI arms race: “Human-level AI is not a zero sum game and nothing is lost by sharing it. On the other hand, competing to be the first to achieve human level AI without first solving the control problem is a negative sum game. The payoff for everyone is minus infinity.” It’s clear that to solve the control problem cognitive and social scientists and the AI community need to start talking to each other a lot – and soon – if we’re to escape the fate gorillas suffered at our hands.

41M3a7N6caL._SX323_BO1,204,203,200_

Human Compatible by Stuart Russell

 

Thinking strategically about platforms

Posted on by

Digital platforms have been very much a focus of policy attention of late, with reports on the problems and challenges they raise published by the UK (several, including the Furman Review), Australia, Germany, European Commission & others. The platforms these various reports discuss are the big ones, and the concerns range from competition policy to employment practices to online harms.

A terrific new book by Michael Cusumano, Annabelle Gawer and Devid Yoffie, The Business of Platforms, points out though that most of the digital platforms that are not big are dead: four in five fail. The book is aimed at people running or starting platforms, offering advice on (as the subtitle puts it). “Strategy in the Age of Digital Competition, Innovation and Power).” The book very nicely links business strategy to the underlying economic characteristics of digital, and I think is probably in this respect the best tech business book since Shaprio and Varian’s (now old, 1998) Information Rules.

It starts by pointing out that there is nothing inevitable about network effects (direct and indirect) kicking in: they have to be nurtured: “Companies and governments have to make the right srategic and policy decisions in order to drive strong network effects.” These can include technical standards, for example, or ensuring competition thrives at the right times and points. The book also distinguishes between two types of platform, requiring different strategies (although there are a groing number of hybrids). Innovation platforms create value by enabling third parties to develop products or service on top of the platform, while transaction platofrms create value by matching different sides of a market.

Key challenges for all, though, involve solving the ‘chicken and egg’ problem (because different sides of the platform depend on each other) by appropriate pricing and cross-subsidy, and figuring out a business model. (And in my view the dependence of so many on advertising is a major weakness & can’t be sustained). The book uses the framework to explore the many platform failures. It also has a chapter on how non-platform incumbents can respond to the digital challenge (it’s tough…), and looks briefly at issues such as the use and governance of data, and also the importance of working with regulators rather than against them and recognizing the responsibilities that come with (market and other) power. “Every major company we cited in this book has been the subjject of government investigations, local regulatory oversight, and intense media scrutiny.”

All in all, highly recommended. If you know the economics, the case studies and management literature covered will be informative, and if you know the business details, the economic framework should be useful. I very much enjoyed reading it.

41rS0XlCETL._SY346_

 

The Technology Trap

Posted on by

Anybody interested in the economic impact of digital and AI, in particular on jobs, will want to read Carl Frey’s new book, The Technology Trap: Capital, Labor and Power in the Age of Automation. He is probably best known for his rather gloomy work with Michael Osborne (original pdf version here) highlighting the vulnerability of many jobs – almost half in the US – to automation in the next couple of decades. The book expands on the issues that will determine the actual outcomes, and is – as the title indicates – still quite pessimistic.

The structure of the book is historical, with sections on pre-industrial technologies, the Industrial Revolution (which saw widening inequalities), the mass production era (which reduced inequalities and created an affluent middle class), the recent polarization in the era of globalisation and digital, and future prospects. The key distinction Frey draws in between technologies which substitute for labour and those which complement it. Whereas the 19th century and the present seem to involve the replacement of people with machines, the 20th century innovations needed increasingly skilled labour to work with them.

Although I am probably not as gloomy about future prospects for work and incomes, I really enjoyed reading the book, which covers a wide range of technological applications in addition to the well-known historical examples. It leaves open two questions. One is about the present conjuncture: what explains the combination of seemingly rapid technological change and adoption with – in at least some OECD economies – very low unemployment rates? The answer might just be ‘long and variable lags’ but the question surely needs addressing.

The broader question, or set of questions, is really about the interaction between technology and labour market and other economic institutions. Although automation is likely to have the same general effects everywhere, the outcomes for workers will be refracted through very different national job markets, education systems, tax systems and so on. How much can any individual country lean successfully against the wind? Frey is not (unlike Robert Gordon) US-centric but does not get into these issues.

And beyond the response to technological change, what is it that determines the direction of technical change in the first place? The book treats the labour substitution or complementing as exogenous. But why were electric unit drives in auto plants and internal combustion engines created as complementary and yet automation in today’s car industry seems like it will substitute for labour? It seems to me this must be an institutional story too, but I don’t think it’s been told yet.

51VabazLy7L._SX327_BO1,204,203,200_[easyazon_link identifier=”069117279X” locale=”UK” tag=”enlighteconom-21″]The Technology Trap[/easyazon_link]

 

Digital arrivals and deaths of despair

Posted on by

There’s definitely a digital theme in the new crop of books arriving at Enlightenment Towers – the left hand mini-pile here.

IMG_0292On my recent trip to Washington (for a fascinating National Academies/Royal Society discussion on international co-operation on AI, culminating in this public symposium) I read the pile on the right.

The Economics of Artificial Intelligence is a terrific collection, edited by Ajay Agarwal, Josh Gans and Avi Goldfarb. It has sections on AI as a general purpose technology, jobs and inequality, regulation and the implications of machine learning for economics. The cast list of contributors is stellar. It’s far from the last word but a must-read as a starting point.

61bIH+8Vs2L._AC_UL872_QL65_[easyazon_link identifier=”022661333X” locale=”UK” tag=”enlighteconom-21″]The Economics of Artificial Intelligence: An Agenda (National Bureau of Economic Research Conference Report)[/easyazon_link]

Tom McLeish’s The Poetry and Music of Science is a persuasive comparison between creativity in the arts and in the sciences, exploring the parallels between the creative process in music, poetry, art and fiction and the discovery process in the natural sciences. Well, I was persuaded. 51wNUley1XL._SX351_BO1,204,203,200_

[easyazon_link identifier=”0198797990″ locale=”UK” tag=”enlighteconom-21″]The Poetry and Music of Science: Comparing Creativity in Science and Art[/easyazon_link]

Matthew Desmond’s Evicted is a distressing piece of reportorial sociology (Pullitzer-winning), detailing through a handful of specific individuals in Milwaukee the reality of the human crisis and housing crisis in America. The book describes the knot of poverty, drugs, ill-health, appalling housing conditions, impossible for any individual to escape. I was shocked on my recent trip to San Francisco to see the desperate condition of its large numbers of homeless people, literally worse than I have seen anywhere in the world. The conditions described in Evicted are intolerable. I recently heard Angus Deaton talk about his and Anne Case’s work on the ‘deaths of despair’ in the US (and some foreshadowing of a similar if less pronounced pattern in UK data). Given the extreme social inequality in the US, its political disintegration is not surprising. The new Deaton Review here in the UK into inequality may uncover ominous similarities, and it would be good to know how other OECD countries compare/contrast.

41qhBahSGLL._SX323_BO1,204,203,200_[easyazon_link identifier=”0141983310″ locale=”UK” tag=”enlighteconom-21″]Evicted: Poverty and Profit in the American City[/easyazon_link]

A digital crop

Posted on by

I spent the holiday weekend sitting in the sunshine reading digital economy books of varios types (in between cooking for the family and playing with the 10 week-old). First up was Steffen Mau’s The Metric Society, one of the slowly expanding genre of sociology of economic measurement books. The underlying theme is the use of metrics to quantify the qualitative, and the consequences of the appearance of objectivity: “By assigning a number to the thing observed, we take a step toward objectivizing it.” At the same time, measurement ‘disembeds’ phenomena from local context and knowledge. “Numbers not only isolate information from its original context but also place it in extended comparative contexts.” The added spice in this book is the ever-growing scope of the use of data as digitalisation marches on. And, like other similar books, The Metric Society is pretty pessimistic – this implies, it suggests, a panopticon society with entrenched structures of inequality. After all, “Categorical systems, once established, become extremely hard to overthrow.” However, I decided the power of numbers gives some reason to be cheerful. As Mau writes: “The nomination power invested in indicators, data and measurements can potentially restructure whole areas of society and impose new logics of action.” As Lenin said (quoted here): “We must carry statistics to the people and make them popular.” My new motto. While people might find an obsession with economic statistics a bit – nerdy – in fact it’s a revolutionary programme!

41sQCo09PuL._SX317_BO1,204,203,200_[easyazon_link identifier=”150953041X” locale=”UK” tag=”enlighteconom-21″]The Metric Society: On the Quantification of the Social[/easyazon_link]

The second book was a proof copy of Democratic Capitalism at the Crossroads by Carles Boix, out next month. I probably shouldn’t give too much of a preview before its publication date, but this is about the interplay between the economics and politics of digital – as the subtitle puts it, ‘technological change and the future of politics.’ The first half of the book compares three modes of capitalism, the 19th century Manchester variety, the 20th century Detroit variety and the 21st century Silicon Valley one. The second part discusses the interaction between digital technology, especially AI, and the labour market. Quite a lot of this covers the economic literature on the issue of the skill bias of technical change, and the resorting of jobs into tasks in extended supply chains, so this is familiar territory. The polarisation of jobs and wages is linked to populist politics and the prognosis is somewhat gloomy – the author is a bit techno-determinist, taking the ‘half of all jobs’ to be taken by robots line as more of a forecast than a thought-experiment. The book ends with some rather generic recommendations – enhance skills, pay a universal basic income. I’m sure it’s right to draw the link between the economic and political polarisations, but I’m more in the territory of taxing multinationals, capping CEO pay, enforcing competition policy etc.

415Rzs1j8qL._SX327_BO1,204,203,200_[easyazon_link identifier=”0691190984″ locale=”UK” tag=”enlighteconom-21″]Democratic Capitalism at the Crossroads[/easyazon_link]

The third was How to Be Human in the Digital Economy by Nicholas Agar. It advocates ensuring there are ‘human’ jobs as more and more activities get automated – in effect, the book takes Baumol’s well-known prediction about the growing share of employment in the least productive sectors, and, labelling this the ‘social economy’, argues against seeking ever greater efficiency in these jobs. Although I agree – and hence it means interrogating what we mean by ‘productivity’ in different types of job – I found the book rather rhetorical. Eg, “AI is the digital superpower that thwarts traditional human responses to technological unemployment.” Whereas Boix has rather too many numbers and charts, Agar has too few. The latter’s suggestion for paying for the “less productive” social economy is the Lanier/Weyl data-as-labour idea, but otherwise it is not very specific about how to create the desired social economy.

51MF72+uFHL._SX336_BO1,204,203,200_[easyazon_link identifier=”0262038749″ locale=”UK” tag=”enlighteconom-21″]How to Be Human in the Digital Economy (The MIT Press)[/easyazon_link]

Anyway, it’s quite interesting to see this crop of books on AI/digital and the future of the capitalist democracies. No doubt there are many more to come.