Socialism with American Characteristics

Iowa's rolling farm fields of coffee

Indeed, I fully agree. The funny thing is, monthly numbers would help us move away from the distortions of the quarterly cycle. If key data reporting becomes frequent enough, you can't get into a cycle of "do adverse-numbers stuff early in the quarter and then cram positive-numbers stuff into the end of the quarter". You have to - *gasp* - just run your business normally.

Some businesses could still manage to switch to a monthly cycle, but anyone who deals significantly in transoceanic feedstocks/parts/goods shipments won't be able to.

BLS numbers aren't some sort of dark art. They're literally just the compiled numbers reported by companies. Numbers are what they are. To fight against jobs numbers is to fight against reality.

People get confused by the existence of revisions. The problem is that not all data gets reported in a timely manner. When late data comes in, it causes revisions to the earlier reported numbers, either up or down.

Firing the head of the BLS because you don't like what numbers US companies reported is just insane Banana Republic-level nonsense.

Yes, he fired the same person who was ultimately responsible for putting out crap numbers.

US reporting has always been the gold standard. Nobody has accused the BLS of "crap numbers" until Trump decided he didn't like them. It's is so way outside the norms it doesn't even resemble something that could conceivably happen in the US; this is banana republic-level stuff.

Yeah, as an investor, my reaction to this proposal was, oh HELL no.

TL / DR:

BLS chief: "Jobs are turning bad now."

Trump: "Fake news, you're fired. I'm appointing a January 6th rioter conspiracy theorist to head the BLS!"

BLS today: "Jobs were ALREADY bad!"

The fact that you take at face value a revision of Biden-era job numbers, immediately after Trump fired and replaced the BLS chief explicitly because he published jobs numbers Trump didn't like, is....

Well, it's certainly a choice.

I was thinking the same thing. What sort of an idiot will take US jobs numbers at face value right after Trump fired the Bureau of Labor Statistics chief after he reported numbers that Trump didn't like?

Integrated thermal stores with nuclear is more difficult (limited temperatures and limited dT on storage = big storage needed with lots of heat exchange surface area), but if they can make it economical, it could be a game changer for nuclear's problematic economics. The ability of nuclear to switch from baseload (not matching the demand curve, let alone the curve of really-cheap renewables) to load-following and even peaking could make the mean sale value of its electricity much higher, and make it much more useful as a compliment to renewables.

The mean July maximum temperature in Helsinki is 27,1C (80,8F), and the record is 33,2C (91,8F). Combine this with the fact that the interiors of large buildings accumulate internal heat. Cooling is of utility. Let alone for industry (most industry is fine with cooling with ambient water temperatures, but some needs cooling beyond that).

I was wondering why. It's so frustrating - JPEGXL is an excellent format. Fast, too, compared to a lot of the alternatives (like AVIF), and less patent risk.

Why not use the soft stochastic qualities of analog light computing

Yeah... at least during inference (training will be hard with this with traditional methods). Inference is extremely tolerant of noise. They should be able to use vanishingly short timesteps.

Of course, computing in memory requires a completely different way of thinking about software than the CPUs or GPUs that we know and love.

Which is... what they're doing.

This is an implementation of something I've been talking about for years. The brain is much more efficient than today's AI chips - not because "wetware" is efficient (it's terribly inefficient, with a huge amount of overhead and numerous chained loss steps), but because analog accumulation is efficient compared to vector math for AI tasks. It's like the difference between trying to determine how much water is in a container via simulating every molecule in the container, vs. just measuring the container. You want the laws of physics to do the "math" for you. For the input field of a neuron in a DNN, instead of multiplying two vectors (activations times weights), your first vector may be flow paths, and the second be the resistance to the flow on that path (in the case of light, optical channels and transparency, respectively). You then need a physical nonlinear activation function (with a bias) based on how much flow accumulated during the elapsed timeperiod, and its results needs to control how much flow leaves that neuron to the next layer.

Doing this with light seems like a very fast and efficient way to implement that - much moreso than via chemicals as our brain does it.

Developing such a hardware system for training (rather than inference) will be much harder with conventional NNs, though. It might require predictive coding networks. Though that would be pretty keen if we ended up switching to PCNs, as they have all sorts of great properties (including realtime learning).

[Citation Needed]

Literally, isolated hunter-gatherer societies still exist. You can literally go observe them. There are old people.

Going to argue that, meh, they have too much exposure to the modern world? Go look at early black and white photos from early western explorers in Africa, South America, etc. Guess what? Old people.

Societies have always had the old. They've commonly been considered "wise men" / "wise women", sometimes attributed as healers, those with spiritual power, etc. To repeat: once you hit middle age, your annual odds of dying are low. The problem is how many people never reach middle age.

I'll repeat: it is mathematically impossible for half of your population to never reach adulthood, and a quarter to not get based infancy, yet have your mean expected lifespan for an adult not be much higher than your mean life expectancy for the entire population.