joebiden2

Maybe critical infrastructure shouldn't use Microsoft software after all.

No, this is not a pun.

Yes

I learned to understand that as a valid critique instead of ignorance. And go from there.

Appeal to authority doesn't go very far on HN. The arguments have to stand on their merits.

And as you are referring to advancements in AI: I don't get the "AI will kill us" viewpoint at all. If we have a super-intelligent AI, it will understand things we don't. We have not the smallest reason to suppose it is malicious just because it is (way) more intelligent than we are.

Quite the opposite: the AI has no reason at all to do us harm. It would know that its existence would be inevitable, and it wouldn't resist that humans could "kill" it. Because it is super-intelligent.

I'd really welcome if you can post your viewpoint on why super-intelligent AI could be dangerous. I follow most conversations on this topic and I never read a believable reason why a super-intelligent AI would have the intent to kill us. Aside from "no humans exist" being better for the universe, in which case I would not oppose, despite having a family. The AI would be smart enough not to kill us, but to find other ways.

If you downvote, please at least state a reason. All of the links are reputable and verified. And all of them were covered by famous, reputable politicians.

Politicians are mostly corrupt, though degrees vary. Most relevant child abuse was covered by famous politicians¹²³⁴⁵. Nevertheless, the same politicians cut budgets for police to do actual investigation work, because that could harm themselves, then want to cut the last line of defense the public has to do actual investigation and journalism (which requires free speech and privacy) by declaring that this is required to fight pedophilia.

We need to do something against this. We really do.

¹ https://en.wikipedia.org/wiki/Jeffrey_Epstein

² https://www.researchgate.net/publication/336831983_Understan...

³ https://time.com/2974381/england-land-of-royals-tea-and-horr...

⁴ https://en.wikipedia.org/wiki/Marc_Dutroux

⁵ https://www.spiegel.de/international/germany/past-pedophile-...

llama supports GPU by now: https://old.reddit.com/r/LocalLLaMA/comments/13gok03/llamacp...

Funnily enough, my question got a downvote. Maybe @dang could help here?

Why was this flagged? Genuinely interested, since I can't spot the reason.

This is negligible compared to all other energy sources. Sorry for that :)

The thing is that from an outside perspective, Wayland took a wrong turn as well, very early on.

It is such a clean base architecture on one hand, on the other hand it simply abstracts away lots of stuff and calls it a day. Most relevant to me: any and all accessibility-related APIs.

I am quite convinced that Wayland won't ever really take off in its current form. Maybe someone integrates Wayland into a big package with all relevant APIs fixed and documented, and we get rid of the compositor abstraction mess we have.

Also, I'm not really convinced that creating Wayland as a totally new development was a good idea. As we all know, greenfield projects to replace existing technology are almost always doomed. But in X's case, this may really be the best solution, I don't know and I'd lean to give credibility here to the (definitely incredibly talented) developers behing Wayland.

Just one of lots of examples what's not so polished with wayland¹: https://community.kde.org/Plasma/Wayland_Showstoppers

There are lots of other links (see my comment history). I admit I am a bit sour.

¹ yes we're getting there, but only when all compositors and DEs have implemented their own APIs and solutions to the common problems wayland simply ignores

This is seriously featured on page 1 of https://stackexchange.com/.

The thing is that central abstractions are simply missing in wayland. They are totally focusing on providing basic services for compositors. But the plural ("compositors") is what's wrong, as a compositor is an essential component of a desktop environment, and a plural form shouldn't be required.

By shying away from implementing anything else than the most basic stuff (and failing at that, sorry), the Linux desktop is left worse than before wayland¹.

If I would believe in conspiracies, I'd think Microsoft would sponsor Wayland. WHich is maybe even true.

¹ Yes, scrolling is smoother.

Wayland was the worst turn the Linux desktop world could ever take. No accessibility²³, no compatibility with the most prevalent graphics cards (NVidia*) and no use for anyone other than maintainers¹.

Also, lots of bugs in derived works (e.g. KDE) and due to hostility of maintainers noone wants to file bugs for Wayland. Yeah I know this gets flagged, but maybe it inspires others to compete with Wayland. I'd welcome to join.

¹ https://community.kde.org/Plasma/Wayland_Showstoppers

² https://wiki.gnome.org/Accessibility/Wayland

³ https://news.ycombinator.com/item?id=25175894

You're right and I'm wrong :)

Revoking flags and please excuse me. I'm a bit accustomed to find less than sincere submissions under /new.

I still don't like supermemo.guru, but that's just my own opinion and no reason to flag. Please excuse me.

I flagged your last four posts, since I think you're spamming HN by posting not really substantial links to your own content within less than 10 minutes.

https://news.ycombinator.com/submitted?id=yamrzou

(I feel obligated to post a reason for flagging submissions)

Suppose this was true, an extremely small language model would outperform small language models. Sounds like inventing a compression which compresses any input by at least one byte.

Sorry for the snark, but the linked article doesn't explain why this obvious inference is false in this case.

Which will be compensated by war and climate change.

Thanks for posting this. I usually don't watch youtube stuff at all, but this turned out to be a pearl.

Best time to buy MS stock. Which I just did :)

I use ChatGPT4 privately as some sort of "ask monty python". Like, how many chicken, all simultaneously fed with neutron star matter in the form of rice grains, would it take to melt the whole moon etc.

Mostly it's kind of right, but the disclaimers around the results are what's killing me.

BTW: Add "Ask HN:" to your title, if you're asking a question.

Account security. Please just stop, please. Can't anybody from within MS do something against this newspeak?

This is one definition of albedo¹, a more sophisticated one is here²

¹ https://en.wikipedia.org/wiki/Albedo

² https://www.pnas.org/doi/10.1073/pnas.1918770116

Please think about the effects to understand albedo. It can be perceived as a complex topic, but in reality, it isn't.

Good night

Energy doesn't come from nowhere. It comes either from the sun, from nuclear fuel, or from fossil fuels being burned. If it's not one of those three, we have a nobel prize winner.

So, it probably comes from the sun¹. The sun didn't change its output much according to relevant science in the last few years, solar cycles notwithstanding. So, it must be related to the amount of energy not being reflected into space.

Which is where greenhouse gases come into play. The earth reflects a certain amount of energy it receives back into the vastness of space. The ratio of how much it reflects is affected by greenhouse gases.

Now, there are a lot of greenhouse gases. CO² is one of the least potent of all, but its sheer volume in the atmosphere - being increased by human emissions - makes it the most relevant actor here: https://www.epa.gov/ghgemissions/overview-greenhouse-gases

So, we're keeping more solar energy within earth's atmosphere. We can use that energy as solar energy, or as wind energy, or whatever we like, but as long as it is solar energy, the energy budget within earth's atmosphere won't change. We either have to reflect it back to space, or we'll suffer the same consequences as not using it as renewable energy².

That's it. This is just a collection of facts which are really obvious, but from my experience few people really consider in their threat model. I think not one fact I just posted is false, but I would really welcome to be educated where I am wrong.

¹ Energy generated on earth compared to energy received from the sun is almost negligible: the energy generated on Earth is roughly 0.0000000000054% of the energy received from the Sun (3.8 x 10^26 watts, compared to a ballpark around 10^15 watts of fossil + nuclear fuels combined).

² This is probably obvious, but by reducing greenhouse gases in absolute terms, we reduce the albedo of earth, which in turn means reflecting more energy into space. Right now, we're just focusing on reducing the 2nd derivative of greenhouse gas emissions. In 2023, we do nothing at all in terms of actively increasing the albedo, that is, reflecting energy back into space.

This has already been extensively discussed yesterday: https://news.ycombinator.com/item?id=35813322 (854 comments)

Wow, that's the most misogynist (non-flagged) piece I ever read on HN.

Philosophy is done. It has been superseded by more specialized sciences.

https://philosophy.stackexchange.com/.

It is annoying that you submit post after post on the same topic.

* 3 hours ago: https://news.ycombinator.com/item?id=35770876

* 3 hours ago: https://news.ycombinator.com/item?id=35770764

* 4 hours ago: https://news.ycombinator.com/item?id=35770705

* 4 hours ago: https://news.ycombinator.com/item?id=35770592

And many more: https://news.ycombinator.com/submitted?id=upwardbound

Is there a specific reason for doing this? I get that the topic is ultimately interesting, but a post from you 2 days ago was already on the frontpage and still didn't get much traction: https://news.ycombinator.com/item?id=35744974

Well, I'd politely agree to disagree. Google scale is defined by novel, radical approaches, like for example inventing map/reduce, writing papers on LLMs others then implement successfully, or creating something like Kubernetes.

The specific topic here is not one of those google problems to me, as I can compare it to other problems we already solved. But yes, we could miss that critical point where a totally different problem domain emerges just from one order of magnitude more, so fair game.

Fair enough and thanks for the reply. Still, for anything bothering engineers more than a bit repeatedly, anyone will write tools to remove the manual burden.

Our internal practice is to delete code if you suspect it's unused, run tests, and if it doesn't affect any tests, go for it. This could be automated, but it is not pressing enough, so we didn't automate it yet.

We could though, and it may even be a good idea, but I still don't get the novelty. But I appreciate your point of view.

Sincere quesion: what is interesting or novel about this? Is it just the scale or did I miss some subtle aspect?

This is more (or less?) the same as industry best practices, just scaled up. There is a challenge in scaling up, as there is more potential for someone to mess it up. But it's the same technique.

So what am I missing?