Scott Alexander doesn't just question a new medical device; he dismantles the Silicon Valley fantasy that artificial intelligence alone can bypass the hard constraints of physics and human biology. While the tech world celebrates Midjourney's pivot to a full-body ultrasound scanner as a revolutionary leap, Alexander, writing from his perspective as a psychiatrist rather than a radiologist, argues that this "bold new kind of machine" ignores why medical imaging works the way it does today.
The Physics Wall
The core of Alexander's critique rests on a fundamental limitation of sound waves: they cannot penetrate bone or air effectively. He writes, "Ultrasound is great, but it can't penetrate bone or air," noting that this immediately disqualifies the technology from imaging critical organs like the brain (blocked by the skull) or the lungs and bowels (filled with air). This isn't a software bug; it's a law of nature.
Alexander points out that while standard ultrasound technicians use their hands to angle probes around these obstacles, a stationary water tank cannot replicate that dexterity. "Most likely their Scanner won't be able to match this level of precision," he argues, suggesting the device would miss the very organs doctors most need to see when looking for metastases or unexplained symptoms. Critics might note that emerging techniques like Full Waveform Inversion Imaging could theoretically solve the bone-scattering problem, but Alexander counters that such math requires "gargantuan amounts of compute" that simply do not exist yet in a clinical setting.
"The trivial reason is that due to the limitations of physics ultrasound will always be less capable at resolving anatomy than MRI or x-ray-based methods that we already have."
This quote, drawn from a radiologist Alexander interviewed for context, underscores the skepticism surrounding the project. It suggests that the industry's hesitation isn't just about cost or logistics, but about the fundamental resolution limits of sound compared to magnetic resonance or X-rays. While historical precedents like acoustic shadowing in older imaging modalities taught us how to work around these issues, Alexander implies that trying to brute-force a solution with AI is a gamble against established medical reality.
The Screening Trap
Perhaps the most provocative part of Alexander's analysis targets Midjourney's proposed business model: using the scanner for annual whole-body screenings on healthy people. He acknowledges this is where the company seems headed, imagining "healthy people go for one of these scans regularly" to catch tumors early. However, he dismantles the logic that earlier detection always equals better outcomes.
He explains that current medical consensus advises against whole-body MRIs precisely because they generate too many false positives—finding harmless anomalies that trigger panic and unnecessary invasive procedures. Alexander writes, "A 'false positive' in this context isn't the scanner hallucinating something that isn't here. It's the scanner finding some sort of boring abnormality that doesn't matter - like a zit but inside your organs - and then making everyone panic."
The argument here is that no amount of AI sophistication can easily solve the human and legal reaction to a "large mass" found on a scan. Even if the software could distinguish between benign and malignant with high accuracy, patients and doctors operate in a risk-averse environment where ignoring a potential finding is legally perilous. As Alexander notes, "Smart technocrats can set the threshold for investigation wherever they please, but a patient who learns that they have a large mass in their brain isn't going to accept 'no' for an answer."
This connects to broader debates in medical history regarding incidental findings, similar to how "incidental imaging findings" in routine scans often lead to cascades of care that may not improve survival rates. The technology might be faster and cheaper than a $2,000 MRI, but Alexander suggests it would simply serve the same demographic: "pushy rich people who are already getting the screening MRIs," just in a more comfortable setting.
The Long Game Bet
If the immediate medical application is flawed, why does Midjourney exist? Alexander posits that this is less about selling scanners today and more about positioning for a future AI revolution. He suggests the company is betting that "AI will benefit ultrasound tomography in particular," potentially solving the bone-penetration problem or learning to interpret data so well that it outperforms human radiologists.
He frames this as an attempt to gather massive datasets: "If you get lots of high-quality ultrasound data... maybe you could train a really good AI on that data and do better than any existing radiologist." This is a classic Silicon Valley strategy—build the infrastructure now for a capability that only exists in theory later. However, Alexander warns that this relies on a specific assumption: that AI will revolutionize ultrasound specifically, rather than improving MRI technology or curing cancer entirely before the scanner becomes viable.
The counterargument here is strong: by the time AI solves these physics problems, existing modalities might have evolved so much that a dedicated ultrasound tank is obsolete. As Alexander puts it, "By the time the AI revolution has solved all the problems that stand in the way of the Midjourney Scanner, might it also have beefed up normal MRI... or enabled some kind of entirely new scanning modality?"
Bottom Line
Alexander's piece serves as a necessary reality check against the hype cycle, grounding an ambitious tech pivot in the unyielding constraints of anatomy and clinical practice. The strongest part of his argument is the dismantling of the "early detection" myth, proving that better data doesn't automatically mean better health outcomes if the medical system isn't ready to act on it without causing harm. Its biggest vulnerability is the assumption that AI progress will be slow in other modalities; if MRI or CT technology leaps forward alongside AI, Midjourney's bet could look less like a visionary play and more like building a phone booth for the internet age.