2018, a breakthrough year for AI
Technology fads come and go. Most of them just go, but one which has come and looks like staying is Artificial Intelligence (AI). It is certainly the marketing buzzword of the year and it is hard to find any new product which doesn't include AI in the spec sheets, or any new research which doesn't propose using an AI methodology.
Look at some of the fashions and fads we have had in the last few years. It is not that long ago when domain names were the bee's knees and people were buying them in the belief, fuelled by a few select news stories, that you could resell them for thousands or even millions and magically get rich quick.
In recent years, BitCoin has attracted the same fervour. It became a bandwagon that people were desperate to jump onto, which promised to be "the next big thing", and which its inventors claimed would shake up the world economic order, but the bubble has burst and now it is mainly scammers and charlatans who regard it as the currency of choice, whilst environmentalists cringe at the megajoules of energy burnt by computer networks to create new digital coinage.
Bitcoin is built upon a distributed technology called the blockchain, and just as Bitcoin was the darling of the popular press for a year or so, blockchain was the buzzword of the marketing departments. There were seminars on blockchain, training courses on blockchain, job adverts requiring blockchain experience, tenders for accounting systems which must utilise blockchain, but where are all the blockchain projects today?
The US Agency for International Development (USAID) has recently looked at 34 cases of blockchain being used in NGOs, contractors, and agencies, and reported:
"We found a proliferation of press releases, white papers, and persuasively written articles. However, we found no documentation or evidence of the results blockchain was purported to have achieved in these claims. We also did not find lessons learnt or practical insights."
Not only did they find no examples of projects which had achieved anything with blockchain, but they also found that none of the vendors of blockchain solutions that they approached was willing to share data about a successful project. Once the hype was stripped away, there was zero evidence that blockchain has ever delivered results.
A similar study earlier this year by the Australian government likewise concluded that blockchain projects held no value for them.
Meanwhile, in Britain, Eddie Hughes, the Conservative MP for Walsall North, confessed to being "just a Brummie bloke who kept hearing about blockchain, read a bit about it, and thought: this is interesting stuff", and promptly proposed "Blockchain For Bloxwich". Bloxwich is a town in the North Walsall constituency. The reason it was singled out for blockchain fame is because "Blockchain for North Walsall" didn't sound cool.
Did you 3D print your Christmas presents?
Another "next big thing" that was going to change our lives was 3D printing, and if we believed the hype, by now, some seven or eight years later, we should all have 3D printers at home and be routinely using them for things we could not imagine in 2010. Unfortunately, we couldn't really imagine them in 2018 either.
However, they have made an impact in industry and we have seen, for example, that they can be used in construction to 3D print complete houses, albeit small and specialised designs, and they are useful for prototyping products.
For most of us, our closest contact with 3D printing is likely to be in the medical fields. If you have ever broken a bone, you will know how uncomfortable and cumbersome the old plaster cast was, and the more modern lightweight plastic casts are much better. 3D imaging and 3D printing allows this to be taken further, to produce a cast customised exactly to the individual patient, as this video illustrates.
Even more likely is that you will one day have a dental crown fitted which was manufactured using 3D imaging and printing. For such a seemingly simple piece of ceramic, it involves a lot of production steps:
An ill-fitting crown is an uncomfortable experience. These advances in dentistry promise much more precise fitting and consistent quality, and that will benefit us all.
2018, the year of machine learning
The hype for previous technologies is insignificant compared to the coverage given to AI or machine learning in 2018, and it feels like it gets mentioned in every press release, every project proposal, every research paper. However, it has made massive inroads into subjects such as voice recognition and natural language translation, and is undoubtedly here to stay.
AI is widely touted as the future of self-driving cars, but here is an area where we need to be especially careful that hype does not exceed reality, and remind ourselves that "artificial intelligence" is just a convenient label for a computational technique, and computers are not really intelligent. Waymo, the self-driving arm of Google, has recently said that AI techniques alone, where the AI system learns to mimic the human driver, will not produce self-driving cars which drive well.
Waymo trained one of its systems with the equivalent of 30 million driving examples, and found that this was still inadequate to give a wide enough range of scenarios. For example, in its simulations, the self-driving car would sometimes get stuck behind cars parked at the side of the road in unusual places, and lacking a human's ability to cope with strange situations, it had no way of determining what to do next. Waymo says it is because of situations such as this that its systems are currently based largely on rules drawn up by driving experts..
Advocates of the fully autonomous vehicles should consider this real-world case of a life or death situation, a forest fire in Canada. This is one of many dashcam videos of residents being evacuated, and sometimes driving on the wrong side of the road to try to keep away from the heat of the flames, or to make room for emergency vehicles to get through.
It is unlikely that a self-driving AI would have had enough data to know what to do in that situation, and the worst case scenario would be that the self-driving cars, confused by smoke and falling debris, would reach the conclusion that they needed to enter a failsafe mode of coming to a complete stop, with no way for the fallible human occupants to override the cold logic of the machine.
Indistinguishable from the real thing
Photography is another rich source of data for AI applications, and graphics specialist Nvidia, who make the graphics chips in many of our computers and phones, have demonstrated how they have used AI to generate photos which, at screen scales, are indistinguishable from real life.
Apart from the few faces shown at the start of this video which are used to parameterise the AI, all of the people you see in this video are figments of the computer program, and yet you might still think you see someone you recognise, and the same techniques could just as easily be applied to any other sort of picture such as cars, cats, or landscapes.
It strikes me that one very real application of the Nvidia system could be in law enforcement, for the creation of realistic identikit images. However, the danger of that realism is that the public will believe the identikit is an exact photo, and fail to alert police to suspects because, for example, the identikit image has the wrong colour eyes or is missing a scar on the cheek.
26th December 2018
This article comes from the SKILLZONE email newsletter, published monthly since January 2008, and covering topics related to technology and the internet. All articles and artwork in the SKILLZONE newsletter are orignal content. If you would like to receive the newsletter direct to your inbox each month, please SUBSCRIBE here. It is free, and you don't get added to any other mailing lists. It uses best-practice confirmed opt-in only, and you may unsubscribe at any time.