Calum Chace's Blog, page 17

Christmas Number One

Sorry, but I couldn’t resist sharing this.  “Surviving AI” is the Christmas Number One.*

A very Merry Christmas to you, and a Happy New Year!

* On Amazon’s top 100 new releases in AI and Machine Learning.  (On Christmas Eve, anyway.)

 

Funding for dedicated organisation to study AI opportunity and risk

It’s great news that the Leverhulme Trust is granting £10m to Cambridge University’s Centre for the Study of Existential Risks (CSER).  The money will fund an important new interdisciplinary research centre, the Leverhulme Centre for the Future of Intelligence, to explore the opportunities and challenges of artificial intelligence, both short and long term.

Dr Seán Ó hÉigeartaigh. CSER’s Executive Director, says that the Centre will look “at themes such as different kinds of intelligence, responsible development of technology, and issues surrounding autonomous weapons and drones.”

Prominent figures associated with CSER include professor Stuart Russell, a world-leading AI researcher at the University of California, Berkeley, and professor Stephen Hawking, who famously said that if and when artificial general intelligence does arrive, “it’s likely to be either the best or worst thing ever to happen to humanity.”  So no pressure, then.

It is interesting that this funding should come from the trust founded by a brilliant entrepreneur and philanthropist, William Lever, who created a lucrative global business from soap – a business which helped create the consumer goods giant Unilever. Lever established a model town for his employees at Port Sunlight, in Cheshire, because he wanted to provide the sort of decent living standards which were rare among industrial workers at the time.  Today, the Leverhulme Trust provides £80m a year in research funding to the UK. 

But like many people of good will in those days, Lever was a deeply paternalistic man, and his commercial activities in the Congo, where he sourced palm oil, caused great suffering.  This provides an interesting metaphor for the future of AI. Our actions are so heavily affected by the morality of our time: how well we understand the world around us, and how well we can predict the impacts of our progress. The next decades of AI research will generate great good, but we also need to avoid the potential for considerable harm.

Discussion of the Economic Singularity, Fondacion BankInter, Madrid

Fondaction BankInter is a leading global think tank based in Madrid.  In 2015 it investigated the idea that machine intelligence may lead to technological unemployment.  A workshop in June lead to a report which was published in November, and the Fondacion asked me and Juan Francisco Blanes, a roboticist, to give talks at the launch.

The volume on the video is very low, so earphones are required.  And with splendid irony, my computer crashed during the presentation, so fans of schadenfreude will particularly enjoy the section at 23 minutes 37 seconds.  Fortunately, the Fondacion staff came to the rescue with great efficiency and aplomb, and the talk re-starts at 28 minutes 27 seconds.

BBC History

Seven vignettes from the history of artificial intelligence, for the BBC’s history magazine.

FiveBooks

The excellent FiveBooks site interviewed me recently, and asked me to recommend five books to read about artificial intelligence. I think this is the century of two singularities, so I chose two books about the Technological Singularity (one each by Kurzweil and Bostrom) and two about what I call the Economic Singularity, the consequences of technological unemployment (one by Martin Ford and the other by McAfee and Brynjolfsson).

The interview (here), by Sophie Roell, is quite long, but a jolly good read (in my wholly un-biased opinion).

My fifth book choice is Permutation City by my favourite science fiction writer, Greg Egan. I don’t know anyone else who deals with the implications of strong AI better than Egan.

Homo sapiens may split in two: a handful of gods, and the rest of us

Charles Arthur, a journalist who writes for The Guardian and other outlets, wrote this intriguing article about the possibility of technological unemployment, and its potential impact on society:

If you wanted relief from stories about tyre factories and steel plants closing, you could try relaxing with a new 300-page report from Bank of America Merrill Lynch which looks at the likely effects of a robot revolution.

But you might not end up reassured. Though it promises robot carers for an ageing population, it also forecasts huge numbers of jobs being wiped out: up to 35% of all workers in the UK and 47% of those in the US, including white-collar jobs, seeing their livelihoods taken away by machines.

Haven’t we heard all this before, though? From the Luddites of the 19th century to print unions protesting in the 1980s about computers, there have always been people fearful about the march of mechanisation. And yet we keep on creating new job categories.

However, there are still concerns that the combination of artificial intelligence (AI) – which is able to make logical inferences about its surroundings and experience – married to ever-improving robotics, will wipe away entire swaths of work and radically reshape society.

“The poster child for automation is agriculture,” says Calum Chace, author of Surviving AI and the novel Pandora’s Brain. “In 1900, 40% of the US labour force worked in agriculture. By 1960, the figure was a few per cent. And yet people had jobs; the nature of the jobs had changed.

“But then again, there were 21 million horses in the US in 1900. By 1960, there were just three million. The difference was that humans have cognitive skills – we could learn to do new things. But that might not always be the case as machines get smarter and smarter.”

What if we’re the horses to AI’s humans? To those who don’t watch the industry closely, it’s hard to see how quickly the combination of robotics and artificial intelligence is advancing. Last week a team from the Massachusetts Institute of Technology released a video showing a tiny drone flying through a lightly forested area at 30mph, avoiding the trees – all without a pilot, using only its onboard processors. Of course it can outrun a human-piloted one.

MIT has also built a “robot cheetah” which can jump over obstacles of up to 40cm without help. Add to that the standard progress of computing, where processing power doubles roughly every 18 months (or, equally, prices for capability halve), and you can see why people like Chace are getting worried.

But the incursion of AI into our daily life won’t begin with robot cheetahs. In fact, it began long ago; the edge is thin, but the wedge is long. Cooking systems with vision processors can decide whether burgers are properly cooked. Restaurants can give customers access to tablets with the menu and let people choose without needing service staff.

Lawyers who used to slog through giant files for the “discovery” phase of a trial can turn it over to a computer. An “intelligent assistant” called Amy will, via email, set up meetings autonomously. Google announced last week that you can get Gmail to write appropriate responses to incoming emails. (You still have to act on your responses, of course.)

Further afield, Foxconn, the Taiwanese company which assembles devices for Apple and others,aims to replace much of its workforce with automated systems. The AP news agency gets news stories written automatically about sports and business by a system developed by Automated Insights. The longer you look, the more you find computers displacing simple work. And the harder it becomes to find jobs for everyone.

So how much impact will robotics and AI have on jobs, and on society? Carl Benedikt Frey, who with Michael Osborne in 2013 published the seminal paper The Future of Employment: How Susceptible Are Jobs to Computerisation? – on which the BoA report draws heavily – says that he doesn’t like to be labelled a “doomsday predictor”.

He points out that even while some jobs are replaced, new ones spring up that focus more on services and interaction with and between people. “The fastest-growing occupations in the past five years are all related to services,” he tells The Observer. “The two biggest are Zumba instructor and personal trainer.”

Frey observes that technology is leading to a rarification of leading-edge employment, where fewer and fewer people have the necessary skills to work in the front line of its advances. “In the 1980s, 8.2% of the US workforce were employed in new technologies introduced in that decade,” he notes. “By the 1990s, it was 4.2%. For the 2000s, our estimate is that it’s just 0.5%. That tells me that, on the one hand, the potential for automation is expanding – but also that technology doesn’t create that many new jobs now compared to the past.”

This worries Chace. “There will be people who own the AI, and therefore own everything else,” he says. “Which means homo sapiens may be split into a handful of ‘gods’, and then the rest of us.

“I think our best hope going forward is figuring out how to live in an economy of radical abundance, where machines do all the work, and we basically play.”

Arguably, we might be part of the way there already; is a dance fitness programme like Zumba anything more than adult play? But, as Chace says, a workless lifestyle also means “you have to think about a universal income” – a basic, unconditional level of state support.

Perhaps the biggest problem is that there has been so little examination of the social effects of AI. Frey and Osborne are contributing to Oxford University’s programme on the future impacts of technology; at Cambridge, Observer columnist John Naughton and David Runciman are leading a project to map the social impacts of such change. But technology moves fast; it’s hard enough figuring out what happened in the past, let alone what the future will bring.

But some jobs probably won’t be vulnerable. Does Frey, now 31, think that he will still have a job in 20 years’ time? There’s a brief laugh. “Yes.” Academia, at least, looks safe for now – at least in the view of the academics.

Roll up, roll up! 50 FREE audio books!

Nice Mr Bezos has given me gift vouchers for 25 review copies of Pandora’s Brain audio books and 25 review copies of Surviving AI audio books.

You get a free book, and when you’ve listened to it, Amazon gets a review of the book – that’s the deal.

Gamification is everywhere these days, so these exciting freebies will be awarded for interesting / insightful / funny or even just plain honest completions of one of the two following statements:

“The best thing about AI is…”

or

“The scariest thing about AI is…”

Email your full name and your answer to calum@3cs.co.uk. Don’t forget to say which book you would prefer, and of course, first come, first served!

Surviving AI now available as an audio book

Surviving AI is now available at Amazon as an audio book – either as a stand-alone purchase, or (free!) as part of a trial of Amazon’s Audible service.

Like Pandora’s Brain, the audio version of Surviving AI is narrated by Joe Hempel, a talented voice artist who is making quite a name for himself in the world of audio books. It is already a best-seller in two categories.

Whether you go to work on the tube or the freeway, turn off the Eagles, and in just four and a half hours, learn all about the promise and peril of our most powerful technology.

SwiftKey announces the first neural net on a smartphone keyboard

SwiftKey pioneered keyboards with a three-word suggestion bar above the keys that could accurately predict your next word. This was powered by a technology called “n-gram”, an approach now used on more than a billion devices globally.

N-gram technology has some limitations, as it can’t capture the underlying meaning of words and can only accurately predict words that have been seen before in the same word sequence. Now, SwiftKey Neural’s intelligent understanding of sentence context introduces a more ‘human’ touch for mobile typing.

Using machine learning and enormous amounts of language data, SwiftKey’s neural model is able to capture the meanings of words and the relationships between them. Within the neural model, words are organised in ‘clusters’, located at varying degrees of proximity to one another.

So, having seen the phrase “Let’s meet at the airport” during training, the technology is able to infer that “office” or “hotel” are similar words which could also be appropriate predictions in place of “airport”. Further, it understands that “Let’s meet at the airport” has a similar sentence structure to “Let’s chat at the office”. This intelligence allows SwiftKey Neural to offer the most appropriate next word based on the sentence being typed.

Until now, neural network language models have been deployed mostly on large servers, requiring significant computational resources. The launch of SwiftKey Neural Alpha is the first time this type of language model technology has been able to operate locally on a smartphone – a huge challenge given the resource constraints.

And the very clever people at SwiftKey reckon they are only scratching the surface of what’s possible with this technology. 

The new Globalisation: products localise as services globalise

We are used to thinking about globalisation as a phenomenon involving products. Economic liberals see it as a good thing, enabling the law of comparative advantage to improve the lives of people all over the world, and uniting nations in peaceful trade instead of sundering them in war. Their opponents on both the political left and right see it as a bad thing, impoverishing their own citizens as cheap goods flood in from “over there”.

Globalisation is not new. When in 1492 Columbus sailed the ocean blue, he kicked off a massive wave of globalisation, but it was neither the first nor the last. The phenomenon gathered pace in the decades following World War Two, as economies recovered from war, and new technologies reduced the cost of transport and communication. It accelerated further during the 1970s, as containerisation kicked in, and China adopted capitalism (in practice if not in name).

But two current technology trends look set to reverse the globalisation of manufacturing and exchanging products. 3D printing has been around for decades, but the lapse of key patents, and the improvements in their software mean that 3D printers are now ready for prime-time. They are being used to build cars, houses, toys, and spare parts for spacecraft. They are still searching for their killer app, and for better ability to use mixed materials, but it is clear that in coming years, many products which would have been manufactured abroad and shipped to a store near you will instead be printed in your own home, or at a neighbourhood 3D printing facility.

The other globalisation-reversing technology is automation. The prices of industrial robots are falling fast, and their safety and user-friendliness is increasing. The competitive advantage of China’s cheap labour is disappearing. It costs the same to buy a robot here as it does there. (China knows this, and is investing in industrial robots faster than anyone else, partly in a bid to make sure its fast-growing domestic consumer market is supplied by Chinese companies rather than foreigners.)

But while globalisation may be going into reverse in products, services are moving in the opposite direction.

Traditionally, services have been local affairs: conceived, negotiated and transacted by people living in close proximity. With a few exceptions (in particular, some financial services) they have been too inexpensive to convey over large distances, and barriers which stood in the way included culture, and the cost and effectiveness of transport and communications.

These barriers are going or gone. The communications revolutions of the 20th century and the powerful global appeal of American culture have plastered a layer of global homogeneity over the world’s multitude of still-vibrant local cultures. Fast-improving bandwidth and interactivity in the internet means that services like media become ever easier to transmit around the world. Virtual reality will sharply accelerate that process.

Professional services like accounting and management consultancy can increasingly be delivered globally. E-commerce continues to take share from the High Street, and artificial intelligence is enabling healthcare and educational services to be delivered remotely, and globally.

As products de-globalise, services are globalising. Something to cheer both the champions and opponents of globalisation, then.