Artificial Intelligence: A Primer

Should we fear AI or embrace it? A dumb question, I know. Thinking we have a choice is as nonsensical as opposing the Internet 30 years ago. Just as the world-wide web changed the way we live, work and relate to each other, so too will AI. It will exponentially ramp up the speed of change, significantly increase productivity, dramatically reduce costs and add $16 trillion to the world economy within a decade.

The term AI was coined by John McCarthy in 1955. It refers to an intelligent program that thinks, learns (self improves) and mimics the cognitive functions of humans. It goes by several names – machine learning, automation, robotics, deep learning and natural language processing (think Siri and Alexa). Each term has a slightly different meaning, depending on whether the intelligence factor is “general” (can think like a human) or narrow (can’t transfer knowledge).

AI is fuelled by data – big, cheap and accessible data that doubles in volume every two years (IBM). It’s communicated ubiquitously via “the Internet of Things” and is stored in what we call “the cloud.” The world’s most valuable resource is no longer oil; it’s data.

Computers are good at pattern recognition and prediction. Machine learning forecasts future outcomes based on past experiences. Amazon predicts your next likely purchases using your previous selections as the rationale. It’s current estimates of your wants are about 5% effective. With time, they will get better. Much better. When returns cost less than shipments, Amazon will send you what you need before you know it. (If you doubt that, Amazon acquired a US patent for “anticipatory shopping” five years ago. It’s intentions were clear.)

Computers are now programmed to “program” human behaviours. Called persuasive computing, it can influence and manipulate opinions and behaviour to advance specific political agendas. As we’ve seen, it works (think the Russian incursion into the 2016 US election or the 2018 Facebook revelations about the use of personal data).

AI learning is progressive and life-altering. Computers are already proficient at picking stocks, translating speech, diagnosing cancer, detecting sarcasm, playing poker, writing music, cracking jokes and even assembling an IKEA chair. The tasks we find difficult are mere child’s play for AI and vice versa. An algorithm running on 16,000 processors taught itself to identify different breeds of cats after analyzing 10 million images. It’s accuracy was 75%. A toddler can do that with 100% accuracy on a walk to the playground. Humans are good at things like perception, mobility and empathy, computers are not.

AI does have limitations. It can solve “how” problems but can’t explain the what or why of its decisions. It’s based on logic and statistical significance, not truth – the “right” answers are those that occur most frequently. It can extrapolate from the past but is incapable of predicting the future with accuracy (of course, neither can humans). It is not creative, doesn’t possess common sense, nor can it speculate. It is a thinking machine incapable of making moral judgements.

It cannot stack up against adaptive biological intelligence (a computer cannot negotiate for example). It doesn’t possess social, emotional or strategic smarts, nor cognitive flexibility. It can learn, but cannot understand, reason, intuit, appreciate, feel the emotions (like love or anger) that guide decisions, know what is unknown, consider trade-offs, side-effects or unintended consequences. That said, AI programmers are now looking into ways to incorporate imagination and curiosity.

Restructuring business models and labour markets is inevitable but the scope and ultimate impact are unknown. A century ago, over 40% of Americans were farmers; today only 2% are. Both Canadian and US government reports say 40% of current jobs will be automated within 10 years. Newsweek recently reported that 90% of Americans believe half of today’s jobs will be lost to automation within five years. In classic human reasoning, more than 90% also believe theirs won’t be the ones axed. In other words, the very people whose jobs will disappear don’t know they’re in the crosshairs.

AI-driven robots demonstrably improve efficiency, reduce costs, improve quality and accuracy, and enhance the customer experience (all 24/7) without getting sick, being disgruntled or needing a paycheque. They can’t get unionized, they work incessantly and they don’t require breaks, vacations or maternity leaves. They are cost-effective and compliant, don’t ask for raises or gossip, and never lie, cheat or complain about their work circumstances.

The jobs most at risk are those that are manual, routine, repetitive or predictable. In other words, tasks that do not rely on critical thinking or creativity. This does not necessarily mean lower-payed workers. Deloitte (2018) estimates AI will automate 39% of jobs in the legal sector and that most accounting jobs “have a 95% chance of being replaced.” AI can now perform candidate searches and job matches better than HR professionals. Even essential services like health care will be impacted. In a recent experiment, a computer algorithm correctly diagnosed 90% of lung cancers while doctors had a success rate of 50%.

It is feared that the competitive consolidation of AI companies by five giant technology corporations (Alphabet/Google, Amazon, Facebook, Apple and Microsoft) will create “the worst monopoly in US history.” China, which has one-third of the planet’s total robot population, aims to become the world leader by 2030. AI will disrupt three industries in a major way: healthcare (e.g., by reducing adverse incidents causing death), finance (e.g., robo advisors) and transportation (e.g., human error causes 94% of the 1.3 million traffic deaths/year, so perhaps driving could become illegal). If you think that observation is a stretch, you’re not paying attention to the nanny state.

Are we preparing for an unprecedented disruption in how business is done? Hardly. While 85% of executives today believe AI will lead to sustainable competitive advantage, 95% of them are playing catch-up (Boston Consulting Group, 2018). Eliezer Yudkowsky, who writes about “friendly AI,” sums up the current challenge: “By far the greatest danger of AI is that people conclude too early that they understand it.” And the primary bottleneck to its implementation in the workplace is where it usually is – in management.

So what can you do as a smart leader to be better prepared? Do a SWOT analysis: list your strengths, weaknesses opportunities and threats in the context of these emerging learning machines that can out-think and outperform what you now do. Understand where the risks to your career are and where new opportunities may await. Think beyond your role and be a continuous learner.

Once AI becomes powered by quantum computing, it could be the last invention humans will ever make. Intelligence without conscience can be dangerous. If the intention of its designers is to serve us, it will. If the goal is to compete with humans and ultimately defeat them, it will. Machines are simply an extension of ourselves – our ideas, hopes and dreams. Therefore, it has the potential to be our greatest achievement or our worst nightmare.