What Will Humans Still Be Better At?

The assumption underneath most conversations about AI is that every capability gained by a machine represents territory lost by a human.

That's historically strange. Calculators didn't make mathematics irrelevant. Cameras didn't eliminate painting. Search engines didn't eliminate expertise.

New capabilities change what we value.

AI will eliminate the scarcity of certain kinds of intelligence. The interesting question isn't whether humans can compete with machines at those tasks. It's what becomes scarce after intelligence itself becomes abundant.

My bet is judgment.

A robot demonstration can make the future look five minutes away.

Then you put the robot in a warehouse where somebody moved the pallet, the lighting changed, a box ripped open and Gary left a ladder where it absolutely should not be.

Welcome to reality.

Humans are phenomenal at operating in environments nobody bothered to make machine-readable.

Robotics will get there. But the last 10% of physical-world autonomy contains about 90% of the weirdness.

A system doesn't have to think like us to outperform us.

That's uncomfortable because we've spent centuries treating intelligence as evidence of human exceptionalism.

If intelligence stops being uniquely human, that doesn't necessarily diminish humanity.

It may force us to develop a more sophisticated definition of it.

They don't.

Humans routinely choose the slower restaurant, the imperfect handmade object, the inconvenient neighborhood and the objectively irrational hobby.

Optimization is useful. Meaning is different.

A future designed entirely around efficiency would be extraordinarily efficient—and probably unbearable.

Machines will become exceptional at answering questions. Humans retain an enormous role in deciding which questions deserve answers.

Particularly when the objective changes halfway through the task and nobody updated the instructions.

We don't merely calculate outcomes. We live inside them.

Every transformative company begins as something the available data cannot fully justify.

Sometimes the most important thing one person can provide another isn't information.

Reality produces edge cases faster than engineers can document them.

The leap from impressive demonstrations to reliable manipulation will happen faster than many people outside robotics expect.

Medicine contains enormous amounts of pattern recognition. Machines will increasingly see patterns no individual physician could.

We should stop pretending humans will retain a monopoly on beauty. The more interesting question is whether provenance changes how we value it.

One-to-one instruction has historically been expensive. AI radically changes those economics.

AI won't simply perform individual tasks. Increasingly capable systems will coordinate tasks, people, information and resources across organizations.

Eventually we stop talking about “AI products” for the same reason we stopped talking about “internet-enabled businesses.”

Intelligence becomes an expected layer of nearly every product and service.

The next enormous shift isn't another chatbot.

It's machine intelligence becoming capable of reliably manipulating the physical world.

Handmade. Live. Local. Human. Unrepeatable.

We will increasingly pay for evidence that something wasn't optimized.

When producing convincing information becomes essentially free, knowing who you trust becomes extraordinarily valuable.

The comparison eventually becomes as strange as asking whether an airplane is better than a bird.

The more important question will be what humans can accomplish with capabilities neither humans nor machines possessed independently.