AINEWS
The AI Image Company Now Wants to Scan Your Body
Midjourney says its ultrasonic scanner would be faster and cheaper than an MRI, but its own blog post admits the hardest engineering step is still unsolved.
Midjourney says its ultrasonic scanner would be faster and cheaper than an MRI, but its own blog post admits the hardest engineering step is still unsolved.
Midjourney, the AI image-generation company best known for turning text prompts into pictures, has announced plans to build a medical body scanner that it says would be faster, cheaper, and less invasive than an MRI. The device, described in a blog post first reported by Gizmodo on Wednesday, does not exist yet. Midjourney uses the word "envisions," and the company itself flags the hardest step as unsolved.
The proposed device would surround the body with roughly half a million ultrasonic sensors, each about the size of a grain of sand. The ring would send sound waves through tissue from every angle, harvest the reverberations, and assemble them into a static internal image. The pitch borrows from dolphin echolocation, a metaphor the company uses itself in announcing the technology, and the comparison does real work for a reader: it explains the geometry of the array and why more sensors, fired from more angles, might in principle see more of the body.
The catch, by the company's own description, is computational. Midjourney says the scanner would produce terabytes of raw data per second, but turning that torrent of overlapping echoes into a clean image remains "a major computational task" the company has not solved. Until that step works, the device is a sensor array in search of an image, not a diagnostic tool. Calling a working scanner "envisaged" is honest, and the honesty is the news.
That gap is the story, and it is a pattern. When an AI company crosses into healthcare with a public announcement, three checks help a reader calibrate. Has the company shipped medical hardware before, and if not, what category of regulatory and clinical work is it walking into? Is the unsolved engineering step named specifically, not waved at as a future problem for someone else? And is "cheaper than an MRI" a cost claim, a speed claim, or a diagnostic equivalence claim, because those are very different things? On all three, the Midjourney announcement reads as a vision, not a product.
Midjourney has no medical-device track record. Its core business is generating images from text, and the company did not say whether the scanner project has a clinical partner, a regulatory pathway, a working prototype, or even a published paper on the signal-processing problem. The press cycle around generative AI has produced a steady drumbeat of companies crossing into healthcare with consumer-facing language, and the pattern is worth naming out loud: the announcement tends to arrive before the engineering, and the wellness framing tends to arrive before the clinical evidence. This announcement is a clean example of that sequence.
The same blog post also unveiled a downtown San Francisco spa featuring "pools of golden light which softly scan your body." The spa is described as a separate venture, but it lands in the same post because the branding is the point. A diagnostic scanner and a luxury wellness experience are being rolled out together, which tells the reader something about the audience the company is courting, and about the level of clinical rigor to expect from the announcement so far.
For comparison, MRIs do not just cost more than ultrasound. They image different things. Magnetic resonance excels at soft-tissue contrast, the kind of detail that lets a radiologist tell a tumor from healthy brain, and that is the diagnostic work the technology was built to do. Existing clinical ultrasound is cheaper and faster, but it has well-known limits on what it can resolve inside the body, and it bumps into the same problem Midjourney would have to solve: turning acoustic noise into interpretable pictures. Saying a new ultrasonic device is "cheaper than an MRI" is true in a narrow sense, but it skips the harder question of what the device would actually see, and what a clinician would actually be able to do with the image.
The next test is concrete, and any reader can run it. Has Midjourney named a research partner, a clinical site, or a regulatory filing? Has it shown a static image produced by the array, even on a tissue phantom or a volunteer? Has it published, or pointed to, a paper on the reconstruction step that converts terabytes of reverberations into something a clinician could read? Until at least one of those answers is yes, the announcement is a vision worth tracking, not a product worth comparing to the MRI suite down the hall.