Nicola Romanò @nicolaromano@qoto.org
Senior lecturer at Edinburgh University and Zhejiang-Edinburgh Joint Institute (ZJE).
Undergraduate Programme Coordinator, Biomedical Informatics at ZJE.
I teach #imageanalysis & #dataanalysis with #RStats & #python.
My research is focused on how #heterogeneous behaviour in #pituitary (and other) cells shapes their function as a population.
I'm also very interested in #reproducibility and #openscience.
Joined Oct 2022
@MCDuncanLab @yvan I agree with what you're saying however I think we should stop using the strawberry example for two reasons.
First, LLMs are not designed to count, but to predict the next most probable set of words so, you're evaluating a task they were not designed to complete. Second, there are more complex tasks where they do get it right most of the time. It's likely that the muscle fibre question is in many websites that have been used to train ChatGPT, so it probably will give you the correct answer, and the whole strawberry thing can be dismissed as a glitch (by the way the latest version of ChatGPT gives the right answer).
The main issue (leaving ethics aside for a minute) is that these systems work the majority of the time, so it's easy to assume that they work all the time. That is, until they don't work on a serious task. Unfortunately, the AI companies narrative forgets to mention that, and prefers to sell their product as having PhD level intelligence (as if having a PhD was a guarantee of anything!)
Nicola Romanò
boosted
I am beyond excited to announce that ggplot2 4.0.0 has just landed on CRAN.
It's not every day we have a new major hashtag #ggplot2 release but it is a fitting 18 year birthday present for the package.
Get an overview of the release in this blog post and be on the lookout for more in-depth posts hashtag #rstats
@infotroph My hunch is that you are correct. Fitting a gamma distribution using empirical quantiles should in theory get rid of sampling noise.
I have a strong feeling that, especially for small empirical sample sizes, the 5th/95th quantiles will be particularly sensitive to small sample sizes (more than the median). Do you only have one sample? Maybe you could bootstrap the quantiles?
@rolandixor And that's why teaching critical thinking is becoming oh so much more important. And not only for Linux, of course!
@khleedril @skyglowberlin @highergeometer Honestly I'm not sure how you got to that conclusion. Just quickly browsing it, that looks like an absolutely fine paper; not everything needs to be unreadable and complicated. Not quite sure why you think the discussion is AI generated either... The researchers who authored the paper used open source Python libraries instead of reinventing the wheel, to simplify what apparently is a complicated problem in their area of expertise using those tools. If more people did that academia would be a better place I think.
Just finished re-reading this classic great paper, and now I am definitely younger
False-Positive Psychology: Undisclosed Flexibility in Data Collection and Analysis Allows Presenting Anything as Significant
Nicola Romanò
boosted
This is what pull requests used to look like.
(The last sentence is particularly remarkable!)
Nicola Romanò
boosted
YES! Thank you...
Say it again, and again, and again. They are not "illegal immigrants".
- seeking asylum is legal and has been since 1951.
https://northeastbylines.co.uk/news/politics/silencing-the-drumbeat-again-seeking-asylum-is-legal/
Nicola Romanò
boosted
More evidence (this time from the public sector in Scotland) that the four day week not only improves workers' wellbeing, it also improves productivity. Moreover, this is a proper four day week - rescued hours but no loss of pay - making it a powerful example, both for the Scottish government's management of its civil service & more widely.
The momentum is building for the four day week & its social benefits....
Nicola Romanò
boosted
Must-read from @taylorlorenz
So-called "age verification" laws are spreading.
They will not protect kids.
They are grossly insecure.
They will be tools of censorship, surveillance, and oppression.
They amount to requiring a license to speak -- and to read.
We have to stop this.
https://www.usermag.co/p/we-must-fight-age-verification-with
Nicola Romanò
boosted
Windows 10 support ends in 2 months, potentially making a lot of computers obsolete and dangerous to use. https://endof10.org/ is a great resource to avoid that by installing Linux - aided by experts and with simple to follow guides.
@almenal99 oh yes... not to mention the following 10 meetings to decide what to do! This year the formative workshops will be done without internet access, so hopefully students will realise in time that they are not able to do simple tasks.
The bright side of the story is that the other half who did pass the exam, did quite well!
Nicola Romanò
boosted
We just published a new Python package for the single-cell analysis of morphological profiles! If you do microscopy of cells and wish there were better solutions than averaging cell populations, now is the time! Built on the scverse, we bring analysis of imaging data into the single-cell era. Check it out now 👇
Paper: https://doi.org/10.21105/joss.08324
Documentation: https://scmorph.readthedocs.io/
@almenal99 well, last year half of the class in one of our courses failed their exam because it was done on computers with no internet access... I hope the message will come across strong this year that you can't learn programming by having AI write code for you...
Nicola Romanò
boosted
After a 13-year immunization campaign, Nepal is declared to have eradicated rubella. https://news.un.org/en/story/2025/08/1165672
@michaelgemar @jcoglan @joe Funny thing about that podcast thing is that after the initial moment of "oh that's a cool trick" you very soon realise that reading the actual text would be faster, less annoying (because you don't have to listen to pointless puns) and give you oh so much more useful information...
I'm also looking for examples of (deceptively) simple programming tasks that genAI doesn't quite get right. Something on the lines of:
Write a function that does <simple task>.
Where in most cases you'd get an output that is either wrong or it kind-of works but not quite.
Hello! I am interested in hearing from anyone who has (un)successfully made use of genAI in any shape or form specifically when teaching programming. I'm thinking specifically university at undergraduate level but happy to hear about any teaching related activities at any level.
Specifically what did you do? What were the challenges and opportunities? Could you have done the same without genAI?
Nicola Romanò
boosted
Remember when the common wisdom was to not pirate commercial software because downloading random executables was likely to infect you with malware that would steal your data or worse?
Now we pay for legitimate, signed copies of software from vendors who load their applications with AI "features" that exfiltrate your data to remote servers and no way to disable that.
I feel old. When I started out, we got our malware for free and were doing something wrong when we did.
Maybe this is payback for all those C64 games I pirated when I was 16.
@katemorley "No one can remember how many pounds in a stone"... unless it's a baby, then it's pounds, all right.
I once had this lovely elderly couple asking me how heavy my dog was. I said 10.5 kg, they looked at me for a second and the wife replied "sorry, we don't do kilos...". 😂
Senior lecturer at Edinburgh University and Zhejiang-Edinburgh Joint Institute (ZJE).
Undergraduate Programme Coordinator, Biomedical Informatics at ZJE.
I teach #imageanalysis & #dataanalysis with #RStats & #python.
My research is focused on how #heterogeneous behaviour in #pituitary (and other) cells shapes their function as a population.
I'm also very interested in #reproducibility and #openscience.
Joined Oct 2022
