Sometimes at work I get slop emails from some VP or other, and the problem isn't the tone or use use of AI per se, it's that 2 sentences of meaningful content is diluted into 3 paragraphs for no discernible reason.
StarLink tops out in the 300mbit range, and has anaemic uplink speeds, whereas most fibre rollouts offer symmetrical gigabit. The latency over StarLink is also 2-5x higher than rural fibre.
StarLink beats the hell out of the copper phone lines / DSL connections a bunch of these rural communities were running on before the fibre rollout, but it's no match for proper fibre. And assuming France has price controls like Spain, the fibre will be cheaper into the bargain
- The obvious political point: that ISP is subject to laws in a country that is testing its legal framework and diplomacy in new (some may say unpredictable) ways.
- The first rule of public funding is that the money goes back to the public. Roughly all (if not all) of the jobs for Starlink go to the US, whereas fiber installation in France employs French people.
- Paying for Starlink is a terrible investment because the entire constellation is disposable. Every satellite will burn up in roughly 5 years and the whole network would go dead without more launches. Fiber, once installed, is at least serviceable.
- Another political point, but the only company that can launch replacements economically is also subject to laws in that same country.
Starlink hands an awesome amount of semi-soft power to whatever country controls it. It doesn't really matter what country that is: it's a huge strategic liability to anyone who doesn't hold the keys. Until there's at least another player in the game I'd expect a lot of countries to be wary.
Ironically, that could mean Starlink/SpaceX really needs a competitor before it can become a universal solution.
We’re on about our third attempt at rural internet in Ireland. The first two involved wasting a lot of money subsidising dubious cellular and satellite stuff which was arguably marginally good enough 20 years ago, but not now. Thankfully on the third go the government decided to stop screwing around and just do 10gbit-capable fibre to all rural premises.
Ultimately if you half-ass this you just end up doing it again later.
This is one reason I’m kind of baffled by spacex’s valuation; its long term market would seem to be countries who are too poor to build out fibre. It’s unclear that there’s a trillion dollar industry in that.
If you need to get internet to rural places quickly then such services are great.
But if you are trying to get something that could last for decades and handle the expected growth in the speed and bandwidth that the average family might expect then fiber is the way to go.
You don't necessarily need to go right to every rural house at first. There are many places where fiber to a terrestrial wireless access point and wireless to the homes would work well.
Fiber can then be expanded out from those access points later.
Satellite services make an end run through a lot of bullshit, that’s true, but the long term costs of fiber are overestimated because cable companies want a very fast profit. Whereas thousands and thousands of satellites will need to be replaced year after year and that rate could increase considerably if Kessler syndrome develops.
At some point I did some research and the Atacama was suggested to me as the best vantage point on earth for simple skygazing. Definitely on my bucket list.
Simple stargazing seems to be selling it short. It's the best seeing on the planet. I can do simple skygazing from my backyard. With a four hour drive I can get to a spot for pretty advanced skygazing. The ESO is building the largest telescope man has ever created in a spot you describe as good for "simple skygazing".
The skies are so dark and clear that some of the indigenous people had "constellations" made up of the dark patches in the sky instead of connecting stars.
I have many concerns with this kind of funding model, but I don't think the measurement problem is so serious. Performance incentives in education typically reward improvement of the student cohort relative to how it was performing the previous year, or even use value-added models that use multiple past years to predict the student trajectory.
It's not an extraordinary claim at all. They're just training the model on recent data. What reason do you have to believe that the model somehow learns significantly worse than it did during the original training run?
It's an extraordinary claim to go from: "the model has some knowledge past it's claimed cut off" (which is common sense given they've trained new checkpoints and has always been true even before those)
To "the cutoff date is now April 2023"
Based on the conversation posted.
The cutoff date is where you can deeply probe the model's knowledge and expect consistent performance: showing that some major headlines got captured in a recent checkpoint is nowhere near that.
_
If anything I'd take not the headlines but the actual disclaimer as something interesting, because it might indicate a future checkpoint is going to use that cutoff and we're seeing signs of their RLHF based on that... but it's nowhere near as impactful as the cutoff changing.
I love cities that embrace street food. My home (New York City) is well-regarded for street food, but I found it that it's nothing in comparison to Mexico City where a vendor seems to dot every single street corner.
With our density, New York could have a much richer street foods scene if the permitting and regulations allowed it.
Go to China in the summer, especially in southern China where it is hot in the day and only bearable at night, night markets with just lots of street food are pretty intense.
Unless you want to be easily exposed to dangerous elements like gutter oil or fake food, which is prevalent in China, I recommend other food stall cities like Singapore or Taiwan.
Seconded. Taiwan is the superpower not just in semiconductor manufacturing but also in street food. Many quirky snacks from there repopulated the culinary desert of post-Maoist mainland, and bubble tea spread all over the world.
Gutter oil has been exceedingly rare in recent years in China due to extremely harsh crackdowns (literally death penalties were handed out). You will be fine these days in bigger cities.
Not sure why you would take Chinese government's word on anything, when they still refuse to accept blame for releasing covid to the world and killing millions, and giving long covid and removing smell and taste for many more.
