Artificial intelligence: how its rise will change the investment landscape
While many investors understand that artificial intelligence is set to be a seismic development, few understand how it works or how it will affect their lives.
Unstructured Learning Time
You may have spoken to Amelia. You may have found her polite and helpful. But you won’t have met her. No one has.
This is because Amelia is a cognitive and emotive engine. “She” has been built for Nordic bank Skandinaviska Enskilda Banken (SEB) to handle customer calls. Amelia is projected to deal with as many as 60% of all telephone queries in her first three months of deployment; designed to learn and react to different emotions in real-time. She will only divert a call to a human employee if she does not know the answer to a question.
Source: SEB corporate website
The adoption of this type of artificial intelligence (AI) is cutting-edge just now, but is just one example of the countless ways AI will affect our lives in the future.
While it is natural for investors to ask which industries and jobs will be affected by AI, it seems more appropriate to ask which will not.
Welcome to the machine
Understanding neural networks is to understand AI. First described in 1943 by McCulloch and Pitts, an instructive example was developed in the 1950s by Frank Rosenblatt, tantalisingly named the “Perceptron”.
In the perceptron model, multiple, binary inputs are used to produce a single, binary output. The inputs could be anything. For example, questions such as “will an Intel processor run faster than a human brain in 2019?” or “did Google’s net income exceed £10 billion last year?”. Each input is then “weighted” according to user preference, and the product of these weights and inputs is measured against a “threshold” value to determine the output. This output might be a question such as “should I buy stock in Intel?”.
This is how things began. The next step, “learning algorithms”, allowed the machine to dynamically alter the weights until the correct output is found. It is the combination of this process, with multiple layer models, that underpins the basis of AI and what is known as “Deep Learning”.
AI models require human input and large datasets in order to learn. This process is computer-intensive and known as “training”. Once trained, the systems can begin to answer problems – a process known as “inference” – and generate solutions to both programmed and un-programmed questions.
An excellent example is IBM’s Watson for Oncology (Watson is the name of IBM’s AI engine). The system was trained by oncologists from Memorial Sloan Kettering, one of the world’s leading cancer hospitals, in combination with every academic paper and textbook on the subject, about 25 million in total. Capable of reading both structured and unstructured data – such as a trial database and a doctor’s handwritten note respectively - it also digests the additional 8,000 scientific papers written daily.
In a study of 1,000 cancer patients at the University of North Carolina, Watson recommended the same treatment as the doctors in 99% of instances. However, in 30%of instances, Watson was able to recommend something better based on papers or approvals the human counterparts had not yet had time to read.
AI should not be thought of as the end of human labour, at least not yet. AI can free up human effort for more productive, creative uses which, combined, could materially enhance corporate and social prospects. IBM does not see Watson as a replacement for oncologists, for example, simply as “augmented intelligence” enabling doctors to spend more time on care and research.
With sufficient data, successful early adopters of AI could enjoy competitive advantage based on lower costs, time to market and insight. New entrants may even emerge if an industry is not moving fast enough. However, sustainable competitive advantage may need to come from proprietary solutions developed in-house. If all solutions were merely bought from external vendors such as IBM or Google, speed of adoption would be the only differentiating factor. Internally developed or perhaps highly customised software may provide something more permanent. Data is scarce but Boeing, Toyota, Manulife and SEB (as discussed above) are amongst a growing number of non-tech companies experimenting on their own in AI.
A tool of our tools?
AI heralds a new era, as prior distinctions of human and computer expertise begin to fade. The impact will not be homogenous or necessarily detrimental. Early adopters can benefit where sufficient data already exists while some labour intensive business models face serious challenges.
The Bank of England’s Chief Economist, Andy Haldane, said the third machine age might result in 15 million UK job losses (50% of employment) and a further widening of inequality. His latter point is well taken, and government interference at the will of the people should not be discounted. However, technological progress has generally resulted in higher standards of living across the population and AI shows remarkable promise in fields such as terminal disease.
Likely to become one of the defining themes of our age, investors can ill afford to ignore its impact on any and every industry which they assess.