“This is getting ridiculous and I’m about to just toss the whole thing and move back to Google”

These people never learn do they?

I am not saying i understand it at all. My comment is genuine, not rhetorical.

On the contrary i express that i do not understand this to the point of deeming it strange or weird, triggering my curiosity.

I was hoping someone would chime in an explain how text generator techniques (which i do know alot about) known to have no capacity to use real world reason or logic can be used to help solve dna puzzles unless as a sub module of general machine learning. In which that later case would be within my understanding

Machine learning sure, but an llm? For anything other then as a module to translate abstract data into words that just seem weird

We are on the same base,

I actually had a friend who jokingly mocked me for liking ai because i was initially very exited ablut Dall-E and ChatGPT 3.5

Back then i could only see the potential that it continues to have. OpenAI appeared to have altruistic goals and was a non profit. Trojan horse it turned out to be.

Had to make pretty clear to my friend that “ yes, but not like this, everything but this” about the current slop situation.

I should nuance,

For a person who already actively uses comfyui, knows how the different nodes work,

Makes complex flows with them,

Making their own checkpoints is not a big step up.

I have not gotten to this level myself yet, i am still learning how to properly using different and custom nodes, and yes

In the mean time yes, i experiment with public models that use stolen artwork. But i am not posting any of the results, its pure personal use practice.

I have already seen some stuff about making your own models/checkpoints, if i ever get happy enough with my skills to post it as art then having my own feels like a must. The main reason i haven’t is cause it does take a lot of time to prepare the training data.

People that don’t use their models while calling themselves artist are cheating themselves most of all.

If you can boot an os from usb (basically the same for all distros) you can try proxmox.

There are these incredibly useful helper scripts that setup entire services in 1-2 copy pasted commands.

https://community-scripts.github.io/ProxmoxVE/

To explain what proxmox is its basicly virtualisation software, it can run vms but also lxc (light linux containers) and share resources very efficiently between all of them

Jellyfin, radarr, sonar. They are all included in the helper scripts, each will be a dedicated lxc.

Its also very easy to setup raid and there own storage format is very efficient.

Its well documented to the point that any decent llm can help you learn whatever you need. In fact its claude that helped me setup my own proper raid on proxmox, also tought me about datasets and how i can make those available to different lxc

Personally i am very hands off with my server, the hardest part is often choosing what ip i want to give a service, i rarely update or mess with it if not strictly necessary.

For hardware i recommend plenty of ram (can Be bought and installed seperatly), more cores is usually better and internal graphics can save you some hassle depending on what you are doing (also allows you to dedicate a Big gpu to some services).

A warning on second hand corporate machines, the performance is often good But quite fans are often an afterthought. I onxe got a beast of machine for free but you could hear it spin from anywhere in my house.

A good practical case is always a blessing when you need to check the insides.

If you download a checkpoint from non trustworthy sources definitely and that is the majority of people, but also the majority that does not use the technical tools that deep nor cares about actual art (mostly porn if the largest distributor of models civitai is a reference).

The technical tool that allow actual creativity is called comfyui, and this is open source. I have yet to see anything that is even comparable. Other creative tools (like the krita plugin) use it as a backend.

I am willing to believe that someone with a soul for art and complex flows would also make their own models, which naturally allows much more creativity and is not that hard to do.

Never heard of this one but i might try it. Looks very clean and practical.

https://github.com/glanceapp/glance

That was a beautiful read.

But do i find myself conflicted about dismissing it as a potential technical skill all together.

I have seen comfy-ui workflows that are build in a very complex way, some have the canvas devided in different zones, each having its own prompts. Some have no prompts and extract concepts like composition or color values from other files.

I compare these with collage-art which also exists from pre existing material to create something new.

Such tools take practice, there are choices to be made, there is a creative process but its mostly technological knowledge so if its about such it would be right to call it a technical skill.

The sad reality however, is how easy it is to remove parts of that complexity “because its to hard” and barebones it to simple prompt to output. At which point all technical skill fades and it becomes no different from the online generators you find.

The pattern are “we have this. we did this, now we believe this that is wrong” and “says what it is not, says what it is instead”

Sometimes a combination of the two

On premise there is nothing wrong with such sentences but llms tend to heavily overuse it and it becomes very formulatic.

Ask an llm to explain any concept and you are bound to find examples of it. Tell it how it made a logical error and your almost guaranteed to see an example of it.

Over the entire text i note about 10 variations of that pattern.

Thats fair,

In irony i probably could have worded my criticism better myself.

Its not because its ai that i don’t like it but rather because it has all the sloppy patterns i started to recognize that are prevalent in ai.

Some of those become increasingly jarring but only because i pay a subjective amount of attention to them. Bad human writers have an advantage in that their bad writing structure is still more unique to only their own writings.

A piece of writing being thoughtfully put together is far from inconsequential for me.

I use a premium tier ai myself and am not against using it for assistance, it can craft a decent snippet (that still needs multiple manual edits) But not at all a full coherent text that reads efficiently.

Its applies structure without understanding the goal of the text resulting in a paragraph salad.

It simultaneously treats the reader like a toddler with oversimplified metaphors while also overcomplicating things for no other apparent reason than filling a word quota.

Above article is twice the length it needs to be. Its lazy, lacks actual understanding and feels “sloppy” in the original meaning of the word.

Having read more of the text i feel my original comment was way too forgiving. Even the opener does not make sense if you try and digest it. Silicon. It even includes misinformation, stack management existed before C was a thing.

Don’t get me wrong, i like using dashes too.

I actually have a contract to sign which requires an em dash that changes the interpretation drastically. I am struggling to get the author to realise its importance because i received 3 updated versions that did not include it. And this is after i replied with a self-fixed document the first time.

But this article really does use them a lot, and thats not the only tell, just a more obvious one.

I dont know how accurate zerogpt is but i gave it the full text and they returned 100% ai writen, not even a mix.

Sorry but even if this was written by a human, ai has ruined this kind of sentences for me:

“This is Hardware Stockholm Syndrome: we optimized the hardware for C, then pointed at the hardware and said “see, C is efficient!” We forgot we made it that way.”

Also, so-many-dashes. Even if the “human author” fact checked all the details, it reads like slop and i cant get trough it.

For a main desktop, absolutely.

For low maintenance servers and vms? Staying with Debian.

For a laptop on which Arch gives issues (mine where with battery when standby). POP-OS has been working well.

The irony is that self hosting in many ways has become way easier, faster and cheaper then using proprietary alternatives.

The catch is what you are already used to do.

A good example to illustrate this is is how easy it is to install software tools on linux using a package manager.

You get used to this so quickly that when i had to install something on windows i really struggled:

i have to find the right site to download from, run that installer, navigate menus, make sure it doesn’t come with bonus malware. And then when i finally run it turns out its a free demo and half the features are locked.

Comparing different plans for o365 and setting up a family with different accounts, again its a pain.

Spinning up a one click install nextcloud instance and getting that configured was peanuts in comparison.

People who wear these cameras to feed into ai should make it a habbit to regularly give the finger to the task between steps. Also take dance breaks when no is watching.

deleted by creator

The technical term seems to be a JBOD bay. (Just a Bunch Of Disks)

Basic ones are probably usb, ideally you have something that has a SFF port. Modern ones might also have thunderbolt.

Finding a micropc that supports SFF out of the box might be a challenge but some do support pci express cards.

Apparently there also exists something like Oculink which is pci over cable but i know even less about that one.

EDIT: if you look for “Nas enclosure 4bay” you actually do find plenty of options (Jonsbro N3 per example) that allow you to build it all in one unit with a mini-itx board. A nas pretty much just is a pc with special software so this would be what i recommend.

Maybe i miss some perspective here because i never had the spare money to consider a storebought nass. The convenience never sounded like it was worth being locked down to its software.

My server is “just a pc”

I got a case with external drive slots (it also needed to fit a gpu), but i suppose external drive cases also exist that can connect to a micro computer build.

The software is proxmox, which imo is amazing. Its virtualisation and backup software and performs really well and has a proper gui.

I have numerous lxc (linux container that is not a full vm) that each run their own docker with a single service. I can ssh into those from my main system or visit the terminal and other panels in the proxmox gui. Many services host a gui to my network and i could probably make it so cli is minimal but i personally am comfortable with that so…

I also run a few full vms on it, including some windows desktops.

You could probably also host actual Nass software this way.

All of these work well next to eachother and share resources. Snapshots and backups of individual systems or data can be made with ease.

If it doesn’t fit your usecases you can get the off the shelf ones i guess but for others interested here, maybe this helps.