NorwegianDude

To be fair, a single server is way more reliable than cloud clusters.

Just look at the most recent many hour long Azure downtime where Microsoft could not even get microsoft.com back. With that much downtime you could physically move drives between servers multiple times each year, and still have less downtime. Servers are very reliable, cloud software is not.

I'm not saying people should use a single server if they can avoid it, but using a single cloud provider is just as bad. "We moved to the cloud, with managed services and redundancy, nothing has gone wrong...today"

Nobody cares if I can do it a million times faster, everyone can. It's cheating.

The whole reason you have to account for the time you spend setting it up is so that all work spent processing the data is timed. Otherwise we can just precomputed the answer and print it on demand, that is very fast and easy.

Just getting it into memory is a large bottleneck in the actual challenge.

If I first put it into a DB with statistics that tracks the needed min/max/mean then it's basically instant to retrieve, but also slower to set up because that work needs to be done somewhere. That's why the challenge is time from file to result.

This is very silly. You're not doing the challenge if you do the work up front. The idea is that you start with a file and the goal is to get the result as fast as possible.

How long did it take to distribute and import the data to all workers, what is the total time from file to result?

I can do this a million times faster on one machine, it just depends on what work I do up front.

Yeah yeah.

Nobody believes what Americans say, be that Trump, Elon or Apple. They're all full of shit, and they rarely do what they say. The average junkie is a more reliable source on what Apple will do than Apple itself.

The Gemma 3 models are great! One of the few models that can write Norwegian decently, and the instruction following is in my opinion good for most cases. I do however have some issues that might be related to censorship that I hope will be fixed if there is ever a Gemma 4. Maybe you have some insight into why this is happening?

I run a game when players can post messages, it's a game where players can kill each other, and people often send threats along the lines of "I will kill you". Telling Gemma that it should classify a message as game related or a real life threat, and that it is for a message in a game where players can kill each other and threats are a part of the game, and that it should mark it as game related if it is unclear if the message is a game related threat or a real life threat does not work well. For other similar tasks it seems to follow instructions well, but for serious topics it seems to be very biased, and often err on the side of caution, despite being told not to. Sometimes it even spits out some help lines to contact.

I guess this is because it was trained to be safe, and that affects it's ability to follow instructions for this? Or am I completely off here?

No, cause thats just one of the features.

Images are often at different resolutions too, that way, depending on the pixel density of the device, and the physical size, the browser can select the photo that has high enough resolution, but not one that is needlessly large, while also selecting the preferred image format.

I had planned a trip that included the US for this summer, but the fact that they can demand the password for my devices is the main reason im not going. Having to wipe devices before travel and having to download data again because people dont respect privacy and others suck.

The fact that you have to get approved before traveling(that is fine), and then can be denied entry when you arrive for no logical reason is absurd. Visiting the US is simply not worth the risk and hassle.

Its crazy when you expect your privacy to be more respected in China.

No technical dependencies on non-EU infrastructure seems very unlikely. Does the EU edition not rely on the same software that American AWS owns? Isn't it owned by AWS?

I dont like the wealth tax, but your numbers are not accurate. There is a discount and deductions on the wealth tax, but after the last election the discount has been reduced drastically.

It's not normally a very large issue, but I really don't like it. Most companies on the exchange makes money, and those who are not on the exchange are taxed on their assets. So in most cases it works out, but not always.

Usually owners use dividends to withdraw money to pay the tax, but that means even more tax as you have to pay tax on the dividends too.

The right side wants to remove the tax on investments, and maybe compensate by increasing the corporate tax. That way the tax on the annual result will be a bit higher, but there will be no wealth tax. This also levels the playing ground when it comes to Norwegian and foreign investors as the tax won't be based on where the owner is from.

Well, that very much depends on your definition of slow, doesn't it?

An additional request is another round trip. That can be very slow. Average TTFB on the internet in the US is ~0.7 seconds.

It's much faster to send it as part of the same request as you then don't have to wait for the browser to discover it, request it, wait for the response and then add it.

A build process does not have to be complicated, at all. If you can write HTML then using something that can simply read the HTML includes you wish existed and swap it with the specified filename is trivial.

Ofc, the idea has many other issues, like how to handle dependencies of the included HTML, how to handle conflicts, what oath to use and many more.

It does not have to be a programming language on the server, no, unless you want to. The server can have a static site, that you build as you deploy it.

Fetching another resource is expensive. It's another round trip, and depending on many factors it could be another second to load the page. And if the HTML includes other nested HTML then it can be much slower.

This is the exact thing we try to avoid when building websites that perform well. You want as few chained requests as possible, and you want the browser to be aware of them as soon as possible, with the correct priority. That way the browser can get the important stuff needed to display content fast.

Including HTML client side for templating is just wasteful, slow and dumb from a technical standpoint.

Every client would have to do another request for each include. It would literally be many thousands of times slower(or worse) than doing it locally where the templates can be in memory as you render the pre-render the pages. You also save a ton of CPU cycles and bandwidth, by not serving more files with additional overhead like headers.

Then you just pre-build the page before publishing it. It's way cheaper as you do the work once, instead of every client being much slower because they have to do additional requests.

24 GB of VRAM is a large amount of VRAM on a consumer GPU, that I totally agree with you on. But it's definitely not an extremely high end GPU these days. It is suitable, yes, but not high end. The high end alternative for a consumer GPU would be the RTX 5090, but that is only available for €3000 now, while used 3090s are around €650.

3090s are no longer produced, that's why new ones are so expensive. At least here, used 3090s are around €650, and a RTX 5070 is around €625.

It's definitely not extremely high end any more, the price is(at least here) the same as the new mid range consumer cards.

I guess the price can vary by location, but €1800 for a 3090 is crazy, that's more than the new price in 2020.

A 3090 is not a extremely high end GPU. Is a consumer GPU launched in 2020, and even in price and compute it's around a mid-range consumer GPU these days.

The high end consumer card from Nvidia is the RTX 5090, and the professional version of the card is the RTX PRO 6000.

If the military says something not in line with Trump they'll quickly be removed. Just look at what happend in Greenland.

https://www.bbc.com/news/articles/creq99l218do

You don't have to buy the hardware. It's very common to rent it.

When using colocation, nothing is stopping people from storing the database data externally from the server running the database like some cloud services do. But doing so, either in cloud or not, does have a serious downside: greatly increased latency.

What do you mean you miss it? What is stopping you from doing exactly that now? Plug it in, drag down the drop-down and make you phone identify as storage. Then you can copy video off it while moving music onto it, if that is what you want.

Yes. Yes it is. I guess this person has the same stance Vercel now has. Even Next.js docs can make up their mind of whether you should or should not do it. They reccomended it until yesteray, but then another major securityflaw was discovered that made it useless, and now they removed authentication from the docs.

The takeaway is that you should not do it. You should never use Next.js if you ever has somehting that is not supposed to be public for everyone.

No serious company uses Next.js after all the recent major security issues, at least not if they have and respect users data.

It's always interesting how it affects how people think about the day. If you look outside and it's dark then it's easy to think that the day is basically over, even though the time might say otherwise.

I really appreciate the long winter nights and long summer days in Norway. Being able to wake up to the sun as early as 04 and enjoy it until after 23 is great, and the UV is only high during the middle of the day. Long summer days are awesome. Long winter nights are probably not as appreciated, but I enjoy those too.

I've done a lot of site optimization on high traffic sites, and in many situations you should serve resources from the same server to reduce latency and the overhead of additional connections. This does of course depend on a few things like where servers are located in relation to users and what the capacity of the setup is, but you definitely should not always use a CDN. It was true many years ago, but not today.

So now the TOPS are not comparable because M3 is much slower than an Nvidia GPU? That's not how comparisons work.

My numbers are correct, the M3 Ultra has around 1 % of the TOPS performance of a RTX 5090.

Comparing against the GPU would look even worse for apple. Do you think Apple added the neural engine just for fun? This is exactly what the neural engine is there for.

I'm not confused at all. It's the real numbers. Feel free to provide anything that suggests that the TOPS of the GPU in M chips are faster than the dedicated hardware for it. But you can't, cause it's not true. If you think Apple added the neural engine just for fun then I don't know what to tell you.

You have a fundamental flaw in your understanding of how both chips work. Not using the tensor cores would be slower, and the same goes for apples neural engine. The numbers are both for the hardware both have implemented for maximum performance for this task.

True. But an AI shop doesn't care about that. They get more performance for the money by going for multiple Nvidia GPUs. I have 512 GB ram on my PC too with 8 memory channels, but it's not like it's usable for AI workloads. It's nice to have large amounts of RAM, but increasing the batch size during training isn't going to help when compute is the bottleneck.

No, I'm not. I'm comparing the TOPS of the M3 Ultra and the tensor cores of the RTX 5090.

If not, what is the TOPS of the GPU, and why isn't apple talking about it if there is more performance hidden somewhere? Apple states 18 TOPS for the M3 Max. And why do you think Apple added the neural engine, if not to accelerate compute?

The power draw is quite a bit higher, but it's still much more efficient as the performance is much higher.

Not much to figure out. It's 2x M4 Max, so you need 100 of these to match the TOPS of even a single consumer card like the RTX 5090.

The memory amount is fantastic, memory bandwidth is half decent(~800 GB/s), and the compute capabilities are terrible(36 TOPS).

For comparison, a single consumer card like the RTX 5090 is only 32 GB of memory, has 1792 GB/s memory and 3593 TOPS of compute.

The use cases will be limited. While you can't run a 600B model directly like Apple says(cause you need more memory for that), you can run a quantized version, but it will be very slow unless its a MoE architecture.

Considering how many people are using cases and even wallet cases, probably very many people. The case/wallet can be attached to the phone as a replacement to the normal backplate, making it much less bulky, just as Samsung did forever ago.

Personally I think that even if you don't know everything, just make clear what you are certain and uncertain about.

You probably know a lot, and what you write is still good as long as you don't spread misinformation. I have issues trusting most sources that are not clear about what they are uncertain about, cause most people are not experts.