Why?

By: John F Groom

We live in an age obsessed with optimization. We optimize our investment portfolios. We optimize our diets. We compare sleep scores and heart rates. We optimize driving routes to save four minutes. We read reviews before choosing restaurants, hotels, movies, headphones, and coffee makers. We compare mortgage rates, airline seats, and mobile phone plans. Businesses optimize advertising conversions, websites, supply chains, and employee productivity.

Artificial intelligence is about to make us dramatically better at all of it. But there is a problem. Optimization requires an objective. And we spend remarkably little time deciding what the objective actually is.

How Much Time Do You Spend Asking Why?

Consider a simple question. Have you spent as much time thinking seriously about what your life is for as you have reading restaurant reviews? Not today. Over your lifetime. Add up the hours spent deciding which restaurants to visit, which television shows to watch, which hotel to book, which phone to buy, which car to drive, which investment fund performed better, which diet is optimal, and which exercise produces slightly better results.

Now compare that with the time spent seriously considering a much larger question: What am I trying to accomplish with my life? For many intelligent and sophisticated people, the comparison may be surprisingly uncomfortable. This is not because people are stupid. In some ways, it is precisely because modern people have become so good at solving smaller problems. There is an enormous infrastructure for answering how. There is much less infrastructure for answering why.

More Is Not an Objective

Money provides an unusually clear example. Ask someone what his financial objective is and the answer will often amount to some version of "more." More income, more savings, more investments, higher returns, greater net worth.

Financial professionals can then construct extraordinarily sophisticated systems around that assumption. They can optimize asset allocation, taxes, diversification, volatility, insurance, and retirement withdrawals. They can measure risk to several decimal places. But there is a much simpler question that should come first: What is the money for? And then: How much is enough to accomplish that purpose?

Suppose, as a thought experiment, that someone knew with certainty that having $1 million one year from now would completely satisfy every financial objective he had for himself and everyone he cared about. Suppose further that additional money beyond $1 million would provide absolutely no additional benefit. His rational behavior would look very strange by conventional financial standards.

Before reaching the $1 million, taking substantial risk might make sense because reaching the objective has enormous value. Once he has the $1 million, the calculation reverses. If another dollar provides no benefit, risking the money he already has in pursuit of additional wealth makes little sense.

The objective changes the meaning of risk. Yet sophisticated people routinely spend decades accumulating wealth without ever establishing what the wealth is supposed to accomplish. If the objective is simply "more," success becomes impossible by definition. There is always more.

Efficient for What?

The same problem appears when we talk about efficiency. Consider a hummingbird weighing only a few grams. Almost every gram appears extraordinarily functional. The animal contains propulsion, navigation, vision, a brain, circulation, digestion, temperature regulation, reproduction, and the ability to find its own energy supply. It seems like an almost perfect example of efficiency. Now consider a 300-pound professional football lineman.

For most ordinary human activities, carrying 300 pounds is inefficient. Walking requires more energy. Running is harder. Air travel is less comfortable. The additional mass can place greater stress on the body. But the lineman does something unusual for a living. For a small fraction of his life, another enormous and exceptionally powerful human being tries to push through him.

Suddenly, those 300 pounds look different. Losing 60 pounds might improve his efficiency during most of his day while making him substantially worse at the activity that produces his livelihood. The contradiction disappears once we understand the purpose. Nothing is efficient or inefficient independently of purpose.

The Missing Objective Function

This becomes especially important as artificial intelligence improves. AI is extraordinarily good at helping answer questions that have an objective. How can we reduce shipping costs? How can we identify cancer in this image? How can we write this software faster? How can we increase advertising conversions? How can we find the fastest route between two locations? How can we structure this investment portfolio to produce a particular combination of expected return and volatility?

These are extraordinarily valuable capabilities. But AI's ability to optimize makes the quality of the objective increasingly important. Imagine telling an extremely powerful AI system to optimize your life without telling it what your life is supposed to accomplish.

What should it maximize? Income? Longevity? Leisure? Social status? Number of friends? Children? Physical fitness? Pleasure? Professional accomplishment? Security? Knowledge? Influence? Service to others? There is no purely technical answer. And optimizing one variable can easily damage another.

A person can maximize income by sacrificing relationships. He can maximize career achievement by sacrificing health. He can maximize safety by avoiding opportunities. He can maximize longevity in ways that make the additional years less enjoyable. He can accumulate wealth long after additional wealth has ceased contributing meaningfully to anything he actually values. The optimization can be technically excellent while the result is completely wrong.

The Old Answers

For much of human history, individuals did not face these questions entirely alone. Religion supplied answers about ultimate purpose, morality, and man's place in the universe. Family supplied obligations, continuity, and identity. Communities supplied roles, relationships, and expectations. Tradition told people what constituted a respectable or successful life.

These systems were far from perfect. They could be restrictive, coercive, and hostile to people who wanted to live differently. The expansion of individual freedom has therefore been an enormous achievement.

But freedom came with a less frequently discussed consequence. When inherited structures weaken, their answers weaken with them. The individual gains greater freedom to decide what his life should be. He also acquires the burden of actually deciding. Nothing automatically fills the void.

Optimization Without Direction

This may be one of the great contradictions of modern technological civilization. We have become exceptionally good at optimizing small things while often remaining remarkably vague about the largest thing being optimized: a human life.

We know how many steps we took yesterday but may not know where we are trying to go. We know whether our investment portfolio beat its benchmark but may never have determined what the money is ultimately supposed to accomplish. We know whether we slept efficiently but may not have decided what we want to be awake for.

We know which restaurant within five miles has a 4.8 rather than a 4.6 rating. But we may never have spent an equivalent amount of time asking what would constitute a successful life. Technology did not create this problem. But AI may expose it.

AI Makes "Why?" More Important

Every major technological advance changes the relative scarcity of different capabilities. Computers made calculation abundant. The Internet made information abundant. Smartphones made communication abundant. AI is beginning to make certain forms of intelligence, analysis, and optimization abundant.

That is extraordinary. But as how becomes cheaper, why becomes relatively more valuable. If it once took 100 people to execute an idea and AI allows one person to execute it, the scarcity moves toward choosing the idea.

If AI can analyze thousands of investment strategies, the important question increasingly becomes what the investment is intended to accomplish. If AI can make an organization extraordinarily efficient, the question becomes whether the organization is doing something worth doing.

And if AI can help an individual optimize hundreds of aspects of daily life, the question becomes what all that optimization is ultimately intended to serve.Without an answer, technology can produce a peculiar result:We become much more efficient at going nowhere in particular.

Purpose Before Optimization

This does not mean everyone needs a grand philosophical revelation. A person's purpose may be complicated. It may change. It may include raising children, creating a business, caring for someone, making art, accumulating knowledge, helping a community, exploring the world, or simply constructing a satisfying everyday life.

There may be several purposes rather than one. And a 25-year-old should not necessarily expect to give the same answer as the person he becomes at 65. The important thing is not that everyone arrive at the same answer. It is that we recognize the hierarchy.

Purpose comes before optimization. Before asking how to become wealthier, ask what the wealth is for. Before maximizing health, ask what the health enables. Before maximizing productivity, ask what deserves to be produced.

Before optimizing relationships, ask which relationships matter and why. Before making something faster, ask where you are trying to get. AI will become increasingly capable of helping us with almost everything that comes after those questions.

It may also help us examine the questions themselves. It can expose inconsistencies, identify tradeoffs, model possible futures, and remind us when our behavior has drifted away from our stated objectives.

But greater intelligence does not eliminate the need for purpose.

It makes purpose more important. The technological question of our age may therefore turn out not to be whether artificial intelligence becomes capable of answering increasingly complicated questions. It almost certainly will. The more important question may be whether humans become better at asking the simplest one. Why?