novariation

Generalization is standard lingo in machine learning – it's about going from known data (training set / test set) to new data which shares the same underlying patterns but wasn't available at learning time.

Taking the posts simple polynomial example, it would be going from

4x^2 + 2x -3 = 0

to

2x^2 -5x +2 = 0

or more generally to

ax^2 + bx + c = 0

Did you read the article ?

They were severely understaffed, and essentially were forced to do a half-assed job because of a storm. Reading the article, it feels as if they didn't even know if the explosives were going to destroy 1,2,3 or all pipelines. They were just 6, with only 4 divers, one of which was a 30yo woman.

They didn't have a large enough time window, they didn't have enough funds, so they did something halfway.

Fun fact: my (french) company audited the french navy and found out they share important nuclear submarine maintenance data as excel files through email.

I have to confirm this as a nerdy french person: I don't like sports and it always felt awkward because it's not something I enjoy, people keep asking for it and it makes me uncomfortable. I tried hard to fit in though: I tried judo, tennis, swimming, running, rowing. But couldn't make any stick, 2 years each (except judo for a few years).

French moms really force you to exercise and my immigrant dad always felt weird about it.

To be very honest I'm tempted to downvote this , not because I would disagree if this was the topic at hand, but because it seems like a deranged rant that's very loosely related to the comment it's replying.

I believe when they say the squatters have a "right to housing" it means "right to use a unit of housing that's not theirs as if they paid for it", which essentially means doing what you suggest would be illegal.

The reality of the legal dimensions is probably much more complex though, but it's helpful to have a simplistic mental picture of what's expected to happen.

This seems to be a copy/rewrite or the 3b1b commencement speech ? [0]

[0]: https://youtu.be/W3I3kAg2J7w?si=7_4z-W5PmnBo302e

That's a great catch ! Thanks for this. For anyone wondering, 25 years is pretty common as a baseline for PV [1] but a good amount of them should be able to go for longer, say at least 30 years. As a comparison, the average nuclear power plant in France is 37 years old, and the french authority considers obvious a 50 year life expectancy, and 60 is being floated [2], but no idea where they got 80 from probably from the US extensions being floated [3] — 92% capacity factor are official measurements though, it's not unrealistic [4]

[1] : https://www.cnet.com/home/energy-and-utilities/how-to-make-s...

[2] : https://world-nuclear.org/information-library/country-profil...

[3]: https://www.eia.gov/tools/faqs/faq.php?id=228&t=3#:~:text=Be....

[4]: https://www.energy.gov/ne/articles/what-generation-capacity#....

More details: You need practicing, so you need experiments/computers/supplies/projects — this is true both in engineering and art school, any time you do anything that's not completely pen and paper.

You also need to divert time from your researchers to teaching activities. You need to pay for classrooms, that's real estate, plumbing, electricity, etc.

Keeping up with the state of the art in research isn't cheap, and using part of those funds to purely teach and not compete with other countries is an investment that's not neutral.

I'm using french costs, these are paid 90% by the government ! The underlying issue does not disappear.

I've got a weird relationship with market socialists because I generally think they're much more sensible and convincing than the average "destroy evil and create utopia" socialists, but ultimately it's hard to understand whether their efforts to rebuild society would yield much better outcomes compared to European style welfare capitalism with more taxation and welfare spending.

I suppose they're trying to build a grand theory of social market economies / social democracy, which is nice.

Even if you dislike markets as a mechanism, education is very tricky because it's very very expensive, and prone to important moral hazards when it's free.

Think your passion is economics or psychology and drop out after 1.5 years ? You've just cost the equivalent of 8 years of tax money from the average joe (ex: France median after tax income is 25k, roughly 2k of tax money, each college year is 10-12k).

It turns out there aren't enough average joes to give every kid the opportunity to try multiple studies, so we have to share the burden, at least a little bit. The average joe can pay 70%, and the students 30%. Or the average joe pays 90% of the first curriculum, but if you "reset" once (change majors without transfers), then the average joe no longer pays that share and only pays 50%. But you have to accept that education isn't free and there should be some incentive, even if small, to make sure you're picking something you'll commit to and make it work.

My guess is they're talking about regional monopolies ?

There's generally a misunderstanding from left wing voters — when people say "markets" they understand "crony capitalism and monopolies with regulatory capture" when most other people just mean "fair competition between suppliers".

We'll all agree the second option isn't easy to get, but the disagreement often comes from the fact one team thinks it's impossible or even undesirable, and something more radical should be preferred.

OMZ is definitely super slow, but there are a few competitors trying to speed start times up, with nice results (50% to 80% reduction).

For style just use starship.rs

It's funny because I just got out of a professional training session where the instructor told us sale-leaseback agreements are a common business practice and helped a few companies he worked for secure long term growth.

Let's use a metaphor HN might understand: sale-leasebacks are a bit like migrating to the public cloud from an on-prem setting, and in some cases the government pays you for migrating to the cloud. It helps you balance capital expenditure among other things.

31k citations, h-index of 81 [0] Cambridge professor & seems to have very serious experience in building models of climate variation across time. He recently published a pretty alarming finding about extreme weather events, so he's clearly not a climate change denier [1]

[0]: https://scholar.google.de/citations?user=3hW6AkMAAAAJ&hl=en

[1]: https://www.tortoisemedia.com/2024/05/20/dont-believe-everyt...

Honestly with the amount of degrowth friendly opinion pieces nature published in the last 2 years, I'm surprised they take the complete opposite approach. They're really trying to open the debate to very broad and different viewpoints, and I think it's healthy. I don't know enough about this researcher to understand how powerful/influential his comments might be

These two comments are a perfect summary of why I stopped debating politics online with anyone.

A good 50% of the online crowd, on both sides (left/right) are *absolutely delusional* when it comes to political perception.

On the right you'll hear Biden is too far left because of the massive public spending, the aggressive NLRB policies favoring labor, the support to unions, anti free trade policy, neo-brandeis antitrust, trans rights speeches and wealth tax support.

On the left you'll hear Biden is actually a neoliberal or center right, because of his stance on Israel, his belief that welfare capitalism is better than socialism, his belief in economic growth and innovation, his past ties with the banking sector, his 1994 crime bill and support for the police, pragmatic climate policy, etc.

When debating online half the time you're just trying to filter out completely delusional beliefs, and it's tiring.

I work a lot with F500 companies which have been around for decades, and that hurts because it's so true.

I hate XLSX with a passion. The main issue is that as a file format it's too heavy, slow, clunky, and does weird data type conversions, but that's not the only problem.

The main problem of excel is "mission creep". It just does too many things in a half-assed way. So when you get a bunch of excel files to process for your data analysis task, you can't be sure someone didn't put a bunch of charts, merged cells, used formulas, used strange currency or unit formatting gimmicks, structured the data in a weird way to make it visually appealing to management, and so on. It's all so tiresome.

If I liked playing devil's advocate, I wouldn't be annoyed at your answers.

To be honest I started a few years ago interacting with comments I would otherwise ignore because I thought it was harmful to keep yourself in filter bubbles, but I think it's pointless and I'm mostly having negative experiences that I don't need in my life.

To address your points: I'm not a marketer, I've got a STEM masters degree and I work as a data analyst, never worked in any marketing related job so far, no interest to do so.

You bring a decent point that often advertising isn't something people want, it's rather something they have to accept because they're paying either zero or less than the producer wants/needs for the service they're using, which I think is a fine point. We could go back to a point in time where most things were not financed by ads: either they're financially viable or they don't exist. That's a tradeoff, and companies like Netflix tried to move away from this and found some success. But it seems to me in a lot of situations most people are fine with the old radio station model where you get a few ads every now and then but don't pay anything.

If we decide to go this way and make it impossible to fund most of your company or activity through ads, then we'll have a lot less ads, but still won't reach zero ads. Think of the very old ads in the newspaper kind of thing. People need to send information about things they're selling or jobs they need, or events that might interest people. You're bound to have some advertising, you probably just mean to say you would like to have much much less than the current level.

The comment I'm replying to is very black-and-white: "X should not exist".

My answer is simple: there might be a reason that X exists, although it could take different forms and have different rules and methodology.

You reply with an equally black-and-white point : "so you're saying they're saints and no one would be able to do anything without their existence".

This is lazy and boring, do better.

Every consumer, and every human being in society faces the "exploration-exploitation" dilemma [0] weekly or daily. They've discovered a set of things they like and dislike, either through family, friends, circumstances, or marketing ploys in the past. Finding new things they like isn't easy. You need external influence. Without these you feel life is boring, repetitive, predictable, lacks texture or excitement.

As most honest people will tell (one recent example I've stumbled upon could be independent game developers) - you could have a great product, but sadly most people will not know about it or will be "too busy to try it", and will prefer sticking to their habits or past experiences.

How would you go about solving this issue in a reliable way, that's not prone to similar distortions and hijacking as advertising and public relations ?

[0]: https://en.wikipedia.org/wiki/Exploration-exploitation_dilem...

That was my thought as well, I believe it's close to the "let the data warehouse sort it out" solution: make the current state separate from the "history" - in their example this supposes a data warehouse, but it could just be separate tables or databases.

The main consumer of Huawei is not the privacy-obsessed western HN reader, but a third world customer that has a fairly lower expectation of privacy compared to the average American.

Think of cheaper phones in China, India, Pakistan, South America, and so on. Huawei can aim for literally billions of customers without these features, just by being an all-around useful OS in every day life for the people who usually don't have phones as part of their daily toolbox yet.

Yes it's all about tradeoffs, but I think it's useful to accept (e.g.) in the laptop case that having thin and portable devices isn't necessarily a bad thing and it's likely that a good chunk of consumers will want that even if we create economic incentives and policies to encourage recycling and reuse.

I personally don't think it'd be bad to have a larger chunk of buildings be more versatile but you'd lose some of the economic efficiencies (agglomeration effects and positive spillovers) from the concentration of firms in business districts etc.

In every scenario you win some and lose some, and it's not always clear to me we should go all in on any of the bets.

> "for public policy it's easier to frame it as traditional food vs industrial junk"

This isn't research, this is simplistic populist fear-mongering.

It's a bit like saying "the economic research does not want to go into the details of the different economic benefits and drawbacks of immigration, instead it's easier to frame it as the smart educated whites versus the hordes of dirty foreigners coming into the country" and expecting this argument to be taken seriously.

I don't understand your point:

Construction companies build what they were told to build by developers, and developers decide to build stuff by guessing the balance between financial viability and usefulness to users.

In the past there was a need for office space, and there was specifically a need for buildings which are cost optimized for office work. So the developers designed their plans and investment that way, and the construction companies built them.

What is your idea? That it shouldn't be possible to build something that's expensive to convert to residential housing ? Then that'll just make all other types of buildings more expensive, so you'll hurt manufacturing, businesses, public institutions, etc. If you want the construction companies to pay for it, they'll just not build it.

Given that I've worked in consulting for years I've seen a good amount of soulless corporate types, but they're usually not that far from the average human being in most respects.

Your understanding of the world isn't workable when you think tens of thousands of people are acting as unhinged mass murderers and society just accepts it.

> if every single person took every single action to reduce their carbon footprint they possibly could, the global economy would simply collapse

This is very hard to believe, and makes the reader believe you think economic systems and societies are a bunch of discrete and discontinuous processes. Their efforts, just by taking inertia into account and limited ability for people to change their behavior immediately would send signals progressively to businesses and governments, allowing a good chunk of the system to adapt over decades. A lot of businesses would definitely go bankrupt, but generally econ people are aware of that and think it's a natural process (Schumpeter isn't new).

What you I mostly see in econ circles is just standard pigouvian discourse I believe, nothing too scary: if externalities aren't priced, people have no "real time" incentive to change behavior.

I'm sorry but how do you deal with the fact governments are complex entities which require important software and IT infrastructure to function ?

From my experience building everything in-house is always going to be 10 times more expensive than any market rate software/service and it's also going to be worse, because IT/software companies are concentrating that knowledge and know-how and the government entities end up doing half-assed engineering with a few engineers where they'd need dozens.