Book review: ‘AI in practice’ by Bernard Marr
Image credit: Dreamstime
A book based on leading examples tries to open up the secrets of how AI can deliver value to businesses and customers.
AI is thrown around as a buzz word, often without being properly understood. Yet, despite this widespread lack of understandingof the secret workings of the AI magic, its real-world applications should surely be tangible to anyone.
A new book now attempts to provide this. The average reader is given a list of 50 companies and their receipt-of-success in AI deployment.
In Bernard Marr's book ‘Artificial Intelligence in Practice: How 50 Successful Companies Used AI and Machine Learning to Solve Problems’ (Wiley, £29.99), the author presents an array of industry-leading examples, showing AI's achievements and often also its weaknesses. Although, the book may in some instances read like a marketing brochure - perhaps similar to a sensational account of the early days of the invention of the telephone or the discovery of electricity - any of the 50 company profiles exhibits a problem AI solves, the practical use, the tech, tools and data involved, results and challenges. This is usually enough to get the gist of it.
While the rather dull sales case studies are fairly obvious to the insider, the hidden gems are buried in chapters where one wouldn't expect AI to be injected.
Nasa, the civilian space programme, is one of such examples. It will launch it next Mars mission next year and faces the problem of the limited bandwidth available for sending information back to Earth.
Another issue is the limitation in power - and where running out of it at the wrong point could be detrimental to the mission. The use of AI is as complex as it is exciting. A vast set of sensors collects data and AI is trained to use that data to 'recognise anomalous' data that could help avoid peril.
So far and prior to the start of the mission, real results are sparse of course - it will be the first rover mission that relies on autonomous decision-making. Nonetheless, the merit seems promising. Thanks to the implementation of AI systems, information can be acted on almost instantaneously by the rover. It can make up its own mind on which locations should be investigated, reducing the risk to put it in jeopardy by needing to wait for instructions from the Earth (which can take time).
Harley Davidson is another example that may surprise some readers. "When you think of Harley Davidson, you may not initially think of high-tech business", Marr writes. Bikes would still be individually assembled on the shop floor. So how can AI play a part here? Marketing experts know that half of the marking budget is wasted. They just don't know which half. With AI, The company's AI-driven customer targeting increased lead generation by 2,930 per cent and as a result, helped to grow sales.
Geographically, AI is also used differently, sometimes even to a greater good, for the sick - and not only to further stuff the wallets of large corporations.
For example, the Chinese company Infervision uses AI to detect lung cancer, the country's most deadly disease. When AI spots such a cancer early on, survival rates are much higher.
Demand for such tech is high and has won (or will soon perhaps win) the hearts of the communist party. At present, it analyses 20,000 health scans every day. Now the company moves to new areas. Next station is stroke detection and other types of cancers.
Within the engineering industry, Siemens, the German industrial conglomerate that sells transport machinery as part of its portfolio, presents an inspiring example. Siemens is building the concept of "Internet of Trains" as Marr calls it. Train delays are not only annoying but expensive for businesses when their employees are not where they are supposed to be on time.
Sensors and cameras across the entire transport system would feed AI data and allow a 'digital twin' model of a railway system to be used to forecast factors likely to lead to delays or inefficiencies. It then helps to advise what can be done to react or prevent delays occurring in the first place.
Within digital technology, one laudable example is described in the chapter on Facebook-owned social media platform Instagram to battle cyber-bullying, a problem in the UK where nearly a quarter (23 per cent) of Brits reported that they have experienced it, according to a YouGov poll. This sounds great.
The only problem is that the lack of results gives a different impression. "The company hasn't spoken about the results it has seen yet", Marr writes. Does this mean, the filters and alerts may not work as may be envisioned?
This is unfortunate and maybe one sign that more scrutiny and criticism is needed. Little of it is Marr's fault. He just points out where he found good examples. But their presentation appears biased. Only presenting successful examples may miss important lessons learned in areas where AI failed.
AI is having its own battle to fight with criticism of bias. Used in law enforcement and government services, it can impose a racial inequality on the public, as we have seen in examples outside of Marr's book.
Future examples will need to come to terms with it and describe and share information on how data and sampling would need to be altered to help remove those ethical concerns. One should not forget that it is early days and that AI is just at the very starting point of its own evolution.
What Marr's writing makes so exceptional is his ability to break down in a few words what really matters to a non-technical audience. As a reviewer with a bird's-eye perspective - it is not clear if he himself fully understands the inner workings of machine learning and AI or whether he ever built an artificial intelligence algorithm himself - a lack of technical know-how doesn't harm his writing.
An average reader, whatever industry he or she is in, may quickly put down a book if it features terms more familiar to graduates or research scientists. Marr knows this.
Whether "AI is the most powerful technology available to mankind today...", as Marr claims will remain to be seen. Critics may say that there wouldn't be any benefits from AI if the infrastructure and a whole lot of smart and innovative minds could work day and night to find that very piece of the puzzle that fits perfectly into the picture of their problems.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.