he’s been salty about this for years now and frustrated at companies throwing training and compute scaling at LLMs hoping for another emergent breakthrough like GPT-3. i believe he’s the one that really tried to push the Llama models toward multimodality
honestly, where NixOS shines for me is in my homelab. i don’t always have time to fully document what i’m doing, but my NixOS config is code-as-documentation for when work burns all of my memories away and has a git log and conflict management so i can manage multiple systems that share common config.
and once you find out you can have services run on systemd with syntax like services.jellyfin.enabled = true you’ll never want to go back to containers, although it has ways to manage those as well.
it’s overall a great OS for tinkering and deploying small services across small networks. not sure how it scales, but for my use case it’s damn near perfect
i guess the point that’s being missed is that when i say “hard” i mean practically impossible
my point is that it’s hard to program someone’s subjective, if written in whatever form of legalese, point of view into a detection system, especially when those same detection systems can be used to great effect to train systems to bypass them. any such detection system would likely be an “AI” in the same way the ones they ban are and would be similarly prone to mistakes and to reflecting the values of the company (read: Jack Dorsey) rather than enforcing any objective ethical boundary.
but what are the criteria? just because you think you have a handle on it doesn’t mean everyone else does or even shares your conclusion. and there’s no metric here i can measure, to for example block it from my platform.
what about the neural networks that power the DSP modules in all modern cell phones cameras? does a neural network filter that generates a 3D mesh or rather imposes a 3D projection, eg putting dog ears on yourself or Memojis, count? what if i record a real video and have Gemini/Veo/whatever edit the white balance? i don’t think it’s as cut and dry as most people think
it’s already the case that the distinction between what’s “AI” and what isn’t is a subjective, aesthetic difference and not a technical one
modifier for window manager nav and general OS controls like wofi/rofi
pretty sure it’s SteamOS, an Arch Linux derivative, on a fairly popular Snapdragon platform. probably not too difficult to hack on it.
there are some teams in companies like this where management doesn’t want to account for upstreaming and some engineers are happy to open a bug report, move the ticket to blocked, and move on to something else
i mean… sure. some neat tricks in here i wasn’t aware of, but asking my mom to open the terminal… i mean it’s not rocket science but that doesn’t make it accessible. all the scripting and stuff that you’re talking; that stuff comes in the Jellyfin box. honestly, it might be worth it to have both if you have users that aren’t comfortable in the terminal
man this brings back memories.
i was able to install Arch on my 2012 Macbook Pro, but the networking was a huge issue. not only did the driver cause terrible screen tearing for some inexplicable reason, but i had the same problem even getting the dang thing installed. luckily i’m an Android developer and was able to share wifi over USB with an Android device.
i’d say so. i was a professional Android dev for years, and security and privacy are definitely one of the reasons i prefer iOS. i don’t have time to play with my phone so much for my personal device. Apple is the lesser of 2 evils since their business model doesn’t depend on this kind of tracking (even if they do it as well albeit to a lesser extent)
ok i’m not saying do this
i recently setup an API proxy, C&C server, Grafana and Prometheus, and Discord bot. now i can send pings via Grafana or with a simple request (provided it’s authed via VPN or proxy) and have my Discord bot use a local LLM on my network to deliver the alert to a Discord channel in the voice of Ultron.
but LLMs do represent a significant technological leap forward. i also share the skepticism that we haven’t “cracked AGI” and that a lot of these products are dumb. i think another comment made a better analogy to the dotcom bubble.
ETA: i’ve been working in ML engineering since 2019, so i can sometimes forget most people didn’t even hear about this hype train until ChatGPT, but i assure you inference hardware and dumb products were picking up steam even then (Tesla FSD being a classic example).
i know it’s popular to be very dismissive, but a lot of “AI” has already been integrated into normal workflows. AI autocomplete in development text editors, software keyboards, and question asking bots isn’t going away. speech-to-text, “smart eraser”, subject classification, signal processing kernels like DLSS and frame generation, and so many more will be with us and improving for a long time. Transformers, machine learning optimized chips, and other ML fields are going to be with us for a long time. the comparison to NFTs is either angst or misunderstanding.
the thing that finally got me using a terminal multiplexer was zellij. you can think of it as tmux with training wheels, but i don’t see a reason to go back.
i don’t know the full nature of the exploit, but zlib has an exploitable integer overflow via the MiniZip project. even though our images don’t use that project.
i know people usually are like, “oh cool new features”
but this has a security patch that will literally unblock my pipelines at work lol 🎉
three, maybe four things:
flake.nixsome things are resistant to documentation and have a lot of stateful components (HomeAssitant is my biggest problem child from an infra perspective), but mainly being in that graph mindset of “how would i find a path here if i forgot where this was” helps a lot