Calum Chace's Blog, page 16

Auto-cars Assemble!

Journalists covering the birth of the self-driving car industry are being kept very busy, and sometimes they must worry if they can keep up with all the new announcements. They haven’t even had time to pause for long enough to give the industry a proper name: “self-driving cars” is too long and too ugly. By way of giving them a helping hand, I’d like to suggest “auto-cars”. That was the name for the horseless carriage adopted by The Times of London newspaper in the late 19th century. Linguistic purists objected that it mixed Greek (auto) with Latin (car), so if anyone still cares about that sort of thing they might prefer to shorten it to “auto”.

In the last couple of weeks alone we have seen a series of significant announcements. Uber announced live trials of self-driving taxis in Pittsburgh, thereby enhancing the Steel City’s claims to rival Silicon Valley as the spiritual home of the auto-car. Uber’s legions of self-employed drivers mostly displayed a sort of battle-scarred weary acceptance of their fate as temporary stand-ins for the machines, although Uber’s CEO Travis Kalanick made a brave (some would say risible) attempt to argue that the move to auto-cars will create more jobs than it destroys because machines will not be able to handle all the routes and journeys.

Also in the last fortnight, Helsinki revealed that it is joining a small list of cities trialling driverless buses. Trikala in Greece was the first European city on the list, back in October 2015. Very sensibly, they are proceeding with caution: the robo-buses will hit top speed at a leisurely 10 miles per hour.

An article published this week listed no fewer than 19 car manufacturers, software suppliers and component manufacturers who are committed to having fully automated cars on the road by 2020. Some of them, like Volvo and Tesla, have said they will be ready years before that – if the regulators are.

This raises again the debate about how much automation to offer drivers in the early stages of the auto-car industry. Tesla has been criticised – mostly unfairly – for making its customers complacent about the need to remain vigilant when at the wheel by calling its driver assisting technology Autopilot. When someone is as wildly popular as Elon Musk has (justifiably) become there is bound to be a backlash, and his critics rail against him for putting the safety of Tesla drivers at risk with a “beta” version of the software. Musks’s response has been typically direct and acerbic.

But it is undeniable that humans struggle to pay attention when a machine appears to be doing its job perfectly. Google noticed this when it found the engineers chaperoning its experimental self-driving cars were doing dubious things like turning round to pick up laptops from the back seat while driving along the freeway at 60 miles an hour.  Google concluded that self-driving cars should be – at launch – what the US National Highway Traffic Safety Administration designates as L4. This means the car’s artificial intelligence can “perform all safety-critical driving functions and monitor roadway conditions for an entire trip”. Tesla’s Autopilot is currently L2, but the company says it will have L4 available in 2018. (L5 means the car has no steering wheel.)

These developments and announcements are all fascinating in their own right, and they certainly keep the journos busy. Collectively, they show that the ecosystem of self-driving cars is becoming varied, and ever more life-like. Only people who are just not paying attention can now deny that the 2020s will the decade of the auto-car.

But what are we to make of it all? Fundamentally, is it a good thing?

Auto-cars will have several profound impacts on humanity. The most important one is that fewer of us will be killed or maimed by cars. Human drivers currently kill 1.2m people a year on the world’s roads, and we injure a further 20 to 50 million. Globally, road traffic accidents are the leading cause of death for people aged 15 to 29, and they cost middle-income countries around 2% of their GDP, amounting to $100 billion a year. We are sending humans to do a machine’s job, and this slaughter cannot end too soon.

A less dramatic but nevertheless profound result will be that commuters will gain time and lose stress. I suspect this will have a bigger impact than we currently realise. One unexpected consequence may be that people will travel around more, as driving becomes less of a tiring chore and more like sitting in your armchair at home: comfortable, and fully in touch with your screens and digital feeds – including virtual reality. No doubt somebody has already written the science fiction story in which many people no longer have homes, but spend their time as nomads in self-driving RVs, slowly touring the world, scheduling frequent rendez-vous with friends in exotic locations, and always plugged into the network.

More prosaically, because AIs will drive more efficiently and safely than we do, they will make more efficient use of space, especially in our cities. Traffic jams and pollution will ease, and if the car-sharing model promoted by Uber and others takes hold, there could be less need for parking spaces, as cars spend less of their time sitting idly in driveways and garages.

For me, though, the most intriguing impact will be that auto-cars will probably cause the first great wave of unemployment caused by machine intelligence. An estimated 3.5m Americans are employed as truck drivers, and it is the most common profession in a majority of US states. There are also 650,000 bus drivers and 230,000 taxi drivers. (This latter figure seems very low considering how many people now make their living by driving for Uber, Lyft and so on.) By 2030, most of these people will have either found a new profession, or they will have joined the long-term unemployed. It is a huge tranche of the working population for the economy to absorb.

As mainstream economists rightly point out, automation has not caused long-term unemployment in the past. In my new book, The Economic Singularity, I argue that it is different this time, and that we need to monitor and prepare.

http://read.bi/2buQcuj

A bet about conscious machines

The estimable Robin Hanson and I are reviewing each other’s books.  We disagree with each other about almost everything, which is great.  I admire Robin’s ability to disagree strongly without a trace of animosity.

The bet arose because of the following passage in my new book, The Economic Singularity:

Geoff Hinton – the man whose team won the landmark 2012 ImageNet competition – went further.  In May 2015 he said that he expects machines to demonstrate common sense within a decade.

(Here is a report of Hinton saying that.)

Robin is famously sceptical about the rapid progress of AI, thinking it is grossly over-stated.  He thinks there is no chance at all that machines will have common sense by 23 July 2026.  And so we have agreed a wager.  He has generously given me 50-to-1 odds, so if Hinton is wrong I pay Robin $100, and Hinton is right, Robin pays me $5,000.  (He’s a tenured professor at a prestigious US university, so that kind of money is loose change to him.)

What do we mean by common sense?  I’m not sure it’s a concept capable of precise definition, but you know it when you see it.  Here’s one description, from the next paragraph in my book:  

Common sense can be described as having a mental model of the world which allows you to predict what will happen if certain actions are taken. Professor Murray Shanahan of Imperial College uses the example of throwing a chair from a stage into an audience: humans would understand that members of the audience would throw up their hands to protect themselves, but some damage would probably be caused, and certainly some upset. A machine without common sense would have very little idea of what would happen.

Rather than try to negotiate a precise definition of common sense, Robin and I have agreed to appoint an impartial judge.  We await a response from our first candidate.

To be clear, common sense falls a long way short of consciousness, or of artificial general intelligence.

The Economic Singularity – out now!

Artificial intelligence (AI) is overtaking our human ability to absorb and process information. Robots are becoming increasingly dextrous, flexible, and safe to be around (except the military ones).  It is our most powerful technology, and we all need to understand it.

In this new book I argue that within a few decades, most humans will not be able to work for money.  Self-driving cars will probably be the canary in the coal mine, providing a wake-up call for everyone who isn’t yet paying attention. All jobs will be affected, from fast food McJobs to lawyers and journalists. This is the single most important development facing humanity in the first half of the 21st century.

The fashionable belief that Universal Basic Income is the solution is only partly correct.  We are probably going to need an entirely new economic system, and we better start planning soon – for the Economic Singularity!

The outcome can be very good – a world in which machines do all the boring jobs and humans do pretty much what they please.  But there are major risks, which we can only avoid by being alert to the possible futures and planning how to avoid the negative ones.

Creative Capital at The Hospital

The Ludic Group runs a series of events at London’s Hospital Club, and on the 4th April David Wood, chair of the London Futurists and I gave talks about some of the technologies shaping our future.  The video, just 3 mins 45 seconds, is here.

ThinkNation talk

ThinkNation is a brilliant initiative. The brainchild of Lizzie Hodgson, it is “where young people, artists and thought leaders tackle how technology is impacting everyday life and shaping our futures.” Given the profound importance of the changes sweeping through our lives in the coming years and decades, there aren’t many more important subjects to address.

So I was very pleased to deliver this talk to an impressive group of 14-18 year-olds in December.

The Big Issue

For readers outside the UK, The Big Issue is a street newspaper written by professional journalists and sold by homeless people.  It is one of the UK’s biggest social enterprises.

So I am delighted that they commissioned this article about, well … a Big Issue.  Click on the logo or the image below to go to the site and read the full article.

Letter from Utopia

When Stephen Hawking, Elon Musk and Bill Gates talked about the promise and the peril of Artificial Intelligence last year, the media only heard the peril part. And the Terminator got a new lease of life. Since then a more balanced approach has been settling in.

As a wise and sapient reader of this blog, you already know that those luminaries were prompted to comment on AI by the publication of Nick Bostrom’s book “Superintelligence”. (If you haven’t read it yet, you should.)

Bostrom is often characterised as a doom-sayer, which he categorically is not. Although the New Year fireworks are but a memory, 2016 is still a young year, so I thought I would post this link to an extraordinarily uplifting snatch of optimism that Bostrom posted on his blog back in 2006.

It’s called a Letter from Utopia, and it needs no further introduction.

Click here to read the rest…

Mark Zuckerberg’s New Year Resolution

Mark Zuckerberg’s New Year Resolution was to programme an AI to become his butler.  He was aiming at Iron Man’s faithful companion, Jarvis.  But surely he knows that Robert Downey Junior based his portrayal of Iron Man on Elon Musk?

Interview with BBC Radio 5 Live.

 

A dozen AI-related forecasts for 2016

An AI system devised by DeepMind will beat the best human at Go. (Rather splendidly, he is called Yoda.)
SwiftKey’s Neural Alpha – a keyboard for phones which uses Deep Learning to dramatically improve predictive typing – will be launched to the general public and will be big news.
Virtual reality will become really quite a big thing.
The existential risk organisations (FLI in Boston, FHI at Oxford, CSER at Cambridge, MIRI in California) will continue to grow resources and awareness despite a media backlash against 2015’s excitement about all things AI.
My book The Economic Singularity, about technological unemployment, will be published.
The debate about Universal Basic Income (UBI) will go mainstream, and become politicised. The left will demand it immediately, but no country will introduce it nationwide.
Siri, Cortana etc will get much better. But we won’t settle on a generic name for them yet.
Google Glass will make a comeback, ducking the “glasshole” cynicism by targeting B2B applications.
Google will announce a significant development with their collection of robot companies.
Intel will decide whether or not it is able to keep Moore‘s Law going in the next decade. If it can’t, IBM, Samsung and others will claim that they can.
The number of items considered part of the burgeoning Internet of Things will reach seventy bazillion.
California will reverse its decision to impose bureaucratic roadblocks to self-driving cars. Buses and trains without human drivers or chaperones will appear in a few more towns and cities around the world, and more new passenger cars will contain self-driving elements.

Happy New Year!

Eight big (AI) announcements in 2015

1. The Oxford Martin Programme on Technology and Employment

In January, Citibank helped establish this programme, to be led by Carl Benedikt Frey and Michael Osborne, authors of a famous paper on AI-driven automation. The programme is monitoring changes in the labour market, and watching for signs of irreversible technological unemployment.

2. Google open-sources Tensor Flow

In September, Google announced an important change in strategy. Having built a very lucrative online advertising business based on algorithms and hardware which produced better AI than anyone else, it was open sourcing its current best AI software – a deep learning engine called Tensor Flow. It only licenses the software for single machines, so even very well resourced organisations won’t be able to replicate the functionality that Google generates, but the move was significant.

The following month, Facebook announced that it would follow suit by open sourcing the designs for Big Sur, the server which runs the company’s latest AI algorithms.

3. Google announces RankBrain

In October, Google confirmed that it had added a new technique called RankBrain to its Search offering. RankBrain is a machine learning technique, and it was already the third-most important component of the overall Search service. It is applied to the 15% of searches which comprise words or phrases that have not been encountered before, and converts the language into mathematical entities called vectors, which computers can analyse directly.

4. SwiftKey announces Neural Alpha

In October, British AI company announced the launch of Neural Alpha, the first application of the AI technique of deep learning to mobile phone keyboards.

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. Where N-gram technology predicts words that have been seen before in the same sequence, Neural Alpha’s intelligent understanding of context introduces a more ‘human’ touch for mobile typing.

5. The Leverhulme Centre for the Future of Intelligence

In December, the Leverhulme Trust announced a grant of £10m to the Cambridge University’s Centre for the Study of Existential Risks (CSER).  The centre will become an important new interdisciplinary research organisation, exploring the opportunities and challenges of artificial intelligence, both short and long term.

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

6. Open A.I.

In December, a group of Silicon Valley luminaries including Elon Musk launched Open A.I., a non-profit artificial intelligence research company with $1 billion in initial funding. Musk, CEO of Tesla Motors and SpaceX, was joined by Y-Combinator president Sam Altman as a co-chair of the new organization. LinkedIn’s Reid Hoffman, investor Peter Thiel, Amazon Web Services and Infosys are the other backers.

“It’s hard to fathom how much human-level AI could benefit society, and it’s equally hard to imagine how much it could damage society if built or used incorrectly,” the launch press release said.

7. Google’s quantum computer success

In December, Google announced that its D-Wave quantum computer was 100 million times faster than a traditional desktop computer in a “carefully crafted proof-of-concept problem”, as Google engineering director Hartmut Neven described it. Google acquired the machine in 2013, but had not previously been able to demonstrate its superiority.

Quantum computers use quantum bits (qubits) instead of the binary bits used in classical computers. These qubits can exist in a ‘superposition’ of either 0 or 1 simultaneously, allowing them to carry out a number of different calculations at once.

8. Baidu claims its speech recognition achieves human performance

December 2015, Baidu (often described as China’s Google) announced that its speech recognition system Deep Speech 2 performed better than humans with short phrases out of context. It uses deep learning techniques to recognise Mandarin.