tfgg

The original post of 11% is referring to leaked numbers about _MFU_, which has erroneously been re-reported as fraction of GPUs being used at all. The parent post is trying to correct this misconception.

And one of the interesting (and hard) computer science problems is how to turn an arbitrary einsum expression into the most efficient series of operations (transpose, matmul, etc.) available in hardware kernels. This is mentioned in the post under "How Contractions Are Actually Computed".

Really sad - he was often the sensible voice in the UK media (e.g. BBC articles) when it came to security issues.

There was actually a lot of pushback against the austere aesthetic by government ministers - they wanted fancy looking photo banners and pretty things. This was pushed back on in the name of making something functional the prioritizes users.

Honestly this sort of climate & environment tokenism frustrates me - spending a lot of energy trying to reduce and elimate things that barely matter (or even actually make things worse by displacement), while ignoring the big ticket items. At best it's innumerate, at worst it's greenwashing.

Harry Nyquist isn't exactly an unknown engineer who doesn't have his own achievements, though - not sure why people are saying he would be fired in a modern company!

What was remarkable is how Facebook's Threads app jumped straight to being full of advertisers and hucksters - they didn't think that maybe the right way to bootstrap a social network would be to make it full of authentic conversations, at least to start with.

Why? At what level do you define "local" - country, state, city, neighbourhood, street, literally next door, just my house?

Social media owners wanted all the benefits of superscale communities with none of the responsibilities - it isn't surprising that moderation got worse.

Reddit vaguely has a workable approach with subreddits, but it's still high variance, and they have to discourage long-lived comment threads.

I'm not going to detail challenges in medical ML - the literature can do that. But just to mention one that other people haven't that goes beyond just precision-recall: algorithms can be biased based on variance in physiology (e.g. more accurate for men), and understanding how an algorithm is biased is very important for the person interpeting the information, who should be a trained doctor.

And it doesn't matter what his credentials are, that's appeal to authority. If he thinks this should be used and trusted by people for decision making, then he should submit it to independent peer review and regulatory approval.

So you should ignore it and see a doctor regardless.