Skip to main content
Interactive map showing 8,000 global castles and fortresses, with highlighted regions and detailed location pins.

Editorial illustration for The Download: An Interactive Map of 8,000 Global Castles and Fortresses

Interactive Map of 8,000 Global Castles & Fortresses

• 4 min read

A team of researchers has built an AI system that can look at a brain scan and reconstruct, with notable accuracy, the image a person was viewing when it was taken. Run the process backward, and the same tool can predict how someone's brain will respond to a given picture before they've even seen the full scan analyzed. That two-way trick is the headline item in today's edition of The Download, MIT Technology Review's weekday rundown on tech news.

The appeal is obvious. Scientists studying the brain get a new window into how visual processing actually works, and people who are locked-in, unable to speak or move, might eventually use something like this to communicate. There's even talk of reconstructing dreams. But the same capability that makes this useful also makes it unsettling: a method that can pull images out of someone's head doesn't need much tweaking before it's pulling out thoughts that were never meant to be shared, raising obvious consent problems.

Also in this edition: the Pentagon has brought Elon Musk on to help lead a project focused on future warfare, and a look at how small, unconventional batteries, tucked into things like induction stovetops and food carts, are finding a role in the grid that giant installations can't always fill.

A new AI tool can guess what you’re looking at just by analyzing your brain scans—and recreate that image with remarkable precision. It can go the other way too, and predict a person’s brain activity based on what they’re looking at.

Why this matters

The castle map is a nice diversion, but the real signal in this roundup is the brain-scan reconstruction tool sitting next to a headline admitting AI's "latest breakthroughs and fears may be more hype than reality." That tension is worth sitting with. A system that can guess what someone's looking at from fMRI data is genuinely useful for neuroscience labs, but "remarkable precision" claims deserve the same scrutiny we'd apply to any benchmark number from a lab with a paper to promote. Meanwhile, Musk co-leading a Pentagon warfare project and 22 nations pushing for a new global AI governance body are the parts of this newsletter that'll actually shape what researchers can build and where funding flows.

If you're a founder betting on defense contracts or a researcher relying on open brain-computer interface data, the governance fight matters more than the next flashy demo. Read the mind-reading result as a research milestone, not a product. Watch the Pentagon appointment and the 22-nation coalition instead.

Those decisions will decide which AI capabilities get built in the open and which get locked behind classification.

Common Questions Answered

How does the AI system reconstruct images from brain scans?

The AI tool analyzes fMRI brain scan data to identify patterns of neural activity and uses this information to reconstruct the image a person was viewing when the scan was taken. The system can perform this reconstruction with notable accuracy, allowing researchers to determine what someone was looking at based solely on their brain activity patterns.

Can the brain-scan reconstruction tool work in reverse to predict brain activity?

Yes, the same AI system can operate bidirectionally by predicting how someone's brain will respond to a given picture before they've even seen it. This two-way capability makes the tool particularly valuable for neuroscience research, as it enables scientists to both decode visual perception from brain scans and forecast neural responses to visual stimuli.

What is the significance of the brain-scan reconstruction tool for neuroscience labs?

The brain-scan reconstruction tool is genuinely useful for neuroscience labs studying visual perception and neural processing, providing a new method to understand how the brain encodes and interprets visual information. However, the tool's 'remarkable precision' claims warrant careful scrutiny, as benchmark numbers from research papers with vested interests in their findings should be evaluated with the same critical standards applied to other AI breakthroughs.

Why does the article emphasize caution about the brain-scan reconstruction breakthrough?

The article highlights the tension between genuine scientific utility and potential overhyping of AI capabilities, noting that AI's 'latest breakthroughs and fears may be more hype than reality.' This serves as a reminder that while the brain-scan tool offers real value for neuroscience research, claims about its capabilities should be examined critically rather than accepted at face value.

LIVE16:46OpenAI's Altman Says New Agent Is a Premium Product