This is today's edition of The Download, our weekday newsletter that provides a daily dose of what's going on in the world of technology.
Kids Outlearn AI—and We Still Don't Know Why
Teaching a computer to use human language requires an inhuman amount of data. A large language model (LLM) can easily process a hundred thousand times more words than a person will encounter while mastering their native tongue.
This stark divide between children and machines is known as the data efficiency gap. It raises a compelling question for cognitive scientists and a significant challenge for AI model architects: How can children still outperform the most linguistically sophisticated machines ever built?
Children demonstrate that learning more with less—far less—is possible. By reverse-engineering how they learn, scientists aim to create more data-efficient AI models and, in the process, settle enduring questions about language and the developing mind.
Here's how researchers are trying to close the data gap.
—Elise Cutts
Space Travel Agents: The Next Frontier in Autonomous Exploration
As humanity pushes further into the cosmos, a new kind of explorer is emerging: AI-powered space travel agents. These autonomous systems are designed to navigate spacecraft, conduct scientific experiments, and make split-second decisions without human intervention. In 2026, several missions are testing these agents, which could revolutionize how we explore distant planets and moons.
Unlike traditional spacecraft, which rely on pre-programmed instructions and constant ground control, these agents use machine learning to adapt to unexpected conditions. For example, if a rover encounters an unforeseen obstacle, it can analyze the terrain and reroute itself—saving precious time and enabling more ambitious missions.
Experts believe these agents could be key to deep-space exploration, where communication delays make real-time control impractical. By 2030, we may see fully autonomous missions to the outer solar system, guided by AI that learns and evolves during the journey.
—Thomas Macaulay
Plus: A New Amazon Data Center Could Become the US's Most Polluting Power Plant
In other news, Amazon's proposed data center in the United States could become the country's most polluting power plant if built as planned. This revelation comes amid growing concerns about the environmental impact of AI infrastructure. The data center, designed to support massive AI workloads, would rely on natural gas turbines, raising alarms among environmentalists and policymakers.
This development is part of a broader trend: both political parties are turning against AI data centers ahead of the midterm elections. Lawmakers are increasingly questioning the trade-off between technological advancement and environmental sustainability.
As the demand for AI services skyrockets, the tech industry faces a critical challenge: how to power the future without polluting it. Innovations in renewable energy, battery storage, and nuclear power may offer solutions, but the clock is ticking. In 2026, the debate over AI's carbon footprint is more urgent than ever.
—Thomas Macaulay
Stay Updated with The Download
For more daily insights into the world of technology, subscribe to The Download newsletter. Get the latest news, analysis, and trends delivered straight to your inbox every weekday.