Loong bio/acc

This is why you can’t have more than 99,999 atoms, more than 10,999 residues (-9,999 to 9,999), or more than 62 chains. Not really a problem until recently. To any modern-day computer scientist all of this is totally insane. But also completely unsurprising. We’ve done much worse in our own history (looking at you, WINAPI).

Then along comes Walter Hamilton. In 1971 he invents the Protein Data Bank file format for storing structural information about proteins. At this point, all our editors are still standardised to a width of 80 characters and so — naturally — he picks this as the width for each line in the PDB file format. I’m guessing. And today, it is still the most widely used file format for proteins. It has since undergone mutation after mutation and has been perverted beyond all sanity by scientists with the best of intentions, but very little computer science background. There have been a lot of attempts to supplant — mmCIF being a great example of what not to do — but it still reigns supreme.

Why is the PDB file format aligned to 80 characters wide? There’s actually a really marvellously stupid explanation for this and it starts in 1848 in China with the invention of the loom.

@VictorTaelin @DavidDeutschOxf Still raising?

@bzlwang @DavidDeutschOxf correct

“there are going to be errors … and the correction of those errors is more important than getting it right in the first place – much more important” — @DavidDeutschOxf (not on AI but a general requirement for learning and creation of knowledge) one of the particularly curious opportunities for Blend2 is that it creates new training data that’s not merely an LLM training on its own outputs. it outputs. it gets it wrong. SupGen helps fix it. and you down have training data that you can actively learn from.

People literally use supercomputers to model airflow in rooms where they’re storing those supercomputers. GPUs are a scam.

Hi, Lipinski? Yes, these maniacs right here.

My current pet peeve: LLM love fully qualified paths because of the way IDEs try to localise edits.

“A lonely berry Patiently awaits my fork Swiftly meets the floor” — Miyamoto Musashi, probably

Lastly, maybe all this big city life actually *is* killing Japanese people. And we are just severely misjudge how much it matters compared to the big things like obesity. For example, in Shiga (way more rural than Tokyo) the life expectancy for women is just shy of 89 years. So clearly the “Tokyo factor” makes a difference.

Secondly, on loneliness. My guess is that the impact of this simply hasn’t made its way through to the stats yet. Loneliness rates in Japan are highest in people in their 20s and there’s no way that will show up in *current* life expectancy stats. But in 40 years we may look back and realise this was a massive ticking time bomb that was ignored. It may happen *a lot* sooner though. Even though people in their 20s report loneliness at 47% other age groups aren’t far behind. Japanese people in their 60s report 46% which is the same number as far as I’m concerned.

Tokyo is the strangest juxtaposition between ultra-legit food and ultra-processed food. Average life expectancy here is 87 for women and 81 for men. This is *more than a decade* above the global average. It’s above the US And Australia (l don’t know if that’s saying much).

This looks like something some circa 2020 gen AI would make. Gilead absolutely cooking this one.