Yeah, those were the problematic cases they highlighted. Image/video generation usually involves generating many permutations, perhaps adjusting prompts, and just keeping/editing the parts they want. The Trump video is likely made up of many edited clips from many different prompts, after discarding many more, not just one prompt. The song may be one prompt; haven’t played around with music generation myself.

OpenAI released much more impressive demo videos last year, and I think Sora is available to the public now. I don’t think most proprietary models/systems allow you to use public figures though, so it’s probably an open-source system.

I actually think it’s more targeted toward the Andrew Tate and Fresh and Fit follower types, and the propaganda does appear to be working on younger men, who are rapidly becoming more “conservative.” They’re extremely insecure in their masculinity, and think the subservience of women would be affirming.

Of course, it actually just hurts everyone, barring the people that benefit from keeping the working class divided.

If you’re running a lot of stuff on the same server, I agree with others that you’d want to use containers or VMs to avoid possible dependency hell. I prefer containers so I don’t have multiple OSs using RAM. I’ve never used Proxmox, but if I understand correctly, it’s an OS specifically built for running containers and VMs more easily, so I’m guessing that’d be a good choice. I personally just use Ubuntu LTS or Debian, Docker, and SSH to administer my servers, because that’s what I’m familiar with.

A cheap used Desktop PC off Craigslist or whatever should be fine. Desktops are more upgradable and configurable. You’d want to make sure the CPU and Mobo support however much RAM you’d want. Ext4 is fine if using a single disk; ZFS for multiple disks with redundancy. Preferably, a smallish SSD for the OS disk, but not required.

*arr stack for pirating: https://wiki.servarr.com/

Jellyfin for serving media. You may want something like the cheapest Intel Arc GPU for transcoding if you’re going to serve HDR video to low-spec devices.

Nextcloud for basic file sharing. NFS for high performance file sharing with Linux machines, if needed. Syncthing for syncing files if you need that.

Immich for something similar to Google Photos, if needed.

Also if they can’t afford that, surely they have state medical insurance

Ha! No, depending on the state, there are tons of barriers, means-testing, work requirements, mandated classes they might not be able to attend (due to childcare/lack of transportation), etc.

Even after the ACA, ER visits are still all the healthcare many people get.

I’d argue that even if gen-AI art is indistinguishable from human art, human art is better. E.g. when examining a painting you might be wondering what the artist was thinking of, what was going on in their life at the time, what they were trying to convey, what techniques they used and why. For AI art, the answer is simply it’s statistically similar to art the model has been trained on.

But, yeah, stuff like game textures usually aren’t that deep (and I don’t think they’re typically crafted by hand by artists passionate about the texture).