We have been talking about artificial intelligence for quite a while now, but it has, so far, failed to really make its mark. It’s showing a great deal of potential, but has not yet lived up to the hype. Anyone who has used Siri or Alexa will have quickly discovered the limitations when they try to step out of a strict set of rules. So what are the benefits of AI?
However, there is no denying that it is only a matter of time before AI starts to make a big difference within multiple industries. Over the next few years, I expect AI to start to make its mark in a number of areas.
Benefits of AI: The top industries that will be transformed
Healthcare is arguably one of the industries that is likely to see the biggest growth in the use and application of AI in the next few years and this is backed up by the huge amount of investment in this industry. There are a number of areas within the healthcare industry where AI is already beginning to gain a lot of ground.
One of the things that the healthcare industry has in abundance is data. Governments and healthcare organisations have billions of data records going back many decades and mining this data to gain an insight can be a big challenge. AI is already been used to mine and analyse this data to spot subtle patterns in the progression, diagnosis and treatment of many medical conditions.
The diagnosis of a medical condition or disease is not as black and white as you may think. A diagnosis is often made by piecing together a number of indications and observations to a point where the balance of probability is sufficiently in favour of a diagnosis. I was recently surprised to hear that very few blood tests for common diseases are 100% accurate. Almost all have a few percentage points of error.
Healthcare professionals rely on years of experience to be able to recognise the signs of illness in order to make a diagnosis or will use pre-built diagnoses algorithms but they can often miss things. For example, doctors may look at an x-ray that, to most of us, looks normal but they will see a subtle shadow which can indicate an illness. How subtle does that shadow need to be before it is missed?
This BBC article reported on new research that is using AI to outperform experienced cardiologists in spotting the early signs of heart disease. The report states that, even the best doctors, get it wrong about 20% of the time. The doctors rely on their experience to help them spot tell-tale signs of disease whether that is a pattern in a scan or shadows on an x-ray. Researchers fed the system data on 1000 patients, including their scan results along with information about whether they then went on to have heart problems.
Using machine learning, the AI system was then able to more accurately spot the signs of heart disease in the scans than experienced doctors. This research is still in the early stages but is a good illustration of the potential of what AI can being to this industry.
As this and similar systems are fed more and more historical data they will be able to get better and better at spotting patterns and signs of disease and I can see AI being used more and more for medical diagnosis across the healthcare industry.
We all know about Tesla and Elon Musk’s claims about the capability of it’s Auto Pilot feature. The ability to navigate from coast to coast in the United States autonomously is often talked about as being very close. I have witnessed Auto Pilot first hand and, on a motorway, it was very impressive. However, that was a motorway which has fairly straight lanes, and nice clear road markings. Navigating through the centre of London or, even harder, a single-track country lane with hedges either side is an entirely different proposition. In my opinion, we are a very long way from AI powerful and experienced enough to safely navigate a very complex journey on its own.
Car manufacturers are now looking at embedding AI services such as Amazon Alexa into their cars to allow passengers to control technology within the car through natural language voice commands. At this year’s CES show in Las Vegas, Mercedes-Benz demonstrated their new AI-powered in-car personal assistant and in 2017 BMW announced that it would start integrating Alexa into selected BMW and Mini vehicles in 2018. It has also recently been announced that Kia will soon be embedding Google Assistant into their infotainment systems.
Undoubtedly, we will see increasingly powerful AI within new cars over the next few years, used for both navigation as well as in-car virtual assistants but I think it will be many years (even decades) before we have cars that uses AI to be truly autonomous.
One of the more worrying trends I expect to see over the next few years is the use of AI in cyber attacks. This has been happening for quite a while now in a relatively basic form. For years, the internet has been awash with bots that are constantly poking at web servers looking for vulnerabilities. As soon as they find a vulnerability they will report this back to their owner or automatically exploit the vulnerability. These bots, however, are fairly unintelligent and will not learn and automatically adapt their behaviour based on what they find.
This is where I see AI being exploited and weaponised. Rather than simply poke at servers looking for holes, AI powered bots will have the ability to learn, adapt and evade cyber security. It is often said that humans are the weakest link when it comes to cyber security. AI tools are already being developed that can learn what phishing techniques are most effective and to automatically create phishing campaigns that are better than those created by humans. This technique was tested by two data scientists from security company ZeroFox in 2016. They built an AI tool that would use machine learning to determine what phishing techniques gained the best results and adapted the emails based on this learning. In tests, the AI tool significantly outperformed a human.
This view that AI will be increasingly weaponised is shared by the industry. During the Black Hat USA 2017 conference in July last year 62% of surveyed attendees agreed that there is a high possibility that AI could be used by hackers for offensive purposes.
While AI is expected to be weaponised over the next few years for offensive cyber attacks it is also expected that AI will be used defensively within the cyber security industry. One of the key roles of any cyber security system is to recognise threats and protect against them. This is usually done by recognising threat signatures that match a pre-defined list.
AI and machine learning can be used to identify malicious behaviour that does not necessarily have a known signature and then defend against that behaviour. While this approach is still at a very early stage, I expect it to become more prevalent, especially as the offensive use of AI becomes more widespread.
E-commerce and customer service
Retail is one of the fastest moving industries in the world and e-commerce retail is even faster. Competition is often fierce and this drives innovation within the industry.
Chances are, you have already experienced AI within e-commerce but you may not have noticed. Every time Amazon recommends a product to you, this is driven by AI. A very complex set of algorithms are used to determine what you are likely to purchase based on your demographic profile, your purchasing history and what other products you have viewed. Amazon generates vast quantities of data and this data can be used by AI to generate highly targeted recommendations.
You may also have used a live chat tool, either on a website or on a platform such as Facebook to communicate with a brand. There is a good chance that, at least once, you have been speaking to an AI-powered bot that is feeding you a pre-set range of replies based on your comments.
Customer services is the perfect area for automation using AI. Most customer service queries follow a very similar pattern such as ‘where is my order?’ or ‘can I change the delivery address?’. Customer service agents will normally have a script to follow based on the query and the majority of queries will fit into a small set of scenarios.
If you consider the scenario where a customer calls to ask when their order is going to be delivered, the customer services agent will probably ask the customer to authenticate themselves, maybe with an order number and postcode and then search for that order within their internal systems to find the status and delivery date. The customer may then ask to change it and the agent may do that for them.
This process could very easily be automated as it does not really take any initiative from the agent as they are following a standard and scripted process. By automating processes like this, humans can be freed up to deal with the more complex queries that AI would struggle to handle.
It will be interesting to see how AI will be used within e-commerce over the next few years. I predict that AI will have its biggest impact within customer services but also in user personalisation to provide more targeted recommendations and experience to users.
Virtual Personal Assistants
There is currently a fierce battle taking place between Apple (HomePod), Google (Google Home) and Amazon (Echo) over home virtual assistant devices. Right now, Amazon seems to be winning with the Echo powered by Alexa. The skills of these assistants are fairly basic at the moment and are mainly limited to choosing music, answering a few questions and controlling home automation devices.
I expect to see big advances in the capability of these devices over the next few years, especially with some of the biggest and most innovative companies in the world behind them. I would expect to see their integration with more consumer devices and home automation systems and also for their AI to improve significantly. The software that powers these devices is already being integrated into other products such as cars, Sonos and even an LG Fridge and I would predict that this trend will start to accelerate in the next 12 months.
In summary, it seems that the development and use of AI is beginning to accelerate and we are likely to see AI becoming much more prevalent across many industries over the next few years. I have picked a few industries here but, in reality, AI is likely to have an impact across almost all industries to some degree.
I do, however, think that we are a very long way from AI that can handle situations outside of a clear set of predefined scenarios. My car is not likely to be driving me all the way to work any time soon and you are still going to need to vacuum your stairs for years to come.