I listened today to one data scientists' job interview. While answering on how to deal with LLMs' hallucinations, he put it like actually all these models do is hallucinate, and occasionally their hallucinations coincide with reality. That was a good one. AI in a nutshell.
Ayer fue un dia libre, y pasé enfrente de ordenador solo una hora. El dia sin ordenador - una sorpredamente buena experiencia.
When I was a student, I have written a little program to model a glissade of a descending plane, as a part of some study task. Not a rocket science, still involved an itty-bitty pinch of differential equations, required some effort (and was not fun). That resulted with a thin white line on a black screen.
Anyone seeing it shrug their shoulders, like "ok, nice".
Also I was interested in graphical effects, and I can tell you that modelling flame is a basic arithmetics. Initialize a bottom of your display matrix with random (or you can play with it) colors, then go over every pixel calculating average of values of surrounding ones, and put that result one pixel higher. Looks awesome. Anyone seeing result was like "wow, you're a hacker man!".
Here probably should be some kind of morale but to be honest, I love that flame effect better myself.
"After Yang" A nice sci-fi drama
Cloud workflows are another level of a mess, not quite a news, is it? A peculiar case of my little struggle with chaos - devops does not have capacity to assign me the access to some db. And I have to deliver little code, involving the other code (not mine), involving that db. Yay, workaround - I do have an access to deployment pipeline, which spins up the container with the access to that db. So my journey is like a hangman game - assume, change code, deploy (10 minutes), check. Trying restart the app within that container relaunches the container. So, this is the way. And I have some time to write this grumpy post.
Great Irish one.