Propelloni

As someone who operates in the planning/approval space I can say it is kind of a relief to just steamroll through some posh neighborhood.

TFA is about students learning a language (German). How can a student critically verify AI output of a supposedly German sentence if the student does not speak German?

I guess you are knee-jerking and confounding two notions: a) learning something, b ) using something, which are not the same. Using something that already knows something else is as old as pie, to wit, my tax advisor. If I want to be a tax advisor myself, I have to learn this shit and not outsource to another tax adsvisor. Typewriters help students learning a language by removing a distraction. Should the students to this themselves? Yes, but we all should spend less time on the internet, yet we don't. This tells us something ;)

Great work! I still think that [1] does a better job of helping us understand how GPT and LLM work, but yours is funnier.

Then, some criticism. I probably don't get it, but I think the HN headline does your project a disservice. Your project does not demystify anything (see below) and it diverges from your project's claim, too. Furthermore, I think you claim too much on your github. "This project exists to show that training your own language model is not magic." and then just posts a few command line statements to execute. Yeah, running a mail server is not magic, just apt-get install exim4. So, code. Looking at train_guppylm.ipynb and, oh, it's PyTorch again. I'm better off reading [2] if I'm looking into that (I know, it is a published book, but I maintain my point).

So, in short, it does not help the initiated or the uninitiated. For the initiated it needs more detail for it to be useful, the uninitiated more context for it to be understood. Still a fun project, even if oversold.

[1] https://spreadsheets-are-all-you-need.ai/ [2] https://github.com/rasbt/LLMs-from-scratch

LOL, they just introduced QoS into the electricity grid.

I have several points to say to that.

1) 2015 saw the iPhone 6s, which was only 15 mm shorter than the Xperia 5 or 10 V, while being about the same width and thickness. It had a tiny screen in comparison. The 6s Plus was larger, and heavier, than the Xperia 10 V, in all dimensions (OK, not thickness, this was the time of "paperthin" phones) while still having a smaller screen.

2) I don't want a tiny 2008 smartphone, I want a phone I can use with one hand. A width of 70 mm or less lets me do that. Today, that is small, in 2015 it was about normal.

3) My perfect phone was the Samsung Galaxy S6 Edge from 2015, which has about the same dimensions like the Xperia 10 V but the rounded screen edges made it easier to use with one hand.

Modern self-help is as much a sham as are management gurus, which is no surprise because they overlap. Who cannot recall "Start with why?" or the "7 habits of highly effective people"? They play on your insecurities and promise silver bullets. If they don't work it is because you are, of course, deficient. You need another self-help book or follow this guy on Insta. Indulging all this self-help stuff is just another form of procrastination, instead of doing it you talk and read about doing it. It's just like learning Org-mode (no offense) to be better organized instead of, you know, organizing.

My waking call was, ironically, another management book "The Management Myth" by Matthew Steward (I think), which just showed me the ridiculousness of it all.

I run a Xperia 10 V. Great phone, great form factor, easy to unlock. It runs for days, almost a week, on one battery charge. Sony is doing something right here.

Then how did we get the laws we have now? How did we get the constitution and the amendments?

Art needs no justification. So, well done ;)

I applaud the technical journey, but I really do not understand the purpose. What's the point of knowing last week's lottery numbers?

Write a short blurb what's it about? You can do that with any submission.

Me too. So as a service to the community: the article is about a noticeable increase of submissions about high-energy theory to arXiv due to mediocre articles quickly produced with or by AI and how to deal with that.

I'm not worried, freedom.gov will save us all! /s

The e-mail posted somewhere in the comments, assuming it is legit, makes it clear that FLOSS Fund requires certain paperwork for tax reasons to the benefit of the receiver. Apparently the Pocketbase developer is receiving the money personally, which means it is income and will be taxed. Apparently, again, it would also be taxed in India (the seat of FLOSS Fund) and the paperwork would allow to avoid double taxation.

This appears much more reasonable to me than the hoops I have to jump through to declare my taxes as an US expatriate and avoid double taxation with my country of residence.

Maybe you could build a heuristic around shipping weight? A single golf ball weighs about 45 to 50 g, so divide the shipping weight by, say, 50 g to account for boxing and so on and you get a rough estimate of the balls in the package.

Preach it. I, for one, welcome my caveman dentist!

True, but we have the same issue with US-based software, or any closed source software really. At least here I can take the source code and check for myself, or let an AI, before building.

I have to use MS products at work and use Collabora (on Linux) at home. Sometimes, I have to edit the same spreadsheet at home and at work and that usually works flawlessly, besides automatic coloring. Now, I also don't try to code in Excel, so that might be one reason. Excel graphs are also nicer.

Here is a comparison by the Document Foundation for spreadsheets [1]. I think it speaks for itself.

Regarding Powerpoint I can't say. I can't recall when I last used Powerpoint for anything. We have an in-house system where I just select slide type, enter my text and attach pictures in a form and it builds a CI-styled PDF for me. I think its basically a LaTeX front-end, but I never cared.

HTH

[1] https://wiki.documentfoundation.org/Feature_Comparison:_Libr...

So, if I rent out my bike to you for an hour a day for really cheap money and I do so a 50 more times to 50 others, so that my bike is oversubscribed and you and others don't get your hours, that's OK because it is just capacity planning on my side and widely accepted? Good to know.

How about a flight coming out of Bolivia with 200 Dutch tourists on board? Is it more or less risky than a flight coming out of the USA with 200 Donald Trumps on board? Is there a list?

> If the world is governed by rules, why does the United States maintain a considerable number of military bases around the world, far exceeding the total number of military bases of all other countries combined?

It's the other way around. Rules are tools of peace. No peace, no rules. But if you want peace then you have to be ready to wage war. It's called deterrence and the EU is learning this just now, again. That's also one reason why the USA has been called the world police... because it was true.*

If nobody enforces the rules any more, things break down and we close in on violence. It is plain to see on the global scale, e.g. Russia's war against Ukraine, and also the domestic scale, e.g. ICE's violence against their own citizens in the USA.

> Why is the American military budget so much higher than the combined military budgets of all other countries?

The US military budget is about three times that of the EU or China's, or about a third of all military spending on the globe. Obviously, this is much higher than any single entity, but not all other countries combined.

* Frankly, being the world police has had a lot of benefits for the USA. Why they are abdicating this position to run a protection racket instead is for wiser people than me to answer.

> if there is a shortage of doctors, why are we trying to solve that by training AI models modelled on influencers that spits out (hopefully improving) advice at 10x the rate of a human doctor?

The crucial part is the training. AI may very well be the solution for underserved communities, but not if it is trained on internet rubbish. Train an AI on curated state-of-the-art, scientific data, imagine the expert systems of yore on overdrive, and you will see much better results, including knowing when to call in a human doctor.

Isn't EUMETSAT data usually under CC-by-SA 3.0? So all you have to do is to register with them and get your client ID for API access, or are there more hoops to jump through?

This only holds for companies that do not have to comply with some regulation or standard, e.g. ISO 27001, to do business. Especially infrastructure, banking, and defense tech have high compliance requirements that also, and sometimes esp. cover software development.

Maybe I'm not getting you right, but IMO it hasn't? I, as a customer/patient, just don't weekly converse with my MD about small issues, and frankly, they have better things to do, for example treating sick people.

Instead I use the health benefits programs of my health care insurer. My insurer has an interest in prevention, so I can get consulting for free (or very low fees), and even kickbacks if I regularly participate in fitness courses and maintain my yearly check-up routine. Now, I live in Germany and it probably is different in other countries, but it just makes economic sense from the insurer's point of view so that I would be surprised if it were very different elsewhere.

> Note that I'm not saying everyone should give the US a pass or maintain as much economic and defense dependency on the US. But I think it's hyperbolic to make all your long-term plans assuming something as stupid and self-defeating as his worst anti-ally policies are a new normal, because they harm the US at least as much as they harm everyone else, and everyone but those two knows this.

It is debatable if everyone but John Bolton and Donald Trump knows this. After all, according to the last NYT poll the current POTUS commands an approval rating of 41 % in the USA. The number of people I meet who do not understand how tariffs work, for example, is staggering.

Anyway, it is smart policy to expect the worst and plan for it instead of being surprised by another insane president voted in by the people of the USA. Call it risk management if you like. It would be negligent of the leaders of the EU and its member nations to not account for that. The EU has to reduce dependence on unrealiable trade partners, this is true whether we are talking about warmongering Russia, dictatorial China (probably the most reliable of the three!), or unpredictable USA.

So, let's hope for the best and prepare for the worst. The EU can't change it if preparation harms US economic interests in the long run. That's on Trump.

It is like in organisational error management (aka. error culture), there are three levels here:

1) errors happen, basically accidents.

2) errors are made, wrong or unexpected result for different intention.

3) errors are caused, the error case is the intended outcome. This is where "bad people" dwell.

Knowing and keeping silent about 1) and 2) makes any error 3). I think, we are on 2) in TFA. This needs to be addressed, most obviously through system change, esp. if actors seem to act rationally in the system (as the authors do) with broken outcomes.

> I hope you are using KDE Plasma instead of the default GNOME which is going the Microsoft way.

That is a disingenuous statement.

Gnome is just as open source as KDE is and there are several forks for those who don't like the direction on Gnome. At no point does Gnome force ads on you, changes default apps under your butt, or takes a nap before opening a menu.

Sure, Gnome is not for everybody and you may dislike the direction it is taking, but saying it is like MS Windows, or the community project is like Microsoft is dishonest and insulting. I expect better behaviour from a fellow FOSS enthusiast.

And yet Gnome and KDE have no troubles doing just that if you activate the corresponding plug-ins for the respective search engines (Tracker and Baloo). How you access them is a matter of taste, but both integrate into your runner dialog or launch menu, if you wish.

Well, obviously, your car can count cycles on the electromotor moving the wipers. Then you apply statistical wear and tear, maybe even geofenced, and your car orders new wipers. Same with tires. Simple as pie ;)

BTW, I would have zero interest in that feature.