Sitemap
Open in app

Sign in

Medium Logo
Write
Search

Sign in

The Go Players Got Better After They Lost

What a 71-year dataset of professional Go, a 1930 Keynes essay, and Gallup’s engagement numbers tell us about finding purpose next to superhuman machines.

--

Press enter or click to view image in full size

In November 2019, Lee Se-dol, an 18-time world Go champion and one of the greatest players of the modern era, announced his retirement. He was 36. Three years earlier he had lost four of five games to DeepMind’s AlphaGo, and he had decided there was no longer a summit worth climbing. “I’m not at the top even if I become the number one,” he told South Korea’s Yonhap news agency, as the BBC reported. “There is an entity that cannot be defeated.”

It is the saddest sentence in the history of artificial intelligence, and also, it turns out, a misleading one.

Because here is what happened to everyone else who kept playing. A team of researchers analyzed more than 5.8 million moves by professional Go players across 71 years, from 1950 to 2021, using a superhuman AI to score the quality of each human decision. After the arrival of superhuman AI, human players “began to make significantly better decisions.” Novel moves — plays never seen before — “occurred more frequently and became associated with higher decision quality.” The machine didn’t end human Go. It broke the game’s orthodoxy open and made the humans more inventive.

Two responses to the same event. One champion concluded his work had lost its meaning. Thousands of other players found new meaning in the same board.

That split is the question every company is about to face. Here’s my thesis: AI won’t make human work meaningless, but it will make meaning expensive. When machines can do most tasks competently, the sense of purpose people used to get for free — as a byproduct of being needed — will have to be paid for, in deliberate effort, preserved friction, and jobs designed around ownership. Organizations that budget for that will keep their best people. The rest will be running very efficient operations staffed by people who’ve quietly checked out.

Keynes saw the real problem in 1930

The fear isn’t new; only the timeline is. In his 1930 essay “Economic Possibilities for our Grandchildren,” John Maynard Keynes predicted that within a century “the economic problem may be solved,” which meant it was “not — if we look into the future — the permanent problem of the human race.”

What worried him was what came next. Humanity’s “real, his permanent problem” would be “how to use his freedom from pressing economic cares.” He suggested we might “spread the bread thin on the butter,” with “three-hour shifts or a fifteen-hour week,” because “three hours a day is quite enough to satisfy the old Adam in most of us!” And he asked, with real dread: “must we not expect a general ‘nervous breakdown’?”

Keynes was wrong about the hours — we found infinite new wants to work for — but right about the shape of the risk. The danger of abundance isn’t poverty. It’s purposelessness. And the AI labs, to their credit, are now arguing about exactly that.

Three theories of meaning from the people building the machines

Read the major lab leaders’ manifestos side by side and you find three distinct theories of where human meaning will come from.

Dario Amodei: relationships, not labor. In “Machines of Loving Grace,” Anthropic’s CEO argues that “it is very likely a mistake to believe that tasks you undertake are meaningless simply because an AI could do them better.” His evidence is disarmingly simple: “Most people are not the best in the world at anything, and it doesn’t seem to bother them particularly much.” Meaning, he writes, “comes mostly from human relationships and connection, not from economic labor.” Then comes his most revealing line: “The economic piece actually seems more difficult to me than the meaning piece.”

Mark Zuckerberg: individual aspiration. Meta’s superintelligence letter frames the question as one of agency. It distinguishes Meta’s vision from “others in the industry who believe superintelligence should be directed centrally towards automating all valuable work, and then humanity will live on a dole of its output.” Zuckerberg’s historical analogy: “As recently as 200 years ago, 90% of people were farmers growing food to survive.” Technology freed us for “the pursuits we choose.”

Mustafa Suleyman: deliberate limits. Microsoft AI’s chief, in “Towards Humanist Superintelligence,” argues for AI that is “domain specific,” with “real restrictions on autonomy,” designed to “keep humanity in control” and “leave precious space for human flourishing, for us to keep improving, engaging and trying, as we always have.” He’s not shy about capability, either: his team’s diagnostic orchestrator, he writes, reached 85 percent on the New England Journal of Medicine’s notoriously hard case challenges, where “human doctors max out at about 20%.”

Notice the progression. Amodei says meaning will take care of itself. Zuckerberg says meaning comes from choosing. Suleyman says meaning needs room deliberately left open. Only the last one treats meaning as something you have to design for. I think he’s closest to right — and that it applies to companies at least as much as to models.

The non-obvious part: the meaning crisis already happened

Here’s the counterintuitive fact that reframes the whole debate. If you’re worried AI will drain the meaning from work, I have news: for most workers, it’s already gone.

Gallup’s latest global workplace report found that just 20 percent of employees worldwide were engaged in 2025, down from 23 percent in 2022 and 2023. Gallup estimates that low engagement cost the world economy approximately $10 trillion in lost productivity last year, or 9 percent of GDP. The steepest declines were among managers.

And now AI anxiety is layering on top. In Gallup’s U.S. survey from the first quarter of 2026, 18 percent of employees said it was likely their job would be eliminated within five years by automation or AI, up from 15 percent. In organizations that have implemented AI, the figure was 23 percent. In finance and insurance, it was 32 percent — nearly a third.

So the “meaning of work” was never a stable asset that AI threatens to destroy. It was a depreciating one that most organizations had already written down. That cuts both ways. The bad news: there’s not much cushion. The good news: the bar for doing better is on the floor.

How meaning actually gets made

If meaning at work isn’t a byproduct of being needed, what is it? The research points to an unglamorous answer: effort that leads to completion.

Behavioral scientists Michael Norton, Daniel Mochon, and Dan Ariely named it the IKEA effect: people place disproportionately high value on things they’ve labored to create — but the effect depends on actually finishing. Build half the bookshelf and the attachment evaporates. Labor plus completion equals love.

Now consider what generative AI does by default. It removes the labor and delivers the completion. That’s wonderful for throughput and corrosive for ownership. Two education researchers, speaking at a Harvard Education Press event covered by the Harvard Gazette this spring, have a name for the worst version. “They’re ‘never-skilling,’” said Stephanie Smith Budhai of the University of Delaware, “which is even scarier than ‘deskilling.’” Her co-author Marie Heath of Loyola University Maryland went further: “For democracy to function, we need to be able to sit in discomfort… One of the things that we give up when we turn to this technology is the ability to sit in discomfort and practice being uncomfortable.”

Harvard’s Christopher Dede offers the most useful image I’ve found for the alternative. Athena, he told the Gazette, is always portrayed with an owl on her shoulder. “Can AI be like the owl that helps us be wiser?” The key, he argues, “is not to let it do your thinking for you.” His colleague, the philosopher Jeff Behrends, is blunter: “I am very worried about the effects of general-use LLMs on critical reasoning skills.” The Gazette, notably, flags that the most-cited recent study on this — MIT’s small EEG experiment on essay writing with ChatGPT — is not peer-reviewed. The evidence on cognitive harm is early. The design question isn’t.

Get Brian Nuesi’s stories in your inbox

Join Medium for free to get updates from this writer.

Go back to the Go players. What separated them from Lee Se-dol wasn’t talent. It was that they kept playing — kept doing the effortful thing, now with a superhuman sparring partner — and the effort, redirected, produced novelty. The machine raised the standard; the humans rose to meet it. That’s the version of AI that creates meaning. It requires people to stay in the game.

The skeptic’s case, honestly stated

The best counterargument is Amodei’s, and it deserves a fair hearing. Maybe I’m overrating work as a source of meaning. Most people already find their deepest purpose in family, friends, faith, community, and play — none of which AI threatens. Amodei notes that he enjoys getting better at a video game and doesn’t care that someone is far better. If AI handles the economic labor, people will simply relocate meaning to where it always mostly lived. On this view, “meaning gets expensive” is a knowledge worker projecting his own identity crisis onto everyone else.

I’d concede more than half of that. Work is not the only, or even the primary, source of meaning for most people. But two things keep me from conceding the rest. First, Amodei himself admits the economic piece is harder — and economics and meaning aren’t separable for people whose sense of competence, status, and provision is tied to a paycheck. Gallup’s 32 percent in finance and insurance aren’t worried about their hobbies. Second, his examples — the video game, the mountain climb — are chosen difficulty. That’s exactly my point. In an AI world, meaningful effort doesn’t disappear; it becomes optional. And optional things are the first to get cut when a quarter’s numbers are tight.

What leaders should do with this

Budget friction on purpose. Identify the tasks where struggle builds skill and judgment — first drafts of analysis, early-career modeling, writing the investment memo — and protect some of them from full automation, especially for juniors. That’s not inefficiency. It’s training spend.

Assign ownership, not tasks. The IKEA effect needs completion. Give people end-to-end responsibility for outcomes that AI helps with, rather than fragments that AI stitches together.

Make managers the meaning layer. Gallup’s steepest engagement drop is among managers. If they’re disengaged, no AI rollout will feel like anything other than a threat.

Redefine “good” upward. The Go lesson: when the machine raises the baseline, raise the human standard with it. Ask people to do what the AI’s output can’t — the novel move, the judgment call, the relationship.

For finance readers, there’s a hard-numbers version of this argument. If low engagement is already a $10 trillion drag, then even a modest improvement in how people experience AI-augmented work is worth more than most efficiency gains on the slide deck. The productivity case for AI and the meaning case for humans are not opposites. They are the same P&L line.

Bottom line

Lee Se-dol looked at a machine that couldn’t be beaten and concluded there was nothing left to play for. The players who stayed discovered moves no human had ever made.

The machines are getting competent at almost everything. That doesn’t make human effort worthless. It makes it a choice — and choices, unlike byproducts, have to be paid for. The companies that pay will get the novel moves.

Enjoying this? Subscribe free and get the next breakdown in your inbox tomorrow morning.

If you want the full map of how shifts like this move through business models and balance sheets — how tokenization, AI, and agentic systems are rewriting who captures value — I wrote it for you. AI Tokenization Revolution: Navigating Business is the playbook: https://www.amazon.com/AI-Tokenization-Revolution-Navigating-Business/dp/B0GW5CSHK6

And if you want these breakdowns every week as the story unfolds — the moves, the numbers, the second-order effects — join me on Agentic Tomorrow, my podcast on the forces reshaping business and finance: https://open.spotify.com/show/3PcECfoWuocMOUpIxyGJKH

And to watch these stories come to life as video explainers, subscribe to my YouTube channel: https://www.youtube.com/@Brianspod

If this is your beat, follow along — a new piece here every morning.

Sources

This story was written with AI assistance and edited by the author.

Meaningful Work
Human Agency
Future Of Work
Ai And Society
Philosophy

--

--

Brian Nuesi
Brian Nuesi

Written by Brian Nuesi

20 followers
101 following

Dominican-American polymath bridging Hollywood, global finance & AI. Executive Director at The Carlton Group, founder of Aggregate+, host of Agentic Tomorrow.