Upgrading badly with AI

This is an anecdotal take on my experience of updating from Linux Mint 20 Una to Linux Mint 21 Vera and my experience of using an AI chatbot tool, Perplexity, for my information searching and troubleshooting processes. I didn’t follow the officially recommended upgrade procedure or properly use the distro’s upgrade tool. I got a broken DE, but everything ended up being okay.

I offer my chat transcript in Perplexity as a resource.

I wasn’t directly following any official guides or community forum discussions. I wasn’t expecting my ultimate success. I had copied all my stuff onto a USB drive, an activity that is basically a personal ritual at this point, since I’ve been sloppily hacking my way through Linux for 12 years. In fact, I had been planning to wipe the system and start over with Arch Linux, but unfortunately the Arch Linux live installer didn’t find my wifi hardware. With several experiences of struggling tediously and often fruitlessly to fix Linux driver problems, I wagered it would be more worthwhile and more fun to try to hack my disappointingly stable Linux Mint installation than to follow the rules.

The major oversight was not ensuring that the desktop environment, Cinnamon, was reconfigured and upgraded from the new Vera repositories. I dealt with the broken DE, going through the process in a circular discussion with Perplexity that highlights to me the reasonable limitations of AI.

The AI failed to suggest that I simply needed to use the command startx to reload the DE after I had upgraded Cinnamon. It seemed to think the problem lay with my NVIDIA driver, and the rabbit hole of web sources cited by Perplexity lead me to nvidia-driver-535. It seems to me that AI currently can’t reason outside of its functional loop, even though that functional loop is now subjective and hard to define. Our newly subjective models can pull off a lot of heavy information work and can spark useful brainstorming, but they need input from beyond the reach of their processes. This is a feature, not a bug. Getting clever responses from AI and using those responses well requires creative reasoning.

Because I was cowboy-hacking based on a language model’s broadly cast artificial-neural dump from the silicone representation of the collective consciousness, I didn’t know that I was supposed to use the tool mintupgrade to manage the repository transition process. Perplexity had suggested I run mintupgrade after I update the repositories from una and focal to vera and jammy. I had rebooted after manually replacing the string ”focal” with ”jammy” in all the repo source configuration files.

I had forgotten at first to also replace every occurrence of ”una” with ”vera”. This was probably the cause of the broken DE. I’m not sure whether or not the DE would have broken after I had upgraded if I had initially done the Linux Mint repository transition as well as the Ubuntu transition.

From the broken system I encountered after the reboot, I simply had to edit the configuration files accordingly, update and upgrade again, reinstall Cinnamon, and then restart the X server with the command startx.

(If you happen to be reading this and are becoming frustrated that I have but vaguely described the process I went through, I can only apologize with an equally vague gesture toward the fact of mental exhaustion. I have had many maddening and even depressing experiences combing over step-by-step descriptions of others’ Linux troubleshooting writeups on forums and blogs and wikis, becoming increasingly perturbed by my promising progress and increasing understanding while tripping up without hope of recovery due to the omission of critical little details that I needed. This is a reason why LLM chatbots are helpful tools.)

Cover image generated using DALL-E.