I was wrong about the Midjourney ultra-sound scanner
twitter.com
twitter.com
There have been a lot of more thoughtful analyses of the scanner circulating on Twitter that have actually highlighted the real problems with what they’re claiming.
An example of an important point brought up by someone with actual domain expertise is to point out that ultrasound doesn’t travel well through bone or air. Once you realize this, you understand why they chose the slices they picked for their marketing images. Get up toward the rib cage and lungs and ultrasound isn’t going to be doing the magical things they claim.
Even this post hedges with “if it’s high resolution” claims, but we already know what the resolution looks like. You can see the images they’re producing and they’re pretty rough. They’re using sensors from another ultrasound company so it’s not like they made a breakthrough. The concern now is that they’re going to start trying to make up for the limitations of ultrasound by having AI process the images into something that looks more impressive and hides the limitations of the technology.
I don’t know why this particular Tweet is trending, because it doesn’t seem to add anything at all to the conversation. If you spend any time on Twitter this feels like template engagement bait designed to ride a popular topic without adding anything to the conversation.
> If the Midjourney ultrasound is high resolution, harmless, inexpensive and convenient, people can get an initial scan...
> If the MIdjourney device can be repeated frequently, like weekly, at a low cost and is harmless..
This particular author is backtracking on his original idea that lots of frequent scans are bad, as long as they are cheap and accurate. But that's a pretty irrelevant side issue IMO when the vast majority of objections I've seen to Midjourney's announcement have been that qualified folks just don't believe the tech is medically feasible - it won't have the necessary resolution to discriminate findings.
I'm not faulting the author, who was quite clear where he did a rethink, but I do fault people who somehow think this is evidence that Midjourney's scanner will be viable.
TBH, while I don't think the Midjourney announcement is the same level of malevolence as Theranos, I don't think it was quite far off. I'm baffled that people think science and medicine should be done by a flashy website and PR-speak instead of the sober language of research reports, but I guess that's just a (sad, IMO) sign of the times.
The original Midjourney Medical thread here on HN was loaded with people confidently declaring that these kinds of scans are a bad idea and net negative because of the risk of false positives.
Conditional on low base rates, and the need for invasive follow-up testing, they are likely to cause more problems than they solve.
But personally, I love all this stuff and have an emotional belief that more data is always better, which is weird in cases like this where it turns out that's not the case.
The irony is I believe that if a medical devices company announced this, it was being sold to hospitals, and it would only cost the patient's insurance $100 a scan, then the medical industry would universally praise this as a breakthrough.
It is very easy to be cynical about change...especially in areas we are knowledgable...because all we see are the challenges.
And there will be lots of challenges with this. For my part, I'm not wild about what Midjourney might be allowed to do with this data. However, dealing with those problems seems better to me than leaving things as they are. This X post is a great example of "yes, and" instead of "no."
yes, that's the power of reputation. if a company with a proven track record of selling effective diagnostic tools decided to stake their reputation on a new system that sounds a bit like something from an ai-generated fairy tale, people might be more likely to give it the benefit of the doubt.
when a company best known for selling actual ai-generated fairytales announces a medical diagnostic tool that sounds like an ai-generated fairytale, i think it's reasonable to treat that with some skepticism.
No, it's not just about reputation, at least for me it's about the fact that there is a relatively strict, FDA-approval process for actual medical devices. Midjourney's announcement was the equivalent of a marketing page for some supplement that claims it will make be sleep great while growing my penis (the page even basically even marketed it as a "wellness" product), not actual scientific evidence that the device would, or could, work as advertised.
Something like this will never cost patients $100. Even if the actual cost to a provider was $5, patients would be billed $500 after insurance…
1. We should absolutely pursuing these kind of ideas but given then nature of technological progress and our history with “democratization” things are likely to get worse before we get better. Matt is hedging a lot here reflecting this.
2. Maybe all this stuff is as promising as the various threads suggest but it’s bizzare that this is all being argued in culture war terms (you vs the gatekeepers) and not like shared human flourishing terms. Again maybe it’s working, but it’s also being marketed to a certain kind of persons fears, not as the future of human understanding.
Buried in tlb's response might be the J-curve, the S-Curve (=integral of J-Curve), the hype curve, and finally Braess/Jevon/Baumol: why is healthcare inflationary if it is so helpful and so needed. The tension between helpfulness and perceived necessity must be explored, ceterum censeo (3.)
In general, I think we should applaud this though.
Any genuine attempt to create novel medical technologies is probably a good thing (assuming they’re non-invasive and non-painful).
Unless it’s a Theranos situation, I think it’s a great thing to attempt, even if it fails. So many things we rely on today are the result of a successful attempt, but the failures were just as necessary for the eventual success.
That ambition is very positive to me.
>>That ambition is very positive to me.
...and this is why people fall for it. Every. Single. Time.
Literally everything they're saying is marketingslop gobbledegook. No studies, no papers, no doctors; just "500k transducers," and "30fps!" Anyone can wire that up. With a little cash, you might even be able to stream, record, and proccess it. Still means absolute diddly squat if you haven't compared it to other imaging or figured out how to train a radiologist to use it effectively or done trials for specific diagnostics or diseases.
They'll figure out if it does anything other than show you an animated cartoon xray of yourself later. After they have your money.
You’ve pretty much demonstrated the criticism is true, that all these grandiose claims made by VC backed medical device companies are… bunk. They made something smaller and shittier, yay?
With regards to Midjourney Scanner, you have no evidence that any fraud is happening or is likely to happen with this device. AFAIK they haven't actually made strong claims about what this will be capable of in a medical context, so fraud would be quite hard at this stage anyway. As such, it seems unreasonable to expect them to already have done studies with doctors.
Those people generally got shouted down when Theranos was still the hot story in SV, and of course after everything came out suddenly everyone knew it was fishy all along, which is just absolute bullshit. I remember reading comments here on HN suggesting that people saying Theranos was fraud were motivated by misogyny, just absolutely infuriating stuff.
So now we have an AI company coming out with promises of a revolutionary leap forward in medical imaging, in terms of cost and information gathered, and they're not using some sort of revolutionary sensor that they've invented, all the "hard" engineering has been done by other companies that have been in this game for decades at this point and how have R&D staffs with deep knowledge and expertise, but who somehow lack the ability to take the next step, they're coming in with the software and automated interpretation, which is basically the "?" step in the underpants gnome business plan.
The real truth about stuff in the biology and the medical world is that its all insanely hard due to the complexity of the systems involved, and there are tons of skilled and smart people who dedicate their careers to moving this stuff forward. There are really no low-hanging fruits being ignored by "the establishment" waiting for an outsider AI company to come in and overturn the table. Progress is basically won by sweating it out at the lab bench and accumulating a bunch of hard-fought incremental wins over a decade or so. Its frustrating that Elizabeth Holmes is still in jail and yet we're all here forgetting every lesson we should have learned from the previous go-round.
Very smart people were trying and failing to effectively split the atom before some very well-funded and well-placed individuals made it happen. Loads of very smart people were working on AI before Transformers were developed and made LLMs viable / intelligent enough for real work. People trying and failing to do something hard doesn't mean it is impossible, it just means they've found 10,000 ways to not make a lightbulb.
A theranos situation is one where you're lying about what you have reason to believe the new device will do - saying it will do things that you have no reason to think are even plausible - that you have prototypes doing the thing even. Not one where you're merely experimenting with something new that might or might not pan out.
The argument is that current full body scans often have false positives, and our treatment of false positives is bad including risky biopsies. Some have gone so far as to literally state early detection does not lead to better outcomes (simply not true - genuine early detection is very helpful but what they mean is it is outweighed at the population level by false positives).
The counter argument is that more information is always good but that you must learn to handle the noise, and we should focus on improving how we handle false positives. Including, but not limited to more frequent and abundant scans of various types.
The reason this is trending is because it both includes the argument and feels nice that someone changes their mind.
Personally, I think a lot of the issues here come down to the fact that we lie about the statistics instead of just show them. A test is neither positive or negative. It’s x% updated probability that you have something.
Maybe we need better treatment of the results rather than less scans?
> And overall, the side effects from all the risky, invasive procedures to track down the over 90% of stuff that was harmless equal or outweigh the benefit from removing the less than 10% of stuff that wasn't harmless.
I accept this (well-used) perspective from a practical, current perspective, but not for abstract diagnostics generally. From the Bayes' theorem, and same logic you use in Kalman filters: More knowledge, if you have data on the confidence, always helps. It only causes these negative outcomes due to acting poorly with the data (e.g. due to emotions and liability concerns, I suspect here)
to generalize accurately, one must forget (irrelevant) details
That is NOT a typical patient response. Go run that in front of any doctors and watch them laugh you out of the room.
1) When most people see something medical flagged, their first reaction is PANIC.
My friend's wife had something flag on her mammogram--the biopsy later showed it to be benign. In the meantime, her anxiety drove her blood pressure to something like 175 over 130 and they had to up her anxiety and blood pressure meds until the biopsy came back. Her probability of stroke was way higher than her probability of breast cancer. My grandfather had a little blood in his urine from a slowly growing bladder cancer. He forced them to treat it in spite of the fact that he was in his mid-80s, and the probability that cancer would kill him was zero. Instead, he basically died feeling miserable from the cancer treatments 6 months later.
2) Most working people cannot take off the amount of time necessary to do a weekly set of followup medical appointments for a couple months. Again, this is very different from tech geeks who can pretty much throw their work schedules around pretty flexibly.
the state of the art is dexa scans but they can be off by 5+% and more error on the distribution of the fat
- imaging that can get down to the cellular level easily and often.
- the ability to process that imaging data to find issues effectively.
- the ability to act on that data in a minimally invasive way.
This is a step in the right direction for one and two and the third has had progress too by others. We aren't there yet, but I can see a future where individual cells are treated and at that point all sorts of things are possible.
Unfortunately it hasn't kept up on image quality, detail, or text rendering. I think this is because they don't have the $ to keep up in the scaling game.
https://radiologybusiness.com/topics/healthcare-management/h...
This post didn't seem particularly AI generated to me (and notable even the OP didn't claim it was - just that they are "also prompting chat to write his posts").
It reads as someone with a moderately qualified opinion changing their mind on something and giving reasonable reasons.
The OP complaining that they "are an AI booster" and that they use AI seemed to be non-specific points rather than anything they didn't like about this post.
The thing is these are more likely than not not sincere posts but are probably bots accounts used to get some skimming amount of money from the generated content. I'd rather read actually interesting details, not the "template engagement bait" (as the top comment) posted so we can debate that instead of this contentless generated post.