INeedMana @ INeedMana @lemmy.world

Posts

7

Comments

411

Joined

2 yr. ago

It doesn't have to be hamsters

Well, they also were after stability

Good. Let the data flow freely

That would actually be a good idea. With people growing up and learning on their own, you don't have a fully controlled system. The world inside the Matrix will be evolving and from time to time there will be some glitch somewhere. If you have a base of defined childhoods, then only physical differences between bodies and random occurences inside are uncontrolled

If I were to construct the Matrix, I would go after cloning (fewer biological lottery) and predefined lifepaths to keep everything nice and stable

The best advertisement of Belarusian windows

But machine will not do the creative part. It can only fill in the time-sinks around our creative ideas. Ask an LLM to tell you a joke no-one has ever heard before and then google it. The creative part still has to come from humans

EDIT; and the truth is that we very rarely come up with something creative. We mostly just recompile previously met combinations

trying to weasel out of putting some effort into something that sounds worth putting some effort into

But that depends what do they need it for

Personally I don't see a difference between legalese boilerplate and 10k word story. But that discussion might lead us nowhere

What about text creation have you learned

In many cases I don't want nor need to learn that. I just need volume about the key points

Why an LLM is any different?

Let's say I want my RPG players to find a corporate mail that gives them some plot info. Why not ask an LLM to write the boilerplate around the info I want to give them? Just as example

Let’s not put any effort into anything: the machine will do it for me

So you are not using a calculator, I presume? Only math done on abacus is not being lazy?

If you want something local and open source, I think your main problem will be the number of parameters (the b thing). ChatGPT-3 is (was?) noticeably big and open source models are usually smaller. There is, of course, an exchange about how much the size of the model matters and how the quality of the training data affects the results. But when I did a non-scientific comparison ~half a year ago, there was a noticeable difference between smaller models and bigger ones.

Having said all of that, check out https://huggingface.co/ it aims to be like GitHub for AIs. Most of the models are more or less open source, you will only need to figure out how to run one and if you have some bottlenecks on PI

1. no rolling-release: around once half a year you have to reinstall the system because it can't update some core library to a more recent version. And it's only the distro's limitation because rolling releases have no issue with it
2. you can't just define a package of your own. So if a piece of software is not in packages, you need to compile and install it manually without packager managing it. It tends to break in the long term and when the software suddenly becomes packaged
3. deb-hell: if you come to the idea to solve the first problem by compiling your own package, the packager will give you hell for that. And compiling your own deb with bumped up version is no easy task. Which means that when your version of the system goes out of life, you have to reinstall. Pray that you thought about this before and put /home and /etc on separate partitions
4. package dependencies are too baked in or stability is too high priority. Even if your issue got resolved recently, it will take a long time for an updated package to appear. And you can't roll your own in the meantime (see 2, or even worse 1)

My mom and grandma are using Manjaro. With grandma I'm the only one doing the updates of course, but with mom she usually can do it herself just using pamac-tray. If that fails a phonecall is usually sufficient. Once in a few years I have to come and do something by myself

And when that happens I work with a distro that just works, instead of some broken crap
EDIT: I tried having Mint on their computers. Big mistake, it's as broken as Debian and Ubuntu

EDIT: Xfce is very nice in such cases. It looks familiar for them while being manageable for me

In the end everything is maintained by the community, the only difference is that AUR is "everyone can maintain" and official is "we have team of official maintainers that decided to maintain these packages". Personally I can't imagine running without using AUR

But it's fair if it doesn't count for you

Yes, but that's only because a generation found some random, specific motion that scored better. Not because it analyzed that doing a skip should be possible

Yes, but that's kind of my point

We see it learn something with insane precision but most often it is almost an effect of over-training. It probably would require less time to learn another layout but it's not learning the general rules (can't go through walls, holes are bad, we want to get to X), it learns the specific layout. Each time a layout changes, it would have to re-learn it

It is impressive and enables automation in a lot of areas, but in the end it is still only machine learning, adapting weights to specific scenario

Pure speculation:
I think it's just not that popular. Does it do something more than rclone or simple rsync? If not, then its main selling point would be GUI. But then, I think, either one can use the remote location via their file manager (like thunar with MEGA for example) or there is not that much difference between opening another app and using web. And if the selling point would be pausing and resuming download, torrents are probably more verstile
It is available in AUR, though, so maybe it's only me that haven't heard about it earlier

Also, it's a java application. There is not much to package or depend on, I guess

It's cool but my question is (I did not see this addressed in the article nor video but might have missed it) did it learn to win the game in general terms or only this one example? I mean, if the layout of the board was changed, would it still solve it?

...

3 minutes ;P

"And now, Einstein will explain WiFi for you"

Oh gods, now I'm starting to count how much time will pass until I see an ad like that...