marcodiego

> [..] models are [...] limited in [...] ability to answer knowledge-intensive questions [...], they did not memorize the relevant facts. [...] This is probably because the required knowledge cannot easily be accessed via papers [...] but rather by lookup in specialized databases [...], which the humans [...] used to answer such questions [...]. This indicates that there is [...] room for improving [...] by training [...] on more specialized data sources or integrating them with specialized databases.

> [...] our analysis shows [...] performance of models is correlated with [...] size [...]. This [...] also indicates that chemical LLMs could, [...], be further improved by scaling them up.

Does that means the world of chemists will be eaten by LLMs? Will LLMs just improve chemists output or productivity? I'd be scared if this happened in my area of work.

Any more information than that? Abuse?

Sure! How about a description?

> we have immense wealth, medicine, etc. that doesn’t necessarily imply we are psychologically healthier or happier.

That may make sense if you compare us to a few generations ago because maybe we're almost in a "diminishing returns" phase and our lives aren't significantly better than our closest ancestors. Nevertheless, I'm not sure if we were happier during periods of famine, plague and inquisition. It is still possible to have a "go-to oversimplified answer to this" even for this "go-to oversimplified answer to this".

> I’d argue the world/humankind is already an immense disaster

Not exactly disagreeing because I really don't know any better. But whenever I hear comments like these, I wonder if isolated peoples or medieval/industrial revolution population have/had a better life then we do. We have conveniences that far surpasses anything a king could have just a few decades ago. Our life expectancy and IDH is (almost) monotonically increasing.

I really expect that efficiency growth will make the world converge to a life style where most people don't have to work to survive, but just to acquire luxury goods. I actually think that maybe we could already be at that point if it wasn't for the "sadists that run the world", but I don't think they can prevent it from happening because that will make them make more money too.

So, I really don't know if people criticizing our current state as "an immense disaster" are overly pessimistic or if the people (like myself) thinking that the world is in a never before achieved good state are realistic or uninformed.

I fear deterioration beyond repair, like irrecoverable climate change, because an event like that would really separate the very rich from rest of the population in a way that only one of these two partitions could survive. But I don't think we'll reach that point my lifetime.

Just port it to Wayland. /s

FLOSS devs do eat daily. It's not developed by fasting people.

You know, a "made with" blender film just won an Oscar. I bet Blender devs also have to make a living and I doubt they work for free.

I tried KDE 1.0 two decades ago. Although it looked like a copy of windows ideas in some points, it already seemed better even at the time.

It might be a good idea for a SETI@home like project.

How expensive is a "brute force" approach to decode it? I mean, how about mapping each unknown word by a known word in a known language and improve this mapping until a 'high score' is reached?

My first x86 was acquired in 1997: a 200MHz pentium MMX with 32 MB of RAM, a 2.0 GB harddisk, a 2 MB S3 ViRGE as video card, Aztech Labs AZT-R 2316 sound card, a 33.600 bps winmodem and 28X (I think) CD-ROM. I soon wanted to install Linux on it. It ran minilinux, a small distro that could be launched directly from DOS without partitioning, relatively well. I also downloaded Debian and prepared the 10 floppies to install it but never got ahead because it didn't support my keyboard layout (ABNT2) out of the box. I eventually installed Conectiva Linux 3.0 (marumbi), a RedHat based distro, on it. It worked somewhat beautifully but had no support for the modem and soundcard.

As my relatives and siblings updated their computers, I inherited their spare parts. My machine got some very interesting upgrades. It got 2 floppy drives (very rare at the time), 2 harddisks, a cd-recorder unit, RAM was upgraded to 64 MB and the modem was replaced by a ne2k compatible network card. I also had a Linux-supported Canon BJC 4200 and a SANE-supported TCE table scanner. Still, I couldn't get the SoundBlaster that my brother had and, sadly, my machine continued without sound support on Linux.

At around 2006 I replaced it with a new self built computer which had better compatibility with Linux than with windows-xp and then this new computer became my first dedicated Linux machine. I found out that Linux eventually got support for AZT-R 2316 at around 2007, but the last time I tried my old computer, it displayed "parity error" probably from oxidation in the memory connectors. I then just gave up on it.

Later on I thought about if it would be possible to install a graphics card with OpenGL support and USB port on one of the remaining PCI ports. I certainly wouldn't be able to install a modern Linux distro on it, but certainly it would work with one of those specially crafted for old computers like TinyCore. With recent kernel changes, swapping maybe smart enough to be usable on SSD with SATA to IDE adapters even in that computer. Now, that would be a dream machine.

Not the first time something like this is shown here. And it is very important to say that such a thing is a anti-feature of the driver and that something like this would probably never be acceptable on Linux, where such devices work beautifully just after plugging.

The operating system where these things happen should also be blamed.

The vendors of the camera have the same interests of the vendor of the software. It is just a matter of time until the software watermarks the video and your camera automatically stops recording.

Users have to resort to (exclusively, if possible) open source tools.

This and DRM and other restrictive anti-features like self destroying messages, un-recordable strings, unprintable files are all fully artificial restrictions. They make no sense when the source code is available since removing it is as simple as removing an if.

I payed for my device, it is mine, it is up to me to decide whatever I'll do with it. It is my right under the private ownership definition. The current situation on modern devices, especially smartphones, is ridiculous and a complete distortion of rights that are fundamental even for the roots of capitalism.

Users should organize and, at the least, avoid using such services even if it means to lose some convenience. Losing my freedom is not a fair price to pay for such conveniences.

A friend of mine who was into alchemy, told me it was because the difference was only three protons. I don't if early alchemists knew that or why not consider metals that are less than three protons different from gold.

Clipper continued popular in Brazil until early 2000's. It was sometimes called 'software de locadora' (video-rental software) or 'software de farmácia' (drugstore software) since they were so often used in these two cases. During the late 90's there was a wave of replacement of such software with applications made in Delphi.

For some time, this world was fully proprietary in Brazil and you had to pay big bucks for tools or resort to unlicensed copies (piracy). Since the second option was unavailable for the public sector and bigger companies, this created an horrible dependency on proprietary inferior software tools that took around a decade to heal. It is a good thing Linux was so successful in some many points of the stack.

An anecdata: a few years ago I bought something in a drugstore that was running a Linux desktop with DosBox running something in Clipper. I also saw Linux desktops accessing something running in Clipper, in apparently a DOS (virtual?) machine, remotely through telnet. Funny but somewhat ugly reality that is fortunately becoming extinct.

Things only got marginally different (better) when during the first decade of the 2000's systems started to go on-line, mostly with LAMP winning against ASP and then, more recently, with the advent of Android apps coupled with cloud services.

What I really would like: something that mimicked the old SGI start up, complete with boot audio and a micro distro for OS setup. These days, with snapshot filesystems, that shouldn't be too hard. Also, I've had to chroot to fix my system in my life a few times; I can't believe that's hard to automate.

If it falls in my backyard. Can I keep it to myself?

I can reason about an "unknown chinese rootkit" as much as about an "unknown US rootkit".

That's is not a new idea. I think Office 97 had an accelerated startup that made windows take a little longer the boot but faster the start office.

Hmmm.. it maybe free from IME! Maybe the FSF want a word with them.

In Brazil the entity "the economy" is called "os mercados" meaning "the markets". It is handled by the press as a personality who has humor, wishes, preferences and power. It is actually somewhat funny or curious when you read things like "the markets didn't liked what was said and shares fell".

Of course, it is clear that politics has impact on investors but, other than being able to vote, there should be no other way to influence politics in a democracy.

Easier to watch if you know french.

I put "1 / (-1~1)" and expected something around - to + infinty. It instead gave me -35~35.

I really don't known how good it is.

> For the second attempt, GIMP was used to get a better window selection algorithm with ffmpeg, and with a slight bit more data (more frames extracted), a perfectly legible result

Take that adobe.

In Brazil, people who support bolsonaro are more likely to believe in conspiracy theories and not trust doctors, vaccine or science.

I also think it is reasonable for everyone to have a tendency to not trust a doctor who is a bolsonaro supporter.

There is a very emblematic case, although not very known, of a doctor who refused to treat the child a a Lula supporter: https://www.jusbrasil.com.br/noticias/medica-se-recusa-a-ate...

I'm a Brazilian. Everything here can be paid through pix. It's very convenient, fast reliable and, for a country like ours where walking on the streets with money is a risk, safe.

There's only one reason I don't use it: there's no FLOSS app (AFAIK) to use it.

Something as common as the dominant payment system should not depend on proprietary software.

> Try doing the same with a Linux binary that's just a year old.

AppImages are very close to fixing this. I'm not sure if it is already solved or very close to.