While spinning, the blades store a miniscule amount of kinetic energy.
After removing power even that small amount ends up as heat through friction ( both in the bearing but mostly in the air turbulence). And the blades end up in the same zero energy state: sitting still.
Most of that energy gets transfered to the air that's being moved by the blades, and who knows what that air does eventually. And we're not even talking about the plant growing light that might be sitting in my room near my house plants literally creating new life from electricity.
We do know what that air does eventually. Given no further inputs of energy, it swirls around generating friction, raising its temperature (heat!) as the currents slow down to nearly nothing.
Overhead high voltage conductors are not insulated with a coating, probably for many reasons but certainly for cost and heat dissipation.
That means the path through the air to some conducting materials needs a certain distance, and that even when wet or iced over or whatever can happen up there.
This view is just very extreme, it is much less zig zag. It is just mounted to the wall at the high points and slack in between. Certainly there is also a reason for the exact amount of slack like thermal expansion.
The money you buy stock with l goes to the former/selling shareholder, which is most often not the company. It is possible the company is holding its own stock and selling for cash, or emitting new shares for cash, but that is much much rarer.
Remember that you are supposed to replace the entire thing because the other components like the sensor or simply capacitors also age. It is a very cheap safety device and simply not worth taking any risks by stretching it to say 15 years instead. The proper way would be to replace them while they were all still fine by making a note in the calendar.
There are two cases:
Your products are faulty and at least one has not made their intended 10 year lifespan. I'd change them all for better ones.
Or
They have reached their lifespan and you only noticed because the first one failed. I'd replace them all.
Fair point, although with a 400+ year half life in the americium source in the detector, I am skeptical that a new smoke detector would be any more reliable than a very old one.
I would think testing them regularly - especially with simulated smoke as done in professional situations, or in my case via bad cooking, is probably more effective than regular replacement on a schedule to ensure they are always working.
If dealing with something that follows a Poisson failure probability distribution with a fixed percentage probability of failure per year (as is the case with most electrical components), regular replacement only makes the system more reliable if you are unable to test it, otherwise it makes no difference.
With a few rare exceptions, is largely a myth that replacing machines or technology at regular intervals increases reliability- people incorrectly assume this to be true, based on observing that most failures happen to things that are old, but this is merely because they spend more time being old, not because the rate of failure per time increases with age (it almost never does). Testing and redundancy are more effective and cheaper.
Now, everything I am saying would be wrong if smoke detectors indeed have components besides the alpha source whose failure rates are known to increase with age, and actually age out within a decade or so. Like you mentioned, this can be the case with electrolytic capacitors as well as non solid state relays. However, I wouldn't be surprised if the lifespan of capacitors at the low temp and low voltages in a smoke detector wasn't 50+ years.
> Couldn't LLM provider just fine-tune their model for these tasks specifically - since they are static - to get ad value?
They could. They would easily be found out as they loose in real world usage or improved new unique benchmarks.
If you were in charge of a large and well funded model, would you rather pay people to find and "cheat" on LLM benchmarks by training on them, or would you pay people to identify benchmarks and make reasonably sure they specifically get excluded from training data?
I would exclude them as well as possible so I get feedback on how "real" any model improvement is. I need to develop real world improvements in the end, and any short term gain in usage by cheating in benchmarks seems very foolish.
It sounds very nice, but at the same time very naive, sorry. Funding is not a gift, and they must make money. The more funding they get - the more pressure there is to make money.
When you're in charge of a billion-dollar valuation company which is expected to remain unprofitable by 2029, it's hard to find a topic more crucial and intriguing than growth and making more money.
And yes, it is a recurring theme for vendors to tune their products specifically for industry-standard benchmarks. I can't find any specific reason for them not to pay people for training their model to score 90% on these 113 python tasks, as it directly drives profits up, whereas not doing it will bring absolute nothing to the table - surely they have their own internal benchmarks which they can exclude from training data.
> If you were in charge of a large and well funded model, would you rather pay people to find and "cheat" on LLM benchmarks by training on them, or would you pay people to identify benchmarks and make reasonably sure they specifically get excluded from training data?
You should already know by now that economic incentives are not always aligned with science/knowledge...
This is the true alignment problem, not the AI alignment one hahaha
They cannot be found out as long as there is no better evaluation. Sure, if they produce obvious nonsense, but the point of a systematic evaluation is exactly to overcome subjective impressions based on individual examples as a notion of quality.
Also, you are right that excluding test data from the training data improves your model. However, given the insane amounts of training data, this requires significant effort. If that additionally leads to your model performing worse in existing leaderboards, I doubt that (commercial) organizations would pay for such an effort.
And again, as long as there is no better evaluation method, you still won't know how much it really helps.
This market is all about hype and mindshare, proper testing is hard and not performed by individuals, so there are no incentives not to train a bit on the test set.
And assumption like that are probably a great source of accidents, our mind needs to take some shortcuts like that and isn't always right. Grandmas car got sold last week and daddy is alone in the car and late for work and will be racing to get in front of you.
I'd rather have a computer keep track of everybody just the same but with millisecond reaction to all changes. Something that I can't do lacking eyes all around and processing power.
I'd guess that they used to have very good discounted contracts but are not able to renew them anymore.
The cost they charge also includes the cooling, so you could see it as for example 30c/kWh for power, 15c/kWh for cooling to compare it to a consumer contract. Maybe they have published their ratio of that split "on average" before.
I do not understand why one would fetch stacks of paper from the bank if one believes toilet paper is the more valuable type of paper you should hoard and maybe also trade with.
Seriously, why would the end of fiat cause a bank run? I would expect it would cause a switch back to trading physical goods directly, or the emergence of another currency-like thing or good (like cigarettes might have been in the past).
When bank runs were common it was normal for a bank to go out of business while nothing really happened to devalue the currency so you'd essentially be racing other customers to withdraw first. The Fed's guarantee to cover accounts in the event of a bank failing stopped bank runs entirely.
Because fiat is created in 2 ways: central bank and bank. Cash in hand means it's central bank money, numbers on your bank account are the multiplied variant. Both are part of the fiat system.
After removing power even that small amount ends up as heat through friction ( both in the bearing but mostly in the air turbulence). And the blades end up in the same zero energy state: sitting still.
So it is correct that a 100% "end up" as heat