Imitative AI is Eating All The Good Tech

Midjourney, a firm most known for its imitative AI image generation, has announced a pivot to medical scanning. Now, normally, this would be the kind of thing that I would applaud. Better medical scans are a great thing, and the kind of technology that I have been screaming that we should invest in ever since the crypto bubble. But Midjourney’s zig appears more about the pathologies of our current technology economy than about finding good products.

First, this is likely another sign of how deep and how fragile the imitative AI bubble has become. Midjourney is one of the premier imitative AI image generation systems. It had wild dreams of avarice, of replacing all human artists at every step of every production. Graphic art, storyboarding, movies — all could be and eventually would be replaced by the outputs of Midjourney. None of that, of course, has come to pass. Not only are there competitors that do the job as or nearly as well, the process of imitative AI is not and likely cannot be profitable. Which appears to be why they are moving toward medical scanning — they cannot make any profit off their core business. We know this to likely be true since their scanning technology also doesn’t appear to work.

The scanning is a kind of ultrasound — basically you sink into a pool of water and get scanned. They are supposedly supposed to produce MRI-level scans in a significantly lower amount of time with, obviously, a significantly lower amount of stress for the patients. No matter how bad you think you look in a bathing suit, I promise you, an MRI is worse. The problem is that Midjourney has produced pretty much zero evidence that their machine works. This is an established idea, so they are not inventing something from whole cloth. But they have provided no evidence that their version works as well or as easily as they claim that it does, and many experts in their field are skeptical. It feels, then, like so much Wall Street driven vaporware. A desperate play to convince the money men that they can still be a billion-dollar firm. And part of that desperation is the notion that people would use this all the time.

Midjourney claims that one of the benefits of their miraculous not quite new, not quite working system is that people could use it very often, collecting an enormous amount of data based on these scans. That, according to Midjourney, would inevitably lead to better outcomes. The idea, though, that more data is always better is insane, as should be clear to anyone who has worked with large data sets with frequent sampling. Not every data set is signal. Any results that indicate an issue would have to be followed up with different, sometimes painful and/or inconvenient tests. Anyone who has lived through a cancer scare, for example, knows how incredibly terrible the wait for confirmation can be. In addition, the notion that early detection inevitably leads to better outcomes is just not true. Some things cannot be cured regardless of when caught, and improved treatments mean that many things can be cured even if found late. A world in which we have constant noise about our health is not a world in which we are necessarily healthier. But it is a world in which Midjourney makes more money.

Different, cheaper, easier scans are a potentially beneficial advance. But Midjourney’s move here is not about such a benefit. This hype of scanners that are unproven as the next big moneymaker is very much in line with crypto, the metaverse, NFTs, 3D television, on and on and on. It is hype meant to drive investment and pay out the investors, not a real advance meant to solve a real problem. Combined with the insistence that everyone should use this as much as possible regardless of the wisdom of that course, and you have all of the pathologies that have driven our technology industry into a series of ditches.

The economy is meant to serve society. Society creates rules that nurture and protect the markets in the hopes that doing so will allow economic activity that improves the lives of its people. And it can work that way. Markets can produce increased living standards and products that improve lives. But American tech has not done that in some time. As the Midjourney scanning pivot demonstrates, it is too wrapped up in fast money and building recurring services that leech money from people regardless of the benefit to the users. We have collectively lost the plot, and if we want real progress, we have to change the incentives and rules that have so warped our markets. Otherwise, the future is filled with underwater scanners that do nothing much beyond scare you into using them.