There are two pieces of nutrition health advice that have stood the test of time:

1. don't eat too much.
2. have a varied diet.

One incarnation of the first one is to eat until 80% full. Or to chew slowly, so your stomach (which is slow to realise) can properly indicate fullness.
One incarnation of the second is the food pyramid, another easy to remember one is 5 fruits and vegetables of different color per day.

If you come across advice of the form "you need this one thing in your diet", it's rubbish (see point 2). Similarly, "the most healthy" - does not make sense. There are many ways to create a varied diet.

Also, food has cultural and social aspects.

Agreed, instead of "Smelling this one specific scent ..." this should be described as "smelling a scent ..." , although even that might be misleading.

It seems part of Meta's issue is that they only have "ban" or "no ban" in their toolbox. Wrongdoing complaints have to be thoroughly investigated, and they err on one side or the other. Why not create more options, like:

* tarpit ban: require logging in again after each action - people can still use their account somewhat, for attackers it becomes annoying, rate-limit how many followers can see new posts (first 10 after 1 hour, first 100 after 2 hours, etc).
* hidden ban: hide all activity of a user to non-friends, unless it receives one like from a friend, shadow-ban messaging to non-friends.
* ...

Not exactly the same, but it seems to align with the findings here: https://martins1612.github.io/...
There appear to be high-dimensional latent correlations.

https://en.wikipedia.org/wiki/... is the anti-hero we need. I wish it was a government-run service -- bring your bricked device and you get a reembursement. Should be easy to check if it is was bricked by a brickerbot. Such a service should pay off easily compared to the cost of botnets.

To fight IoT botnets, Brickerbot used to take out such devices, make the issue visible and force the users to take action on security by replacing the device if they want the functionality. Not sure if there are any alternatives to effectively achieve IoT upgrades over time.

Julius Caesar had coherent writing.

IMHO the larger point of view of this class of problems is to allow multiple specifications. This introduces an ambiguity, which is fine and nice from some perspectives, but then you have to maintain this long-term.

A different example of the same beast is to introduce two ways to call a functionality in some library. Some codes will use one way, others another. Now you are stuck maintaining both ways. Diagnostic scripts become complex because there is more than one way to do it.

So there is a benefit to being stringent and avoiding ambiguity.

Potentially, this means the entire 4chan user interaction chain and all posts and their language will be available for download. Then AI researchers will use it to study social networks and language. Then we will get 4chan regurgitated back to us by AI tools.

Can't they ask the NSA for their backup copy?

I hope there is something like this for code soon https://slashdot.org/story/25/...

The other lesson is that a key unspoken assumption, that this is a single phenomenon with a single explanation, can be wrong.

So what is the plan to avoid astronauts going blind and dying of radiation poisoning?

Currently, 2/3 ISS astronauts develop eye problems after only a few months in space, due to pressure changes in the body. A 18 month journey will leave them blind on arrival. https://www.sciencealert.com/w...

The current solution to radiation poisoning is either lots of lead (expensive to accelerate and decelerate) or going there fast (current NASA strategy).

Living inside lava tubes below the moon's surface is, in comparison, feasible and emergencies can be dealt with -- but admittedly it is not as sexy as a dome under a clear Martian sky (with radiation from above and toxic soil below).

Exactly -- what is the false-positive rate of this approach?

If you throw trillions of data sets compacted into a 50-d cube, sure you will find some in the neighbourhood of your target. Probably it is not just one person, but many thousands.
Both machine learning and mass surveillance have to be gauged by the false positive rates, the false negative rates and cost (monetary and otherwise).

This article, "No encryption was harmed in the making of this intercept", may be clearer in what the Australian politicians meant: https://risky.biz/bannedmath/ There are solutions for companies outside breaking/weakening the crypto while still allowing law enforcement to follow through with warrants.