Links from December 2024: pattern games, probability, more genes, markdown, gene activity
Pattern games for dogs, probability doesn’t exist, more genes, AI helps better teams, markdown conversion, gene activity from slide images
Pattern Games – A New Way to Help Change Behavior
“The main goal of Pattern Games is to help your dog process their environment and give them a set of behaviors to follow”. Sounds interesting, and the article points to the definitive source of games: Leslie McDevitt’s Control Unleashed. The publisher “no longer ship[s] to the United Kingdom since Brexit” but I’ve found a UK distributor. There are also videos on Youtube.
Why probability probably doesn’t exist (but it is useful to act like it does)
David Spiegelhalter: “Meteorologists make predictions of temperature, wind speed and quantity of rain, and often also the probability of rain — say 70% for a given time and place. The first three can be compared with their ‘true’ values; you can go out and measure them. But there is no ‘true’ probability to compare the last with the forecaster’s assessment. There is no ‘probability-ometer’. It either rains or it doesn’t.”
Generative AI is making traditional ways to measure business success obsolete
“Historically, project management and business success was largely defined through a simple formula: Cost x Time = Quality […] AI has upended this thinking, as firms can now achieve both speed and accuracy at the same time by leveraging AI”. For some use cases, I would have added.
A firm randomly assigned its scientists AI: here’s what happened
The good teams got better, while there was no change for the weaker teams. “Top scientists leverage their domain knowledge to prioritize promising AI suggestions, while others waste significant resources testing false positives”.
‘Dark proteome’ survey reveals thousands of new human genes
We thought Humans have around 20,000 genes. It seems there are short sequences in the genome—noncanonical open reading frames—that produce “miniproteins”. How many genes now? “50,000 is in the realm of possibility.”
AI tool ‘sees’ cancer gene signatures in biopsy images
The hunch was that gene activity would produce tiny, but potentially visible, changes in slide images of tissue. SEQUOIA is the resulting model: give it an image of a tissue, it estimates gene activity. This is important for selecting cancer treatment, without the cost or delay of having to do any RNA/DNA analysis. Too good to be true?
The ABCs of Alphafold 3, Boltz and Chai-1
Comparing the performance and open-source-ness of various to predict protein interactions. TL;DR use the Boltz model.
Banish the PDF-hunting blues with these AI and digital tools
I did not know that PubMed search has a ‘create RSS’ option, nor RabbitResearch or Consensus (a sort of Perplexity-like tool).
MarkItDown
“The MarkItDown library is a utility tool for converting various files to Markdown (e.g., for indexing, text analysis, etc.)”. Do: pipx install markitdown