Bernd Porr
banner
berndporr.me.uk
Bernd Porr
@berndporr.me.uk
Realtime AI, Bio-inspired AI, robotics, biosignals, neuroscience, research ethics and signal processing. Senior Lecturer at the University of Glasgow. Occasional filmmaker (https://www.eigenproductions.co.uk/). Oh, and recovering anti Brexit activist.
IEEE journals are full of blurry diagrams and pixelated formulas. It's amazing that "engineers" cannot even create vector graphics. Below not just blurry but a summation node with no outputs but 3 inputs!
bsky.app/profile/bern...
Talking about terrible papers I just found this beauty on IEEE explore (but there are many more) attempting to draw a data flow diagram. Note that the summation node has only inputs. No outputs! And of course drawn in something like MS Paint.
ieeexplore.ieee.org/document/101...
December 3, 2025 at 7:33 AM
In my engineering bubble it's this endless vacuous bullshit of "I'm delighted to have been promoted to regional sub chair of IEEE for south Glasgow" or "I'm delighted to show in this paper that simulating two digital twins talking to each other in MATLAB will make sure we'll reach net-zero."
November 29, 2025 at 11:49 PM
Never had this before: the classifier is a function.
November 27, 2025 at 11:30 PM
So that's all what they tell us about their ML model. Totally explainable. 🥳
November 27, 2025 at 11:27 PM
..and the so called graphs also because they are all terribly blurry. Looks like they took their mobile phone and took a photo of excel.
November 27, 2025 at 10:58 PM
Reposted by Bernd Porr
I will never tire of these wonderful comics about Ada Lovelace (and Charles Babbage] from Sydney Padua

sydneypadua.com/2dgoggles/co...
November 27, 2025 at 9:01 AM
It also make sure that they will never bother to settle so that then a couple of years down they line she can say: "See, they don't integrate. Get rid of them!"
November 15, 2025 at 9:11 AM
Fuer das Geld koennte man auch eine grosse goldene Kuh errichten und alle koennten drumherum springen. Politiker koennten Reden davor halten.
November 15, 2025 at 8:53 AM
The whole article hinges on: "...these large training events only happen every few years." Just googled (!) this claim and there is virtually no info out there. Training a new model is not a one-off thing but will go wrong many times and starting from scratch most likely over months.
November 11, 2025 at 11:53 AM
Wondering if it's giving in to bullies or not pissing off their mates. Or What Goodall pointed out: it's been mostly so sanitised that there is no actual message left. It's not just news output but also resorting to inoffensive period pieces or other inconsequential entertainment.
November 11, 2025 at 8:09 AM