A few things I learned from James Manyika's "Getting AI Right: A 2050 Thought Experiment."

Let's say it’s 2050.

You wake up, pour your coffee, and scan the headlines. Nowhere does it say, “AI: Mission Accomplished.” There’s no triumphant moment, no collective applause. Instead, progress appears in subtler forms: ethical AI systems trusted across borders, medical breakthroughs saving millions, algorithms that quietly balance climate grids. This, James Manyika argues, is how the future unfolds, not in miraculous leaps, but in deliberate, cumulative steps.

I had the privilege of meeting James at a Chatham House event. Before I could fully engage him in conversation, Jay Stoll shared a fascinating anecdote. He recounted his experience as General Secretary of the LSE Students’ Union and his tenure on the board during deliberations about the future of the campus. At the time, decisions were being made about buildings that would take a decade or more to complete. Students felt ten years was an eternity. Yet here we are, nearly 15 years later, walking through those very halls.

It’s a reminder that the long term arrives faster than we think. The same is true for AI. What we choose to do today shapes 2050, and 2050 is much closer than it feels.

After reading James Manyika's thought-provoking publication, "Getting AI Right: A 2050 Thought Experiment," I felt compelled to reflect and share my thoughts.

1. Define what we truly want, not just what’s possible

The biggest challenge with AI isn’t the algorithms; it’s deciding what they’re for. This isn’t a coding problem, it’s a people problem. Humans are messy, full of contradictions. Progress for one group can mean devastation for another. Studies from the World Economic Forum show how wildly interpretations of “ethical AI” vary: what Silicon Valley calls innovation, parts of Europe might call exploitation.

Before AI can align with humanity, humanity must align with itself. This isn’t about finding consensus on every detail but establishing a common baseline for values. Otherwise, we risk building tools that reflect our divisions more than our aspirations.

2. Trust isn’t negotiable, it’s earned

The most damaging thing an AI can do isn’t to fail; it’s to fail confidently. We’ve already seen this with “hallucinations,” models outputting polished nonsense with absolute certainty. Trust isn’t built on promises of perfection; it’s built on transparency and reliability.

Here’s what matters: admitting where systems fall short and proving they’re getting better. The companies that thrive in the AI age won’t be the ones shouting the loudest; they’ll be the ones proving, day by day, that they’re worth trusting.

3. Regulate to accelerate, not restrain

The word “regulation” tends to divide a room. Some see it as the bureaucratic brake on progress; others as the scaffolding that keeps skyscrapers from collapsing. The European Union’s AI Act is a case study in thoughtful regulation: it doesn’t shackle innovation but creates guardrails so the benefits of AI don’t come with catastrophic costs.

Unchecked, AI development risks becoming a digital Wild West, exciting, but dangerously chaotic. Regulation isn’t the enemy of progress. It’s the price of longevity. Without it, AI risks moving too fast for its own good, or ours.

4. AI should be a partner, not a competitor

There’s a growing fear that AI will render humans obsolete, but the reality is more nuanced. McKinsey estimates that AI can automate 70% of data processing tasks but only 5% of those requiring creativity or judgment. That’s a clue, not a threat.

The best uses of AI won’t replace us; they’ll empower us. Machines will handle the repetitive grind so humans can focus on the creative, the empathetic, and the complex. The sweet spot for AI isn’t replacing humanity, it’s amplifying it.

5. Bet big on breakthroughs

When AlphaFold solved the protein-folding mystery that had stumped scientists for half a century, it wasn’t magic. It was the result of sustained investment, visionary collaboration, and a willingness to play the long game. These breakthroughs don’t happen by accident; they happen because someone, somewhere, decided to make a big bet.

The lesson? If we keep pushing resources into fields like climate tech, medicine, and materials science, the next AlphaFold moment won’t just be possible, it will be inevitable.

6. Redefine the human experience

When a machine can compose symphonies or diagnose cancer better than humans, it raises an existential question: what’s left for us? Reid Hoffman calls AI a “steam engine for the mind,” and the analogy is apt. Just as the steam engine freed humans from backbreaking labor, AI is poised to liberate us from cognitive monotony.

But it also challenges us to redefine what makes us human. Creativity, empathy, resilience, these aren’t just buzzwords; they’re our competitive edge. In a world where machines handle the mundane, our humanity becomes the differentiator.

7. Think in milestones, not miracles

Here’s the thing about progress: it doesn’t announce itself with fireworks. The UN’s Sustainable Development Goals aren’t achieved overnight; they’re built incrementally, step by step. AI will follow the same path. By 2050, we won’t celebrate a singular AI success. Instead, we’ll reflect on a series of victories: ethical frameworks, smarter healthcare, reduced inequality.

The future doesn’t arrive in leaps. It sneaks in through the side door, one milestone at a time.

8. We need to act while it’s still our choice

The scariest idea Manyika puts forward isn’t about what AI might become. It’s about what happens if we lose control of what we’re building. MIT Tech Review warns we’re in a “narrow window” where humans still shape AI more than it shapes us. That window is closing.

The takeaway in James' is simple; AI isn’t about what’s possible. It’s about what we choose to make possible.

For me, I am still curious about two things:

1. As AI reshapes what’s possible, are we prepared to redefine what it means to be human?
2. If AI can reflect our values perfectly, what happens if we don’t agree on what those values should be?

They say, “time will tell,” but I’d argue we have the power to tell time what to say. The future doesn’t unfold on its own, it listens to what we do today and echoes it back to us tomorrow.

Getting AI Right: A 2050 Thought Experiment by James Manyika