Will you lose your job to a smarter machine? Possibly. Artificial Intelligence (AI) has been unavoidable in the media recently - from cars and trucks without drivers to lawyers and accountants without jobs. We dive into AI to separate the facts from the headlines and find out what AI can, can't and could do to benefit us all.
Hollywood and the media are fond of painting horror scenarios of machines developing minds of their own, of smart rogue robots turning on their dumb masters. That’s not what AI is about. Most new technologies are big on hype and false expectations. It takes a while for the dust to settle, and for a clear picture to emerge.
First of all, AI is a sloppy moniker: there’s nothing artificial here, nor does it signal a new level of intelligence. People tend to confuse intelligence and consciousness, yet ‘machine learning’ doesn’t mean either. ‘There may well be a breakthrough that makes higher levels of intelligence possible,’ says Andrew NG in the Harvard Business Review, ‘but there is still no clear path yet to this goal.’
AI – what is it?
AI is simply an extension of what we’ve been doing for decades - programming machines - but computers are now much faster with far higher capacity. They can complex programs that would have choked up your old Windows PC for 3 years. Big data analysis is a simple example: it’s a computer program that can scan vast amounts of data in very little time, and pick out just the items of interest to the user. For instance, if you're in B2C marketing, you'll be interested in consumer behaviour; if you’re in enterprise security, you be keen to find clues of intruders.
Computers perform these kinds of functions with more speed, accuracy and consistency than humans ever could. However, these functions are restricted to
rule- or parameter-based repetitive tasks that can be programmed - like assembly of motorcars by robots on a production line. As soon as a situation
requires judgement, AI's limitations become stark. Current AI systems are nothing more than sets of programs that will faithfully execute the instructions
they’re given. But that's might be about to change.
What AI can’t do
Current AI systems use supervised machine learning, but over the horizon are systems that can learn on their own and that's a whole new ball game. We're talking about neural networks such as Google's DeepMind, which does not rely on pre-defined algorithms but learns on the run.
DeepMind’s founder and CEO Demis Hassabis is a former child chess prodigy. With AI he targeted the ancient Chinese game of Go, which has more possible positions than there are atoms in the universe. Last year, DeepMind’s AlphaGo beat world Go champion Lee Sedol. In order to master the game, AlphaGo played itself millions of times until it had figured out how to win.
Unsupervised learning is an obvious next step, but we’re still talking about acting on knowledge or data. It’s not so easy for machines to learn to be creative or inventive, or play music, paint and write novels, or make great strides in scientific discovery.
What can AI do for us?
Manufacturing was an early adopter, with the first Lexus car put together largely by robots. Chess was another early battleground, when Deep Blue beat Gary Kasparov two decades ago. These days, chess champions use supercomputers to test new strategies. Chess is rule-based, so it’s a perfect application for computers that can run through thousands of possible scenarios in seconds.
Most applications of AI are more mundane: when you search for a new camera on the web, you’ll find camera ads popping up wherever you search for the next few days or weeks. No humans are involved in the process, except for signing up with a remarketing company. It’s the same with the pop-ups for other items you might be interested in; they’re generated by recommendation engines.
When you call Telstra, you’ll talk to a speech recognition system that sounds like a woman. Other applications include search engines, photo tagging, language translation and face recognition. Even the writing of simple news stories can be computerized, according to Satya Ramaswamy from Tata Consulting.
What Jobs will AI replace?
Toyota famously built its Lexus cars in a factory of robots, but now the car maker is replacing some robots with humans. Toyota says too much automation resulted in workers feeding parts to robots, which simply keep doing what they’re programmed to do. The workers Toyota has put back on the production line will refocus on ‘Kaizen’, the art of continuous improvement.
The lists of jobs that the media tell us will be lost to computers is long and varied, ranging from telesales people and taxi drivers to lawyers and accountants. In effect all repetitive, rule-driven jobs are at risk, but this is a simplistic view of the future. It’s more likely that AI will take over the boring manual jobs, and many routine admin jobs, and allow humans to focus on value-add activities.
A recent Tata Consulting survey found that ‘AI leaders predicted that by 2020 AI will have its biggest impact on their companies’ back-office functions of IT and finance/accounting.’ If you take IT for instance, especially IT security, it is nearly always under-resourced with personnel being overwhelmed with the exploding variety and number of attacks. It's why successful attacks keep happening. AI won’t necessarily replace staff here but give them better tools and more time to do their jobs.
Footnote: In his book Average is Over, George Mason says the jobs and gains in the economy of the future are going to go to people who can work with and improve machines.
This subject has grabbed most of the headlines so far, perhaps because we’re all familiar with cars. And some may feel uneasy about handing over full control to a robot. Yet, current prototypes only cruise around at a bit over walking speed, so a crash would not be life threatening. Saving lives is what the driverless car zealots cite as the main benefit of this technology, while they blithely ignore all related problems. Driverless cars:
- Are much more expensive and much more complicated than current cars (unreliable)
- Their control systems will make attractive and easy targets for hackers and terrorists
- Sensor technology may fail in adverse weather conditions (dangerous)
- Sensors cannot detect random pedestrians or animals straying into the car’s path
- Cannot avoid another driver whose car is about to hit them
- Rely on GPS systems for navigation (danger, danger: my GPS frequently tells me that I’m going down a familiar road the wrong way. I'm not)
Better then Driverless Cars
Saving lives is a fine idea, but isn't driverless technology a chance to do a whole lot more? Solving the traffic congestion in cities, for example. Why not deploy driverless shuttles that can transport a dozen people at a time and replace a dozen cars carrying just one person? These shuttles could drive prescribed routes and leave / arrive at various stops at known intervals like buses. Because they’re smaller and more numerous, they could run on more routes and smaller routes.
And what’s the point of sticking driverless cars into their own lanes on the freeway? Why not take a train, or a bus? Or, better still, use the technology for driverless buses in special lanes or driverless trains on existing tracks, so a heap more cars could be removed from the road?
These are early days and we have the chance to think outside the cabin and consider wider uses of this liberating, game-changing technology. There could be many better uses than just handing over our cars so we can sleep or text, while robots drive.
Do you have a technology needs a bit of 'un-complication' so buyers understand it enough to want it? Contact us. It's what we do.
- Trackback Link
- Post has no trackbacks.