I thought it was easier to find a possible cut in a MoE where the amount of data transferred between layers is very small (kilo bytes) while in dense architecture this is much harder?
I consider that an unnecessary leak of private topology, it's not info that needs to be public. IPv6 addresses are not really ever internal, and an exploitable vulnerability in a firewall product leads to the possibility of a directly addressable target. Guessing a v6 address without a name pointing to it is impractical at best.
Same. And now, my AI agent uses Vagrant to validate it's ansible yaml before it creates pull requests. It even does `vagrant up` inside a running vagrant when making changes to Vagrantfiles.
Similar things happen in hiking, people who shouldn't be there get encouraged by by how accessible information is to do things whereas before (most) people got info from someone knowledgeable.
May be it all depends on how AIs can help us solve the problem. By now, they are more capable than many people that I deal with. So, may be the problem can give us the solution. May be they can figure out a way out of the mess. And sorry to say this, but China and the US are not going to solve this problem, they may even get it worse. Every country is controlled by politicians.
As expected, higher altitudes, higher mass, and lower surface areas correlate to longer deorbit times. It looks like altitude has an extreme effect on deorbit times, as you can see the 100 KG satellite (solar min) deorbits in a little under 2 years at 400 KM, but over 15 years at 500 KM. So 1.25x the altitude results in 7.5x the deorbit time.
Stuff at 800-1000 KM can take centuries to deorbit, and that's within both NASA's (under 2000 KM) and the ESA's (under 1000 KM) definitions of LEO. There is a definition for VLEO of under 450 KM, which would have fairly short deorbit times, and therefore a relatively mild Kessler Syndrome.
I asked for proof of claims and I just get "it absolutely does" and a bunch more assertions, and word salad like "the validity of the scientific method is ultimately a metaphysical extension of our intuitions about inductive reasoning and causality" -- there's nothing "metaphysical" about it ... the scientific method is a disciplined application of an effective process of discovery. And ""Our intuitions" is why I said metaphysics is tautologically the domain of humans" is more word salad -- "tautologically" has no business in that sentence, and none of the rest of it makes any sense either. Humans have multiple domains, not just "the" domain. Metaphysics is a field of study that non humans capable of studying can also study. Metaphysical facts govern everything, not just humans. etc. I consider this sort of junk to border on bad faith ... it certainly isn't of any use to me.
> There is no scientific proof that the scientific method is valid.
The scientific method is provably effective in a lawful world. That the scientific method isn't provably effective is because the world is not provably lawful.
The Orthogonality Thesis (by Nick Bostrom) says that intelligence and ultimate goals are independent. Those non-instrumental goals can’t be stupid nor right or wrong. Increasing intelligence will not change the goals only the capability to reach them.
> And this time around, it seems like only a minority that are billionaires will be able to move forward, and we all will be left behind.
I don't think this is true. Of course, rich people will always benefit the most from any technological advances. But there is no indication that the average Joe will be worse off in say, 20 years, compared to today. Medical advances alone coming down the pipeline will likely tip the scales towards future average Joe being better off compared to today. If I have to make a choice, for example: do I want to cut the deaths from diseases by half and fill the sky with Starlink satellites, or do nothing? I am picking the better medicice and Starlink-filled sky.
Anything up there needs to reflect as much as possible to avoid building up heat. That which it can’t reflect is absorbed and needs to be emitted as efficiently as possible. Vantablack would likely make it absorb heat readily and glow in the near-IR.
For war it is. Drones and other unmanned aircraft are the future of warfare. That's the whole reason why every country now heavily invests in low orbit sats.
It's not about consumers. Also not for spacex. Defence contracts are zillion times more worth. Once you are in you reach the end level as a business.
I've spent tons of time in NYC, Barcelona and Paris and never ever encountered petty crime even as an observer. It certainly exists to some extent but this whole issue is a hype the kind of which villagers on the net like to argue about.
Companies take cultural cues from leadership. When you have a puffed-up sociopath who has never accomplished anything but lying his way to the top, this is what you get.
I'm both infuriated and worried that such a flim-flam man has put himself at the center of the U.S. stock market.
Their brightness is a mixture of a lot of things, including the huge PV arrays and the angle they have with the sun when they cross the terminator between night and day.
Starlink have already put a lot of effort into their satellites being much less bright than most satellites, including tilting their PV away from earth during the terminator crossing, so from what I've read you'll mainly see them while they're being deployed and while de-orbiting.
(Part of my still-expanding draft blog post about space data centres is to work out how bright a million much larger objects would look. If they were in the orbit with the most sun, that's a terminator-following sun-synchronous orbit, which is maximum brightness).
> LLMs... impedes the ordinary process of theory-building
As I have said on here before, I actually really enjoy LLMs for code understanding precisely because they are imperfect.
They're good enough to point you in the right direction, they might even see something you miss, but you can't trust anything they say.
This forces you to deeply understand what you're looking at, and to constantly re-evaluate your mental model against the code you're reading.
This works the same way as "writing forces clarity".
I could get to the same place on my own - I did this for many years before LLMs existed - but I feel when an LLM works well it gets you to the same place (or better) quicker.
(as idiomatic code generators I have mostly found LLMs to be poor, even Opus, but this comment isn't about new code generation)
Yes. I'm moving the goal post a bit here, but I was actually thinking about design mistakes and dangerous constructs - the kind of thing you do when you write in a "hacky" way - either deliberately or because you don't know better. Although these are not directly under the scope of type checking, a more restrictive language can have a positive influence.
Most bots on reddit are too cheap to use LLMs, from when I moderated most seemed to just have harvested a bunch of old comments and links that they reposted.