• 0 Posts
  • 9 Comments
Joined 2 years ago
cake
Cake day: July 23rd, 2024

help-circle

  • richmondez@lemdro.idtoLinux@lemmy.ml•*Permanently Deleted*
    link
    fedilink
    English
    arrow-up
    8
    arrow-down
    1
    ·
    14 days ago

    As a sysadmin with years of experience I can tell you now chatgpt is almost never right, giving outdated or subtley wrong advice and commands that are fine as a shortcut if you already know the commands and can correct or ignore the mistakes, but exceptionally dangerous if you are just learning the ropes. The documentation is long for a reason, there are often lots of corner cases or exceptions to be aware of.


  • What is stupid is referring to AI as if it was just one tool when actually it’s a bunch of tools, some of which are easier to use sensibly amd some it’s near impossible to. Fancy auto complete like it’s being compared to sure, fully automated do what the fuck you want agents, yeah sooner or later that is going to go wrong and no one is going to know where or why.

    They might be fun and interesting, that doesn’t make it a great idea to go all in in my opinion, certainly not for anything you would consider important.




  • Their areas of expertise and research are not areas that offer insight into mechanisms of cognition, they are not a neurobiologist or a computer scientist involved in machine learning. Their high standing in related areas is being used to give credence to an argument that is no better than a gut feeling.

    You also have no ideas what conflicts of interest exist here, there may be none or the professor could be heavily invested in LLM companies.

    LLMs dont think and the “trains of thought” is just strapping a bunch of LLMs together to try and tidy up and correct if one of the models in the chain outputs nonsense which they are wont to do. They are useful tools, they don’t spontaneously generate output or request input. It’s a very complicated auto complete that is capable of generating plausible valid output (when considered by humans) to a given natural language input.




  • I personally manage my services using ansible, I only set up the actual infrastructure, the virtual machines that run the services, with terraform/opentofu. Docker is one of those in the middle tech between infrastructure and software distribution and it makes more sense to me to treat a service as a role in ansible do I can deploy it (docker, podman package install or whatever), sort it’s networking and handle it’s configuration all in one place. I’m not saying the way you do it is wrong, but this is just a step down the automation rabbit hole.

    It doesn’t appear your setup provisions the actual hosts for docker so I guess you are provisioning manually for that layer? That is another area you might want to leverage opentofu for?

    Also congrats on actually documenting it in a consumable way for others to learn from.