"Timescapes" demonstrates recovery from backup.

Behold the man as an off by one error made me laugh. :)

I should follow these with pictures of the same hotel treadmill in Beijing where I run. View is worse but weather is consistent.

We can't see both your faces at once. Still suspicious. Need to do an eyebrow count to verify identities. (I don't think you look *that* alike.)

The AI Revolution in Math Has Arrived | Quanta Magazine AI is being used to prove new results at a rapid pace. Mathematicians think this is just the beginning.

Can confirm the headline is backed by the text

Be honest though, I mean you definitely would explore it. What sane person wouldn't?

In v2 the autonomous vehicle will engage you in automated borderline racist banter through your journey. While testing a reform councillor will sit silently as a safety racist in case left or liberal ideas are introduced.

An academic conference in Finland had a sauna on the programme as the Finnish organisers believed it would relax people. I didn't go to that conference and twenty years after still feel a shudder of horror at the thought of a sauna with work colleagues and senior people I am trying to network with.

I don't think economically Open ai can majorly subsidise typical users. If they subsidise even $5 a month that is $5billion a month. I don't *think* they have that burn rate.

Ok this is a "vibes based" guess but my guess is that a year ago the typical user did not get close to using even 10% of the allowance paid for. If that is true you can subsidise the big hitters who use 100% but you are caught out when typical users up usage.

I agree with you here Arthur. I suspect it is a change in usage patterns. Narratives have changed and usage with it. In the last six months people complain about running out of tokens and usage limits loads more. It is a change in the pipelines they deploy.

If you're curious curtains hung on a wall cannot be recycled. A farmer proudly holding a bunch of carrots is food waste.
The project was able to get great accuracy on test/train image classifier methodology. Turns out it's easy to tell curtains from farmers.

Not much of a surprise. Some kaggle datasets are genuinely awful. I reviewed a project about using a camera to classify recycling (photo it, the app answers "does it go in recycling"). ~25% of the images were of curtains hung on a wall. Images of "food waste" were farmers proudly holding veg.

Honestly so much. One of the reasons I hate budget airlines is the extra 30 minutes of yapping as they try to sell you stuff.

If you think about it an activity which is only profitable if electricity is cheap is really very marginal as a business. ("Oh, electricity went up in price now we are out of money.")

In other systems though the profit is sufficient that it is more cost effective to just keep blazing away at maximum speed and power no matter what the electricity cost. You are making sufficient cash the price of electricity is a "tweak".

The problem is it is only for a few days a year and in many things are heavily optimised for time of day usage already. Data centres for example do a load of stuff to time shift load and compute to take place where there is cheaper rates of electricity.

I suspect unscientifically the increase in uncertainty here is crucial to "vibes recession". It is easier to cope with zero wage growth if you are at least certain you will keep those wages. Zero wage growth coupled with "I might lose my job soon" is horrendous.

I don't disagree. I would classify them all as "near zero growth" even after 5 years 0.5% growth isn't going to feel much different to -0.1%. Five years you are barely above 2.5% better off. It doesn't change how you Iive day-to-day even compounded.

Or are you simply saying that it could be technically a very shallow recession?

Not sure I understand your point here. I don't think there is a tipping point at 0% given the value is part of a distribution. Measurement errors are not great obviously but I don't think much changes if 0.5% measured is actually -0.1% for a while.

Negative power price is the thing to look for if you want to know about it.

energy.drax.com/intelligence...

It's why the final mile for reewables is such a pain unless you have (say) big hydro/geothermal available (all places with > 80% renewable have something like this.)

This isn't some weirdness of privatisation, it's physics/electronics where the core problem is. Distribution across the grid is a problem and "shedding load" is a problem and stopping/starting things is a problem. If big batteries were a cheaper thing life would be easier but they aren't.

The thing is, there sort of aren't. There is a point where electricity is free and, indeed, where the price is technically negative. Anyone who can use electricity is or is charging something. It's a big problem for renewables and has been for a few years now.

Type 1 diabetes cured in mice with gentle blood stem-cell and pancreatic islet transplant An “immune system reset” cured autoimmune, or Type 1, diabetes in mice in a Stanford Medicine study. The approach may be useful for other autoimmune conditions as well as organ transplants.

Luckily diabetes is cured (in mice)

med.stanford.edu/news/all-new...

These are wild examples of rejecting students

(original article is from The Times)

Now added to our Gendered timeline of technology
cs4fn.blog/2023/05/07/a...

I did a physics degree. Relativity interested me, I read around the topic and studied I still know the core of it 35 years later. Solid state physics I found a bit tedious, I crammed just about enough to scrape a good exam mark. I can remember a scattered handful of the more interesting concepts.

Complete agreement here. As a student there were subjects I cared about much more less where I simply crammed for the exam rather than tried to learn it. I did not really want that knowledge having headroom. I needed the exam mark but I did not want the knowledge.

As side product here is that AI could allow the academic community to develop higher and enforceable reproducibility standards (at least for insilico research) because of the lower cost of reproduction studies