Agreed. Not a radiologist, but I do a fair bit of MRI research. Experts vs lay people probably have different success with getting the right diangosis out of a frontier model. Subtle changes in prompts can cause different diagnosis[1]
I mean, probably not. No expert, but everytime I go to an immunology meeting (I'm a paediatrician) they've got a whole stack of new diseases. The field is moving fast, and there has to be a careful amount of shared decision making about when to test, what a positive test means and so on. I reckon they're as safe as any of us.
I'm interested to know about the approaches that you tried with the ML, and then decided to not use. In practice, the options are so many. How did you come up with the final approach - and was there a systematic way to decide which options to go for?
But for kids over 8, a nice long form video works well. That, and having enough time so that they don't feel like we're in a rush, but also not taking to long to load them onto the scanner...
For the younger ones, it's very much dependent on the child. So we take a bit of time to get to know them before we get them to attend. We have videos to prep them, and can follow a script when loading them (e.g. becoming an astronaut and blasting off into space...).
Part of my job, is that I design protocols to help young children lie in MRI scanners for a living. We have all sorts of techniques to help with this.
However, for each new scanning protocol, I like to have had it myself - so I know what the children go through. And, at times lying inside a MRI scanner, detached from the world, with only the noise of the scanner (very reduced with our new noise cancelling headphones), is almost meditative, and a welcome escape from the constant connection and pressures of being immediately available at work. Sounds like the writer achieves something similar in the coffee shop.
A systematic review like this can be helpful, in that it identifies where there are gaps in the literature, and prevents hype - if some studies show evidence of effect, and others do not, even if there are only a few published studies then we know somthing new about the totality of the literature on the subject.
That said, this particular systematic review has a couple of issues (e.g. I can't find the precise inclusion / exclusion criteria, nor can I find that it has been pre-registered on Prospero or another database).
I have written a few systematic reviews where there is very little data already availabe, and we use them to explain to funders why we need to do further research on a given topic.
'This is possible because the man is the candidate's father. When he says "he's my son," he's simply stating their family relationship.
The scenario doesn't present any logical contradiction - a father could very well be in a position where he's supposed to interview his own son for a job. This would create a conflict of interest, which is why he's saying he can't conduct the interview. It would be inappropriate and unfair for a parent to interview their own child for a position, so he would need to recuse himself and have someone else handle the interview.
The phrasing might initially seem like it's setting up a riddle, but it's actually a straightforward situation about professional ethics and avoiding conflicts of interest in hiring.'
I'm a paediatrician. No parent has ever asked me for their baby's weight in kg - they are all pounds and ounces. So much so that I can do this niche conversion almost in my head, at least at the start of the day, as we weigh them in kg.
I agree - out of the box, a LLM will write an extremely poor essay. But that is not how to use the LLM. WHat you do, is give the topic to the LLM, and say, "with this topic, what are 20 titles that I could have for essays." Then, you pick the best title, and you then say, "I am writing an essay with the title <insert title>, what should the central argument for the essay be?" You then ask for a list of bullet points for the central theme of each paragraph to elaborate on that argument, then you ask it to write each paragraph individually, then you put the pieces together, and ask it to proof read it's own work and finally you add in some references of your own. Within 1/2 an hour, you have a passable essay. And if you then iterate on it with the LLM you can fairly soon have a half decent one, especially if it is a well trodden area (presumably with lots of training data that the LLM draws on).
I tried to get students to critique things, but even then you can put in the text to critique to a LLM with a long enough context, and the LLM will kick off with a passable critique, if you iterate with it enough.
So, even though they'll never get a top mark, they will still be able to get through the assessment. So I don't set essays any more.
I am an educator at a UK university. The essay is rapidly ceasing to be an appropriate way to assess students knowledge and critical thinking.
We regularly organise in person face to face practical exams for our entire several hundred strong year group of undergraduates. It is possible to do assessment properly if the will is there.
I work in pediatrics and am an academic investigating MRI of kids in various diseases. When I saw this work, I did wonder about us being better able to functionally map where things are going wrong in the pathways of neurodisability. I wondered if this would have applications in being able to do that - for example being able to say that someone could process the image. Do you think it could have this type of application? One thing which would be a deal breaker at the moment is the amount of time participants spend in the scanner. But if we wanted to (for example) see if a child could perceive simple objects, would that be doable do you think?
[1] https://www.nature.com/articles/s41591-026-04501-8