When the details get cheaper, think bigger
AI could let more people act on worthwhile ideas. That asks us to think more ambitiously about the goals, and carefully about the results.
AI, Actually
A closer look
When AI explains itself, what are we actually seeing?
A clear explanation can help us check an answer. Understanding what produced that answer is a harder—and important—question.
AI brought a suitcase full of mathematical keys. Which doors do they open?
OpenAI released hundreds of mathematical manuscripts. A small prime-number puzzle helps explain the promise—and what it takes to turn an AI's argument into trusted knowledge.
When does promising medical AI become better care?
A good prediction is a beginning. Clinical benefit depends on whether a tool works across settings, changes decisions safely, and improves outcomes patients actually experience.
The most useful AI skill may be knowing when not to answer
An AI system does not become trustworthy simply because it can produce an answer. In many settings, it also needs a way to show uncertainty, stop, and hand responsibility to a person.
When a data center needs the grid to grow, who pays?
A new House bill puts a familiar utility question at the center of the AI build-out: which costs belong to a huge new customer, and which benefits belong to everyone?
Data centers: separating the claims from the local reality
Electricity, water, jobs and neighborhood impacts—and the questions that make the debate more useful.
Data centers: what the numbers actually tell us.
Rising demand is real. So is the difference between a global percentage and the impact on one community.