In a speech to the Economic Club of Washington in 2018, Jeff Bezos described how Amazon made sense of the challenge of if and how to design and implement a loyalty scheme for its customers. This was a highly consequential decision for the business; for some time, Amazon had been searching for an answer to the question: “what would loyalty program for Amazon look like?”
A junior software engineer came up with the idea of fast, free shipping. But a big problem was that shipping is expensive. Also, customers like free shipping, so much so that the big eaters at Amazon’s “buffet” would take advantage by free shipping low-cost items which would not be good for Amazon’s bottom-line. When the Amazon finance team modelled the idea of fast, free shipping the results “didn’t look pretty.” In fact, they were nothing short of “horrifying.”
But Bezos is experienced enough to know that some of his best decisions have been made with “guts… not analysis.” In deciding whether to go with Amazon Prime, the analysts’ data could only take the problem so far towards being solved. Bezos decided to go with his gut. Prime was launched in 2005. It has become one of the world’s most popular subscription services with over 100 million members who spend on average $1400 per year compared to $600 for non-prime members.
As a seasoned executive and experienced entrepreneur Bezos sensed that the Prime idea could work. And in his speech he reminded his audience that “if you can make a decision with analysis, you should do so. But it turns out in life that your most important decisions are always made with instinct and intuition, taste, heart.”
The launch of Amazon Prime is a prime example of a CEO’s informed and intelligent use of intuition paying off in decision-making under uncertainty (where outcomes are unknown and their likelihood of occurrence cannot be estimated) rather than under risk (where outcomes are known and probabilities can be estimated). The customer loyalty problem for Amazon was uncertain because probabilities and consequences could not be known at the time. No amount of analysis could reduce the fast, free shipping solution to the odds of success or failure.
Under these uncertain circumstances Bezos chose to go with this gut. This is not an uncommon CEO predicament or response. In business, decision-makers often have to act “instinctively” even though they have no way of knowing what the outcome is likely to be. The world is becoming more, not less uncertain, and “radical uncertainty” seems to have become the norm for strategic decision-making both in business and in politics. The informed and intelligent use of intuition on the part of those who have the nous and experience to be able to go with their gut is one way forward.
Human intuition meets AI
Turning to the uncertainties posed by artificial intelligence and winding the clock back to over half-a-century ago, the psychologist Paul Meehl in his book Clinical Versus Statistical Prediction (1954) compared how well the subjective predictions of trained clinicians such as physicians, psychologists, and counsellors fared when compared with predictions based on simple statistical algorithms. To many people’s surprise, Meehl found that experts’ accuracy of prediction, for example trained counsellors’ predictions of college grades, was either matched or exceeded by the algorithm.
The decision-making landscape that Meehl studied all those years ago has been transformed radically by the technological revolutions of the “Information Age” (see Jay Liebowitz, Bursting the Big Data Bubble, 2014). Computers have exceeded immeasurably the human brain’s computational capacity. Big data, data analytics, machine learning, and artificial intelligence (AI) have been described as “the new oil” (see Eugene Sadler-Smith, “Researching Intuition: A Curious Passion” in Bursting the Big Data Bubble, 2014). They have opened-up possibilities for outsourcing to machines many of the tasks that were until recently the exclusive preserve of humans. The influence of AI and machine learning is extending beyond relatively routine and sometimes mundane tasks such as cashiering in supermarkets. AI now figures prominently behind the scenes in things as diverse as social media feeds, the design of smart cars, and on-line advertising. It has extended its reach into complex professional areas such as medical diagnoses, investment banking, business consulting, script writing for advertisements, and management education (see Marcus du Sautoy, The Creativity Code, 2019).
There is nothing new in machines replacing humans: they did so in the mechanisations of the agricultural and industrial revolutions when they replaced dirty and dangerous work; dull work and decision-making work might be next. Daniel Suskind, author of World without Work thinks the current technological revolution is on a scale which is hitherto unheard of. The power with which robots and computers are able to perform tasks at high speed, with high accuracy, at scale using computational capabilities are orders of magnitude greater than those of any human or previous technology. This one reason this revolution is different and is why it has been referred to as nothing less than the “biggest event in human history” by Stuart Russell, founder of the Centre for Human-Compatible Artificial Intelligence at the University of California, Berkeley.
The widespread availability of data, along with cheap, scalable computational power, and rapid and on-going developments of new AI techniques such as machine learning and deep learning have meant that AI has become a powerful tool in business management (see Gijs Overgoor, et al). For example, the financial services industry deals with high-stakes, complex problems involving large numbers of interacting variables. It has developed AI that can be used to identify cybercrime schemes such as money laundering, fraud and ATM hacking. By using complex algorithms, the latest generation of AI can uncover fraudulent activity that is hidden amongst millions of innocent transactions and alert human analysts with easily digestible, traceable, and logged data to help them to decide, using human intuition based on their “feet on the ground” experiences, on whether activity is suspicious or not and take the appropriate action. This is just one example, and there are very few areas of business which are likely to be exempt from AI’s influence. Taking this to its ultimate conclusion Elon Musk said at the recent UK “AI Safety Summit” held at Bletchley Park (where Alan Turing worked as code breaker in World War 2) that: “There will come a point where no job is needed—you can have a job if you want one for personal satisfaction but AI will do everything. It’s both good and bad—one of the challenges in the future will be how do we find meaning in life.”
Creativity and AI
Creativity is increasingly and vitally important in many aspects of business management. It is perhaps one area in which we might assume that humans will always have the edge. However, creative industries, such as advertising, are using AI for idea generation. The car manufacturer Lexus used IBM’s Watson AI to write the “world’s most intuitive car ad” for a new model, the strap line for which is “The new Lexus ES. Driven by intuition.” The aim was to use a computer to write the ad script for what Lexus claimed to be “the most intuitive car in the world”. To do so Watson was programmed to analyse 15 years-worth of award-winning footage from the prestigious Cannes Lions international award for creativity using its “visual recognition” (which uses deep learning to analyse images of scenes, objects, faces, and other visual content), “tone analyser” (which interprets emotions and communication style in text), and “personality insights” (using data to make inferences about consumers’ personalities) applications. Watson AI helped to “re-write car advertising” by identifying the core elements of award-winning content that was both “emotionally intelligent” and “entertaining.” Watson literally wrote the script outline. It was then used by the creative agency, producers, and directors to build an emotionally gripping advertisement.
Even though the Lexus-IBM collaboration reflects a breakthrough application of AI in the creative industries, IBM’s stated aim is not to attempt to “recreate the human mind but to inspire creativity and free-up time to spend thinking about the creative process.” The question of whether Watson’s advertisement is truly creative in the sense of being both novel and useful is open to question (it was based on rules derived from human works that were judged to be outstandingly creative by human judges at the Cannes festival). In a recent collaborative study between Harvard Business School and Boston Consulting Group, “humans plus AI” has been found to produce superior results compared to “humans without AI” when used to generate ideas by following rules created by humans. However, “creativity makes new rules, rules do not make creativity” (to paraphrase the French composer Claude Debussy). The use of generative AI which is rule-following rather than rule-making is likely to result in “creative” outputs which are homogeneous and which may ultimately fail the test of true creativity, i.e. both novel (in the actual sense of the word) and useful. Human creative intuition on the other hand adds value by:
- going beyond conventional design processes and rules
- drawing on human beings’ ability to think outside the box, produce innovative solutions
- sensing what will or won’t work
- yielding products and services that stand out in market, capture the attention of consumers, and drive business success.
—as based on suggestions offered by Chat GPT in response to the question: “how does creative intuition add value to organizations?”
Emotion intelligence and AI
Another example of area in which fourth generation AI is making in-roads is in the emotional and inter-personal domains. The US-based start-up Luka has developed the artificially intelligent journaling chatbot “Replika” which is designed to encourage people to “open-up and talk about their day.” Whilst Siri and Alexa are an emotionally “cold” digital assistants, Replika is designed to be more like your “best friend.” It injects emotion into conversations and learns from the user’s questions and answers. It’s early days, and despite the hype rigorous research is required to evaluate the claims being made on behalf of such applications.
“The fact that computers are making inroads into areas that were once considered uniquely human is nothing new.”
The fact that computers are making inroads into areas that were once considered uniquely human is nothing new. Perhaps intuition is next. The roots of modern intuition research are in chess, an area of human expertise in which grand masters intuit “the good move straight away.” Nobel laureate and one of the founding figures of AI, Herbert Simon, based his classic definition of intuition (“analyses frozen into habit and the capacity for rapid response through recognition”) on his research into expertise in chess. He estimated that grandmasters have stored of the order of 50,000 “familiar patterns” in their long-term memories, the recognition and recall of which enables them to play chess intuitively at the chess board.
In 1997, the chess establishment was astonished when IBM’s Deep Blue beat Russian chess grand master and world champion Garry Kasparov. Does this mean that IBM’s AI is able to out-intuit a human chess master? Kasparov thinks not. The strategy that Deep Blue used to beat Kasparov was fundamentally different from how another human being might have attempted to do so. Deep Blue did not beat Kasparov by replicating or mimicking his thinking processes, in Kasparov’s own words:
“instead of a computer that thought and played like a chess champion, with human creativity and intuition, they [the ‘AI crowd’] got one that played like a machine, systematically, evaluating 200million chess moves on the chess board per second and winning with brute number-crunching force.”
Nobel laureate in physics, Richard Feynman, commented presciently in 1985 that it will be possible to develop a machine which can surpass nature’s abilities but without imitating nature. If a computer ever becomes capable of out-intuiting a human it is likely that the rules that the computer relies on will be fundamentally different to those used by humans and the mode of reasoning will be very different to that which evolved in the human organism over many hundreds of millennia (see Gerd Gigerenzer, Gut Feelings, 2007).
In spite of the current hype, AI can also be surprisingly ineffective. Significant problems with autonomous driving vehicles have been encountered and are well documented, as in the recent case which came to court in Arizona involving a fatality allegedly caused by an Uber self-driving car. In medical diagnoses, even though the freckle-analysing system developed at Stanford University does not replicate how doctors exercise their intuitive judgement through “gut feel” for skin diseases, it can nonetheless through its prodigious number-crunching power diagnose skin cancer without knowing anything at all about dermatology (see Daniel Susskind, A World Without Work, 2020). But as the eminent computer scientist Stuart Russell remarked, the deep learning that such AI systems rely on can be quite difficult to get right, for example some of the “algorithms that have learned to recognise cancerous skin lesions, turn out to completely fail if you rotate the photograph by 45 degrees [which] doesn’t instil a lot of confidence in this technology.”
Is the balance of how we comprehend situations and take business decisions shifting inexorably away from humans and in favour of machines? Is “artificial intuition” inevitable and will it herald the demise of “human intuition”? If an artificial intuition is realized eventually that can match that of a human, it will be one of the pivotal outcomes of the fourth industrial revolution―perhaps the ultimate form of AI.
Chat GPT appears to be “aware” of its own limitations in this regard. In response to the question “Dear ChatGPT: What happens when you intuit?” it replied:
“As a language model I don’t have the ability to intuit. I am a machine learning based algorithm that is designed to understand and generate human language. I can understand and process information provided to me, but I don’t have the ability to have intuition or feelings.”
More apocalyptically, could the creation of an artificial intuition be the “canary in the coalmine,” signalling the emergence of Vernor Vinge’s “technological singularity” where large computer networks and their users suddenly “wake up” as “superhumanly intelligent entities” as Musk and others are warning of? Could such a development turn out to be a Frankenstein’s monster with unknown but potentially negative, unintended consequences for its makers? The potential and the pitfalls of AI are firmly in the domain of the radically uncertain and identifying the potential outcomes and how to manage them is likely to involve a judicious mix of rational analysis and informed intuition on the part of political and business leaders.
“The potential and the pitfalls of AI are firmly in the domain of the radically uncertain.”
Human intuition, AI, and business management
Making any predictions about what computers will or will not be able to do in the future is a hostage to fortune. For the foreseeable future most managers will continue rely on their own rather than a computer’s intuitive judgements when taking both day-to-day and strategic decisions. Therefore, until a viable “artificial intuition” arrives that is capable of out-intuiting a human, the more pressing and practical question is “what value does human intuition add in business?” The technological advancements of the information age have endowed machines with the hard skill of “solving,” which far outstrips this capability in the human mind. The evolved capacities of the intuitive mind have endowed managers with the arguably hard-to-automate, or perhaps even impossible-to-automate, soft skill of “sensing.” This is the essence of human intuition.
Perhaps the answer lies in an “Augmented Intelligence Model (AIM),” which marries gut instinct with data and analytics. Such a model might combine three elements:
- human analytical intelligence, which is adept in communicating, diagnosing, evaluating, interpreting, etc.
- human intuitive intelligence, which is adept in creating, empathising, feeling, judging, relating, sensing, etc.
- artificial intelligence, which is adept in analysing, correlating, optimising, predicting, recognizing, text-mining, etc.
The most interesting spaces in this model are in the overlaps between the three intelligences, for example when human analytical intelligence augments artificial intelligence in a chatbot with human intervention. Similar overlaps exist for human analytical and human intuitive intelligences, and for human intuitive intelligence and artificial intelligence. The most interesting space is where all three overlap and it is here that most value stands to be added by leveraging the combined strengths of human intuitive intelligence, human analytical intelligence, and artificial intelligence in an Augmented Intelligence Model which can drive success.
This blog post is adapted from Chapter 1 of Intuition in Business by Eugene Sadler-Smith.