Midjourney’s announcement that it is developing a medical scanning system drew quick attention from the AI community, and a detailed critique from Astral Codex Ten author Scott Alexander. In a post published Thursday, Alexander argued that the company’s prototype may be impressive as a technical demo, but is unlikely to be an immediate substitute for mainstream medical imaging.
Midjourney said its scanner would use a water tank and a ring of ultrasound sensors to create a three-dimensional image of the body. The company described the device as a new kind of machine intended to reshape healthcare. According to the announcement, the system would use ultrasound tomography to generate rich body data without the radiation associated with CT scans and in a matter of seconds.
Alexander, who said he was offering only a preliminary view rather than a specialist medical opinion, said the scanner’s promise may be overstated. He argued that conventional ultrasound has major physical limits, especially when doctors need to image areas blocked by bone or air. That would exclude many common diagnostic targets, including the brain, lungs, and bowels, he wrote, while also limiting the scanner’s usefulness for situations where clinicians need a full-body view.
He also questioned whether the device could even displace ordinary ultrasound, which is cheap, portable, and widely used at the bedside. A tank-based system, he suggested, would be far less practical for routine hospital care. In his view, the scanner would need to outperform standard ultrasound by a wide margin to justify the added complexity.
One area where he said ultrasound tomography can make sense is breast imaging, especially when dense tissue makes mammograms less effective. He noted that this is not a new concept and pointed to existing medical devices that already use a form of ultrasound tomography for breast scans. Those systems, he said, are narrower in scope than Midjourney’s concept but show that the approach has at least some established clinical use.
The bigger question, Alexander wrote, is whether the scanner could work as a kind of whole-body screening tool for healthy people. That is the use case Midjourney appears to be emphasizing. He compared the idea with whole-body MRI, which already exists and can image more structures, but is not broadly recommended by doctors because of false positives and limited evidence of benefit.
According to Alexander, the main obstacle is not just technical capability but medical utility. He argued that whole-body scans often identify incidental findings that look alarming but do not require treatment, leading to more testing, anxiety, and in some cases unnecessary procedures. He suggested that simply adjusting the threshold for what counts as suspicious does not fully solve the problem, because patients and doctors may still push for follow-up once a scan shows something unusual.
He said the most plausible near-term argument for Midjourney’s scanner is convenience and cost. A system built from ultrasound hardware could, in theory, be cheaper to run than MRI equipment and less claustrophobic for patients. That could appeal to wealthy consumers already paying for preventive scans, even if mainstream medicine remains skeptical.
Alexander also left open the possibility that the technology could become more useful later if paired with better AI models, better data, or new methods for imaging through bone. He framed the product less as a finished medical breakthrough than as a bet on future improvements that might make diagnostic ultrasound more powerful than it is today.
For now, though, his verdict was cautious. The scanner may be interesting as an experimental platform, he wrote, but it does not yet look like a replacement for established imaging tools or a clear answer to the problems of screening medicine.