hazbot

Mass divided by volume encased by the event horizon?

Cool! If you end up wanting to expand beyond just the nearest weather station (forgive me if I've misunderstood your process), you could look into ERA5 - free Numerical Weather Prediction 'reanalysis' of past weather on a regular grid. openmeteo has some open source tools for extracting time series data from it.

But, although you get good spatial coverage, the drawback is 'the map is not the territory' - the model's representarion of reality doesn't perfectly mesh with the weather on the ground.

This is probably because of either poor sensor coverage, or a stale (old) forecast. Many weather services do not issue 'nowcasts' that constantly update with the latest weather observations (it's a hard and interesting problem), but rather a single forecast say 4x a day as the latest Numerical Weather Prediction model run comes in.

Fwiw, I agree with your bemusement and scorn - it's not good enough! (I say this as someone who has had roles where I issued these 'always stale' forecasts)

Depends, the "optimal" forecast can be very sensitive to the scoring metrics used.

E.g. Darwin in Australia's tropics - persistence forecasting (as you describe above, just predicting the weather the day before) does very well on a metric like 'mean absolute error'. But has no practical skill at forecasting a severe tropical cyclone (aka hurricane/typhoon)! Many are willing to accept some level of false positives and a higher mean absolute error, because the cost of a surprise cyclone is so devestating.

But sport is by definition set up so what is measured is the result we care about.

We would perhaps not care so much about the "vicious cycle" if the incentives were aligned with "good" research. If cancer gets cured, at the end of the day I'm not going to lose a lot of sleep over how. But the concern is the incentives are misaligned in favour of exploiting the metrics instead of "good" research.

model.predict() is pretty easy to call, but judging the validity of the model is still hard and very manual. Linear model is less flexible and powerful, but easier to analyse/validate.

Eventually, those who disagree will die out.

What a conclusion!

I think they are quite upfront about their values, and I suspect they would quite likely be OK with being labelled human-centric, as would I!

OP assigned -3 units of charge to the electron, so all works out.

Love to hear you're still teaching a decade later - I'm starting my Masters of Teaching next week to become a math teacher!!

What do you think the biggest mistake that people who don't last as a teacher make?

The paper is here https://dspace.mit.edu/handle/1721.1/111114 and is very accessible reading.

My takeaway is these ‘Harbingers of Failure’ have tastes not shared by the broader market, and so success only among the harbingers is unlikely to spread.

They also allude to an idea that overtuning tech products to early adopters is dangerous because they are possibly weird and unrepresentative of a bigger potential market.

How many plumbers feel Like most of the math they did was soulless and irrelevant, but don’t mind or even like that they got to do some art in school?

I read the first few chapters, then went off and just hacked on some math for the joy of it.

It was incredibly freeing for me at least to hear the message of ‘think deeply about simple things’ with the realisation I don’t care like I used to that there are other people are smarter and faster than me.

I recommend it.

Substack has been broken for me Firefox on iOS for a little while now.

Previous doomsday predictions must be wrong, otherwise we wouldn’t be here to say they were wrong.

But part of the reason the other doomsday scenarios were wrong is because people took them seriously and tried to mitigate the risks.

Fair point, defs should be mentioned by any serious article.

I have use cases where forbidding writes is a feature, and if it means more stability and performance it’s a trade-off that’s worth it to me.

I’ve been burnt by trying to use FsX as a mount-s3-as-a-readonly-filesystem. Flaky mounts, complicated workarounds for maintenance windows - if this lives up to the promise I can delete a whole module (and the cognitive overhead of reasoning about its state and behaviour) that we’ve written to manage a local disk to mirror an s3 bucket.

Every time I get a new windows device, I have to go hunting for the arcane registry settings to disable the start bar ads. It is really jarring when I use a relative’s computer with the default ad vomit.

In Australia we use hydrogen in our weather balloons. It’s essentially a safety/cost trade-off.

Start living on the smaller salary now - see how it feels to cut the luxuries, squirrel away the rest into savings.

I’ve been to a colleague’s ‘demotion party’ from team lead to IC in another office - they had no regrets.

Be prepared for employers to be nervous about employing you - some will be worried about you being too big a flight risk because you might still have an open door to senior PM roles. So plan to manage that fear on their side.

In Australia we refer to ourselves as part of the “western world” - despite the geography being nonsensical!

Also your approach is more honest than saying ‘keep science apolitical’.

I found the generation technique fascinating, but the images felt 'wrong' - I felt like I didn't know where to look to get an overview of the scene, and when I tried to focus on details it was like me eyes kept 'slipping off' because nothing would resolve itself.

I remember as a kid doing:

100/3

ans×3

on different calculators, and at the time being baffled at how they often said 99.9999999.

My android calculator app gives me 9.00000... and lets me scroll through the decimal expansion, it's zeros at least to 1000 places, so presumably the calculator's answer is exactly 9.

I wonder if our calculator apps are doing some symbolic computations? Or just have >>64 bit precision (which I presume is the default for excel/python)?

I think your analogy is pretty apt. However, even though I don't think you can give corporations a free ride just because 'people could choose not to use the product', this is not an absolute! 'People could choose not to use the product' does hold a lot of truth.

True, but that didn't stop me from finding parts interesting, and just generally enjoying the earnestness.

I like the focus on tools to assist the human, rather than automating merge conflicts. Even if you come up with a new clever algo that reduces merge conflicts by x% from where we are now, how do you decide that these two pieces of now non-conflicting don't interact with eachother to produce undersirable behaviour? Unless you solve that pretty well, I much prefer the pattern of keeping a human brain in the loop.

I was chatting to a speech pathologist the other day, apparently when they were being awarded their degrees, the entire class submitted the IPA transcription of their names to be read out at the graduation ceremony.

I don't think there's not a universal theory of AI for lack of trying - it's just really hard and non obvious.

e.g. physics and chemistry took centuries of empirical fooling around to develop the highly predictive theories we have now.

I think one of the points of this article is that we need to think deeply about what our empirical foolings around are actually telling us.