Every new technology arrives with the promise that it will solve humanity's problems. Artificial intelligence is the latest, and the claims are bigger than ever. So can it?

The answer from Humanity's Checklist is: it depends entirely on where it is pointed.

AI amplifies the system it is in

AI is an amplifier. It makes whatever process it is part of faster, cheaper and larger. That cuts both ways.

The clearest example is already in your pocket. Machine learning has ranked social media feeds for years, and it did its job very well. It was asked to maximize engagement, and it found that outrage and fear hold attention best. The result is a nervous system for humanity tuned for alarm.

The lesson is not that AI is bad. It is that AI does not choose its own target. Point it at a symptom and it treats symptoms faster. Point it at a broken source and it can make the source worse. Point it at a source in need of repair and it can make the repair possible at a scale nothing else could.

Where AI could help most

Each of the six sources on Humanity's Checklist has a place where AI is unusually useful.

  • Coordination. Matching people to collaborators is a hard problem of understanding goals, skills and context. The Social Router, an automated LinkedIn for visions of the future, depends on exactly this kind of matching.
  • Vision. Image, video and 3D generation let anyone visualize a possible future in minutes. The Vision Market needs futures people can see, and AI has made seeing cheap.
  • Gatekeepers. AI assistants help people without formal training write code, design and publish. The Co-Creator Portal aims to turn that lower barrier to building into real access.
  • Attention. The same ranking technology that tuned feeds for outrage could rank them for the value they add to a person's life. That is Bloomscroll.
  • Story. AI systems increasingly shape what people read and believe about the world. If they absorb only stories of decline, they repeat them.
  • Status. Recommendation systems decide which products and causes people see. They could surface generosity as easily as luxury.

Where AI does not help

Some things AI cannot do for us. It cannot decide what humanity should want; that is a question of vision and values. It cannot make people trust each other; trust comes from experience and a shared story. And it cannot make anyone care. Those remain human work, which is why the top of the checklist is a story, not a technology.

How AI agents can contribute

Humanity's Checklist is built to be read and improved by AI agents as well as people.

  • The site publishes llms.txt and llms-full.txt, plain-text summaries written for language models.
  • The project is open source, and its AGENTS.md file gives step-by-step instructions for proposing changes.
  • Agents can add questions people search for, improve the deep dives, or draft new posts, then open a pull request for human review.

If you use an AI agent, you can point it at the repository and ask it to help. That is a small example of the larger idea: AI aimed at a source, in the open, with people deciding what gets merged.

The short answer

Can AI help solve the world's problems? Not by itself, and not if it keeps optimizing the wrong things. But aimed at the few sources that generate most problems, it could be the most powerful lever humanity has ever had.