Hype fuels the technology industry. A few years from now, we’ll be shuttled about in driverless cars, lose our jobs to AI and robots, and forget our troubles in virtual worlds. Hyperloop will zip us across countries, chatbots will organise our lives, and drones will deliver our shopping paid for with digital currencies. Don’t buy it? You’re not alone.
While we’ll get to plenty of arguments against overhyped future tech, it’s worth noting that gleefully imagining a sci-fi future isn’t foolish. Driverless cars could cut deaths from car crashes, robotics in factories could end industrial injuries, and digital currencies could give us financial freedom from big banks. Such ideas are worth working towards, but while dreaming of a better future, we must beware the snake-oil salesman shilling technologies that don’t exist and may never happen.
Why are such technologies overhyped? We in the media love a good headline as much as you love reading them. Startups and Silicon Valley giants alike need attention to win investment. The analyst firms are often paid by the tech industry itself, and positive predictions mean repeat customers.
That all combines to make scepticism rare, but doubt should be more common – because successful products aren’t, notes Duncan Stewart, director of research for technology, media and telecoms at Deloitte Canada. “Not only is tech not very good at inventing the next big thing, they’re not even good at inventing the next little thing.”
The global market for consumer hardware was $930 billion last year, he explains, and $830 billion of that was from three product categories: smartphones, computers and TVs. “There isn’t really anything else that threatens them,” he said. “Everybody else wants to come up with the next big thing, but since the tablet in 2010, there has not been another consumer technology that has cracked the $US10bn dollar a year barrier. Not only has there not been another $US100 billion product, there hasn’t been one a tenth that big.”
Forget the spoon-fed hype about the next big thing and read on to find out why it’s time to be dubious about virtual reality, bitcoin and the blockchain, artificial intelligence and even driverless cars.
Virtual reality headsets have long been the next big thing, but that’s not based on anything approaching reality. At the beginning of 2017, analyst firm CSS Insight predicted sales of 14 million VR headsets that year. They were wrong.
Figures from Canalys at the end of 2017 showed quarterly sales of VR headsets finally breaching the one million mark, half of which were Sony’s PlayStation VR. These headsets remain a niche product not meeting analyst predictions. And in case half a million in sales in a single quarter may not sound low, for context Apple sold 46.7 million iPhones during the same period.
VR headsets have improved and prices have fallen. So why aren’t they selling? Stewart points to one main reason: headsets are uncomfortable. They also cause motion sickness in some, and are socially isolating, as vision in both eyes is blocked. Oh, and they’re heavy: the PlayStation VR headset weighs 600g. “Wearing half a pound of electronics on our heads is not something the average consumer is keen on doing,” he said.
“People actively hate wearing them”
Those who do shell out for a heavy headset find they don’t use them as much as expected, Stewart added. “In my experience, I have never found a hardware technology that is more loathed by its users than VR headsets,” he said. “People actively hate wearing them. That’s seldom a driver of significant user adoption.”
The same challenges have followed augmented reality (AR) hardware such as Google Glass and Snap Spectacles. Both failed in the consumer market – the Snapchat developer reportedly had hundreds of thousands of the photo-snapping glasses left over in a warehouse – although Stewart predicts AR will continue to be popular on phones for games, selfies and “looking at furniture” for interior design.
VR and AR technology aren’t a total bust, as they’re already useful in the business world, particularly in manufacturing, medicine, and architecture. However, their lack of success in the consumer market still presents problems for corporate use because scale matters. Look at smartphones: they cost hundreds of pounds now, but if they weren’t manufactured in such large volumes, they’d cost ten times as much. “If VR is not a consumer success and only an enterprise tool, there will be discomfort [learning to use the] device, the developer base will be small, the pace of innovation will be slow, and it will be expensive, because it’s a million unit market and not a billion,” said Stewart. Virtual reality bites, doesn’t it?
Bitcoin has made some people very rich indeed — and made some of us wish we’d bought into the digital currency back when we were first reporting on it (darn!). But you don’t need to be bitter about missing your chance to be a Bitcoin billionaire to see there are flaws in the system.
Whether Bitcoin and other cryptocurrencies become little more than an odd historical footnote or a financial force to be reckoned with remains to be seen, but it’s not looking good. Costs are rising and exchanges can’t keep up with transfers. Scammers are creating new coins for virtual Ponzi schemes, leading Facebook to ban adverts for new currencies and to credit card companies blocking purchases of them. Hackers keep targeting the exchange left standing.
And all of this is before we consider the significant sustainability issues, as mining bitcoins chews through an immeasurable amount of energy. “None of it works right, it attracts scammers like flies, the exchanges are incompetent, and now with the bubble there’s irresponsible press coverage making people think they could get rich quick,” said David Gerard, the author of Attack of the 50 Foot Blockchain. “And of course a lot of people are going to get badly burnt when this pops.”
What’s the point of such digital currencies, other than making millionaires out of speculators? It’s not for anonymous transactions anymore, if that was ever the intent. Early Bitcoin stories marvelled at the ability to anonymously purchase anything from drugs to pizza online, but those looking to use bitcoins for shopping now may be disappointed. The drastically fluctuating value and unstable exchanges mean it’s not all that useful for shopping, as fees are high and transactions are slow. “It can’t possibly work as a useful currency – it’s really bad at the job of being a currency,” Gerard said.
Indeed, payments firm Stripe stopped supporting the currency, with founder Tom Karlo noting in a blog post that fees of tens of dollars are common, making paying with bitcoins as expensive as a bank wire. “By the time the transaction is confirmed, fluctuations in Bitcoin price mean that it’s for the ‘wrong’ amount,” he added.
Karlo said Stripe is still “optimistic” about cryptocurrencies, pointing to Bitcoin Cash – which forked from standard Bitcoin last year – and other rivals as potentially learning from the original digital currency’s mistakes. But some of those have been no more than scams, with authorities shutting down BitConnect, My Big Coin and Proof of Weak Hands Coin, while others collapse of their own accord. “I’ve never owned any bitcoins,” Gerard said. “I did have six Dogecoins, but I lost them when I reformatted the laptop they were on. If I’d held onto them they could be worth six cents now.”
“If it is digital gold, it’s still undervalued… If it’s not digital gold, its price will go to zero”
If Bitcoin isn’t for buying, what else could it be useful for? It could become an asset like gold, says Professor Ferdinando Maria Ametrano, who teaches classes on Bitcoin and blockchain at Politecnico di Milano and Milano-Bicocca University. He’s a fan of cryptocurrencies, calling it “incredible” that we’ve built a digital asset that can be transferred but not duplicated. But he said that, if Bitcoin is money, it’s useless – if it’s an asset like gold, we could be onto something.
“If it is digital gold, it’s still undervalued… If it’s not digital gold, its price will go to zero.” However, Jack Bogle, founder of the Vanguard Group, has suggested the idea of Bitcoin as an asset is based on nothing more than “the hope that you will sell it to someone for more than you paid for it”. Hence the hype.
Even if you care nothing for Bitcoin, defenders can rightly point out that it delivered blockchain unto the world. That refers to the distributed ledger that’s at the core of Bitcoin, tracking all the transactions without letting anyone meddle with the figures. Plenty believe that blockchain technology is the true innovation, with startups using blockchain to offer smart contracts, organise electronic voting, and even track consent in sexual relationships.
Gerard isn’t convinced that such projects even use real blockchains. “The main problem with ‘blockchain’ is that it’s become a hype word,” he said. “It used to mean a full Bitcoin-style blockchain, with a currency and competing coin miners and so on, but that’s completely useless for anything business does. So you have ‘blockchain’ getting redefined to mean bits of stuff that works sort of like a Bitcoin-style blockchain – but the trouble there is that the good bits aren’t original, and the original bits turn out not to be much good.”
He added: “Transaction ledgers that you can only add to, with cryptographic tamper-proofing, are the sort of idea that’s obviously useful. And of course we had them for years before bitcoin. But they seem to be getting more attention with the buzzword ‘blockchain’ attached.”
As an example, he pointed to Estonia, which is pushing the idea of a blockchain to manage its digital citizenship efforts. “But their ‘blockchain’, Guardtime KSI Blockchain, isn’t actually a blockchain, and was only rebranded a few years ago,” he said, noting the technology is essentially a tamper-evident ledger.
(Above: Tallinn, in Estonia. The country has embraced new technology. Credit: Shutterstock)
Ametrano agrees that blockchain has become a catch-all term without much meaning. “Bitcoin is a nuclear explosion, and there’s a radioactive fallout which is applied cryptography,” he said. “Many people are realising that applied cryptography can be used to harness existing business processes. They may want to call this blockchain technology, but it’s not – it’s just cryptography. That’s the confusion nowadays.”
Ametrano noted that most so-called blockchain projects are merely proof-of-concepts that haven’t been properly tested in the real world. “The fact they can work in a restricted lab doesn’t mean it could work in an adversarial environment,” he said. At the core of the problem is people use the Bitcoin blockchain because it makes them money. “Without an active digital asset or token providing economic incentive, distributed consensus cannot be reached.” Why would anyone do the work, after all, if they weren’t getting paid?
Blockchain could have uses, particularly for digital notarisation, but Ametrano argues the rest isn’t likely to work in the real world. “Frankly speaking, after two or three years of blockchain hype… they haven’t delivered a single application in production,” he said. In other words, if it’s so useful, why aren’t we already using it?
Artificial intelligence (AI) is coming for our jobs, according to a myriad of reports from the economists at PwC to academics at the University of Oxford. These suggest a third of existing employment roles will disappear into a black hole of smart computing in the next decade or so. But anyone who has used a chatbot knows it’s not time to panic – the AI-based online helpers can be irritatingly useless – and there’s plenty of examples of neural networks incorrectly labeling images or otherwise tripping up to hilarious effect.
That said, there has been a string of AI successes of late, particularly in deep learning work on games such as Go from researchers at Google and its DeepMind division. Last year, the deep-learning company’s AlphaGo Zero project beat a world champion at the game after learning the rules and tactics through trial and error, acting as its own teacher, meaning it “is no longer constrained by the limits of human knowledge”.
But such “claims are overstated”, said Gary Marcus, professor of psychology and neural science at New York University, and formerly the director of Uber’s AI labs. That’s because AlphaZero came with game-playing techniques built-in rather than starting with a blank slate. It didn’t need to learn to play games, it only needed to learn to play Go.
Those exaggerated claims aside, Marcus argues that a wider problem is “mistaking solutions on narrow, closed problems like Go for broader challenges in the open-ended world, like common sense reasoning and natural language understanding”.
Indeed, AI may be able to play Go, but that same system can’t learn to do your job. Well-designed AI, such as AlphaZero, can do a single task, but it’s not good at multitasking like us humans.
“Thus far we have lots of narrow intelligence, but nothing broad and flexible”
“I think of AGI [artificial general intelligence] as artificial intelligence that is genuinely intelligent, rather than more like a specialised idiot savant,” Marcus said. “Thus far we have lots of narrow intelligence, but nothing broad and flexible.” So we’ve had plenty of AI wins, but they’re at limited tasks; that’s still useful, but right now it means most of us are more likely to use smart computing tools in our work than be replaced by machines.
There’s another reason AI is overhyped: we only hear about AI successes, not their failures. Marcus notes in a paper that it seems unlikely the researchers using AI to win at Go haven’t tried their system on other games, but we haven’t heard about it because it failed. That he refers to as the file drawer problem, when academics leave results to rot in their desks. “People never want to report their failures; successes are much more exciting,” he said. “But the net effect (which we have recently seen come home to roost in other fields like medicine and psychology) is that you wind up with a misleading picture about how robust something is. Medical treatments in which failures aren’t reported turn out not to be reliable, and AI techniques that work in papers that selectively report successes rather than failures typically turn out to be less robust than you might expect when applied to new problems.”
Medicine makes a good example of the benefits and limits of AI. DeepMind is perhaps best known for its work in the UK’s health service, and its first project is a kidney-illness-detecting app – which uses a pre-existing algorithm to analyse symptoms, rather than AI as initially intended. Instead, the work has centred on rolling out an app that’s useful for both patients and clinicians in a care setting. Doctors and developers, your jobs are safe.
Think self-driving cars will ferry you about in the next five years? You can hit the road. That includes Chancellor Philip Hammond, who said we’ll have driverless cars on British roads by 2020, Google with its Waymo driverless car due in 2020, and Elon Musk, who promises a fully autonomous model by the end of 2018.
Transport pundit and author of Driverless Cars: On a Road to Nowhere, Christian Wolmar hasn’t always been sceptical about driverless cars. A few years back, he even wrote a letter to a local newspaper joyfully expounding the many benefits they could offer. “I was sold on the hype… but then I started reading up on it and found there was so much exaggeration.”
Wolmar offered an example of driverless cars being used to deliver pizza, a marketing stunt from Domino’s Pizza last year. “You read about it and find out there’s two people in the car with the pizzas, one’s an engineer and one’s the overseeing driver,” Wolmar said. “There are no pizzas being delivered with driverless cars. Why would you have pizza delivery with a driverless car – you want a guy to take the pizza to your door! It’s patently nonsensical.”
It’s no surprise that a much hyped, exciting idea such as driverless cars is being used for marketing pizza companies, but what of the technology itself? Tech firms and universities alike are racing to be the first to develop a viable autonomous vehicle, but demos of the prototypes aren’t always impressive. Some follow preordained routes, like trams without tracks. Others take to real roads, but Wolmar offered the the example of a Nissan trial in London where the car overtook a cyclist too closely, with the company reportedly later admitting that cyclists are a challenge because they are unpredictable. “Maybe we don’t want your driverless cars if they can’t handle cyclists,” Wolmar said.
There are plenty of other challenges. Wolmar noted that most driverless cars don’t take to the road at night, as their vision isn’t as good in the dark. To be truly autonomous, they’ll also need to manage unmapped dirt roads, navigate when snow covers road markings, and understand car parks, petrol stations and everywhere else we go.
While those may sound like problems that can be solved one by one, driverless cars aren’t developed gradually. There are six levels of automation in cars, from Level 0 with no machine help at all to Level 5 with no driver required, but development isn’t progressive. Adding a few skills doesn’t bump a car from Level 3 to 4; instead, the differences in development are exponential. For Level 5, cars will need artificial intelligence well beyond current capabilities.
“The reality will be that access to the technology will be very uneven”
That’s why many self-driving systems are actually Level 3, with tasks such as steering and braking done by the car until it meets a challenge it can’t handle. It will then pass off the driving to a human. That doesn’t work very well because we humans get distracted and stop paying attention when we’re not actively driving. It’s the model behind the fatal crash in a Tesla – the driver was watching a film and neither he nor the car spotted a lorry in the road – and Google has all but given up on the idea of Level 3 cars, leaping straight to Level 5, Wolmar claimed.
But Dr Jack Stilgoe, senior lecturer in science and technology studies at University College London, believes that Level 5 may never be possible. “The only prediction I’m willing to make is that the so-called Level 5 automated car – able to drive on any road in the world in any weather – will never exist,” he said. “There are reasons why cars are being tested in predictable, well-behaved cities like Phoenix rather than in Rome or rural Wales. The promise of self-driving cars is that everyone will benefit, that road deaths will be cut and people without access to transport… will be able to get around. The reality will be that access to the technology will be very uneven.”
Some of those benefits aren’t clear anyway. Driverless cars may promise safer roads, but if they’re forced to stop for all people, we’ll have to tighten up against jaywalking and ban pedestrians from roads, or autonomous cars will be continually halted in their tracks. Driverless proponents suggest we’ll no longer need to own cars, as autonomous vehicles can be hailed as needed, but Wolmar notes that’s no different from car sharing and Uber today. Others suggest it will mean the end of large-scale parking lots, as the cars won’t sit idle all day, but Wolmar argues they’ll still need a place to wait when not in use. “They are presenting these changes as technical changes when in fact they are social changes,” Wolmar said.
And, argues Stilgoe, if the companies selling this tech don’t admit its uncertainties, then there is a real danger of a public backlash. “When a bystander gets killed and companies start getting sued, the backlash could set back self-driving technology by years.” A little less hype, or autonomous cars may have to hit the brakes.