My Personal Ethical AI Journey

Disclaimer: Every post I write on this site is 100% human generated. My little fingers typing on a keyboard and nothing more. This post in particular is an opinion post and not a guidebook for ethical AI use. Maybe there’s no ethical AI usage? I don’t know, but this is where I’ve landed.

The Need vs Hype

When AI first started gaining traction, I was excited. Wow. a tool that’s great for pattern recognition, summarization, research, and more. But, I think we can all agree that what it has become isn’t what was truly promised. We’ve tried to replace search engines with it and it’s just not what it was meant for. Heck, we’ve tried to replace EVERYTHING with it. It’s a prediction based model; asking it questions you need 100% certainty on is foolish. Generating “art” with it is a creative nightmare that makes the world more grey.

Beyond that, so many people think it’s a silver bullet to replace workers, artists, and their own thinking power. The environmental impact, the disruption of the supply chain for data centers, all of it just left a bad taste in my mouth. I wanted to fully reject it but if I’ve learned anything from technology hype cycles, it’s that there actually can be a use or a need buried in the hype.

What Do I Use This For?

What’s funny about AI is the wrestled question I’ve had - what do I use this for? Everyone around me has been using it to summarize and write emails. Okay, I like to write though (clearly.) I don’t mind reading. And a summary can sometimes be as long as the email. The prompt people type up can be the same length as the email they needed to send. Like, just send me the prompt next time. I’d rather read that than paragraphs of fluff.

As I’ve already mentioned, search engine replacement is just a bad idea. The hallucinations of AI has somehow become worse over the years, not better. I caught myself arguing with AI more than getting the answers I actually needed.

I seriously have been wracking my brain to find a place to slot (or slop) in this technology.

Use cases that I intend to use AI for:

  • Summarization of long winded/heavy legal documents (ie Privacy Policies, ToS documents, etc)
  • Bespoke technical setups/research
  • Boiler plate for firewall rules. Looking at you, IP Tables!
  • Dungeons and Dragons Oracle (No seriously, I’ll explain with another post later.)

Ethically Running AI (Just for Me)

As mentioned above, I just don’t think cloud use of AI is ethical. Data centers are poisoning water supplies and disrupting communities. Billionares are fighting for user retention, data, and control. Eh, I’d rather do these tasks manually than engage with that.

I’ve determined offline/personal AI use is the only “ethical” way to engage with AI. At least, ethical to me. I’ll be honest, I don’t think there’s a true ethical way to engage with AI. To create AI, companies had to siphon all the data from the internet. So, true ethical use of AI just isn’t going to happen. But this feels less evil to me.

My Offline LAN Only AI PC Setup
  • Old gaming PC running a GTX 1070 and 16 GB of RAM
  • Centos Stream 10 that auto sleeps after 15 minutes
  • Wake on LAN capabilities to wake it up
  • Ollama locally installed

When I want to ask the AI a question, I SSH into my home lab, wake up the PC using WOL, and SSH into it. It’s not perfect but I’ve been using it for the use cases above for about a week. Since the computer sleeps after 15 minutes, I’m not using unnecessary power and am only paying for the prompts I ask it. That actually incentivizes me more to only use it when I’m stuck.

What’s next?

I’ve been doing some reading on PewDiePie’s Odyssey application. I’ll be real, I don’t think I’m quite ready for that but that’s probably where I’ll go next if I need more functionality. For now, this is working and I feel like I’m able to participate in a new technology without feeling like I’m dooming the planet further.

Technical Stuff

Okay, so you want to do this yourself? Well, here are a few things I’ve discovered.

  • You need the official nvidia drivers. There should be a lot of tutorials online how to do this. Make sure you blacklist any other drivers!
  • You need some grub arguments to make sleep work nvidia-drm.modeset=1 nvidia.NVreg_PreserveVideoMemoryAllocations=1
  • The script I use to auto-sleep is below with the crontab (sleeps every 15 min of inactivity.)

auto-sleep.sh

#!/bin/bash
# taken from https://dgross.ca/blog/linux-home-server-auto-sleep

logged_in_count=$(who | wc -l)
# We expect 2 lines of output from `lsof -i:548` at idle: one for output headers, another for the 
# server listening for connections. More than 2 lines indicates inbound connection(s).
afp_connection_count=$(lsof -i:548 | wc -l)
if [[ $logged_in_count < 1 && $afp_connection_count < 3 ]]; then
  systemctl suspend
else
  echo "Not suspending, logged in users: $logged_in_count, connection count: $afp_connection_count"
fi

Root Crontabs */15 * * * * /root/auto-sleep.sh | logger -t autosuspend

References: