SCaLE 22x & Reaffirming Hope

Yesterday was a really good day. The first really good day I’ve had in what seems like a very long time.

Some months ago I submitted a proposal for a talk to be given at the Southern California Linux Expo, or SCaLE for short. As you might expect based on the name, SCaLE is an event where Linux users of all ages, shapes, and sizes gather to share information and come together with other like-minded persons. It is usually held in Pasadena, CA, at the Pasadena Convention Center, on the opposite side of Green Street from the Paseo, a popular place for shopping and dining.

Given that artificial intelligence (AI) and large language models (LLMs) have become my latest technological obsession, I decided to talk about how people can set up and customize their own local LLM for fun, profit, or just because they have security and privacy concerns about using third-party LLM services like ChatGPT, Claude, Copilot, DeepSeek, or Grok.

I titled my talk, Keeping Things Local – Making Your Own Private LLM, and while I will give this talk and workshops based on it again several times over in the next few months, this was the first time I would be presenting on this particular topic before a live audience. Because LLMs are the new hotness in the tech field, I expected there would be some interest, but I didn’t know how much interest my talk would garner. I was just glad for a chance to get out of the house, hang with some friends I don’t get to see IRL very often, and try to give something back to the universe as thanks for opportunities I’ve been given, personally and professionally, by other geeks.

When I got to SCaLE Saturday morning, one of the first things that struck me was the diversity of the attendees. There were people dressed in modern corporate costumes, with polished coifs and sleek attire. There were also young geeks whose deportment screamed of social awkwardness and too many hours spent in dark rooms, staring at monitors, running commands on their Linux variation of preference. There were the old guard tech bros with grumpy faces, grizzled beards, and long silvery locks barely constrained in their ponytails. And there were many other variations of geekdom, with multi-hued hair and sticker-covered laptops. Yeah, it was good to be home.

In addition to the various geek personas in attendance, I noticed that the ethnic and gender diversity seemed more in evidence than in years past. The dominant skin tone was still decidedly on the light side, but the number of shades present, ranging from from brown to black, seemed higher. Likewise, I noticed more girls, fems, and somewhere-in-the-middle people, all there to spend time with others, sharing and smiling and laughing together, their body language reflecting how at ease they were. It was beautiful, and I couldn’t help but smile at the sight.

I checked in at the speaker’s registration desk, got my badge and swag bag, then found a spot where I could assemble my badge and add some ribbons. I debated about applying the “expert” ribbon, but finally decided to go for it, if only to beat my imposter syndrome into submission. Then I wandered around, located Ballroom F, where my talk was scheduled, and settled down to wait.

The talk immediately before mine was a deep dive into data science techniques for standardizing job descriptions using open source AI tools. It was fascinating, and dense, but the presenter knew his stuff and was very much deserving of the applause he received.

His talk ended early, so about 15 minutes before I was scheduled to begin, I set up my laptop, took the headset mic from the SCaLE volunteer working the room, then did a little pre-show banter as the room began to fill, getting a feel for the audience by asking a few warm-up questions like, “How many of you are here at SCaLE for the first time?”, plus some questions to determine the level of knowledge in the room re: AIs and LLMs, specifically.

I was especially grateful for the chance to ask people about their existing knowledge about AIs. I had been worried that my talk, which began with foundational level information, might be too focused on the basics. As it turned out, the level was just right for the bulk of the audience, so that was reassuring.

What I didn’t expect was how MANY people were interested in the talk. When I switched from the pre-show to the official content, the room was already full, and as I kept talking, more people kept showing up.

I began with an overview of AI in general and how LLMs fit into the larger AI ecosystem, moved on to how LLMs and other generative AIs work, and then dove into system requirements and software options for working with LLMs.

Then came demo time.

I’ve been doing public speaking for longer than I like to admit to anymore, but it wasn’t until fairly recently that my talks were of the sort where live demos were involved, or at least not live demos that relied on a functional internet connection. Then I tried giving a talk a while back on prompt engineering with ChatGPT, and the internet bandwidth at the conference venue was so bad that the site would not load at all and I couldn’t do any of the demos I had planned. Now I record my demos to thwart the cruel demo gods and ensure that my demos work every time. It also lets me skip past boring parts like downloading files.

Not too surprisingly, my talk went long. This wasn’t too much of a problem as it had been scheduled right before a long break. The recorded demos began with installing Ollama in a WSL instance, moved on to pulling and running a couple of LLM models, then wrapped with a simple, then complex, customization of one of the models I had downloaded to the system.

Sadly, because I went overtime, that last bit got cut off in the YouTube stream. The good news is that I can re-record the demos (the video got glitchy toward the end) and add a voiceover so I can post it to LinkedIn or YouTube.

When I wrapped my talk, the audience applauded and there were lots of smiling faces. Then the SCaLE volunteer took a mic around the room so people could ask questions and be heard clearly. Then the official event was over and a bunch of us stood around talking, asking and answering questions, and eventually the last of us had to be kicked out of the room so the next speaker could give his presentation.

On my own again, I stopped by the Starbucks adjacent to the conference center for a snack and a drink. Then I found a quiet spot where a table and chair were set up outside the ballrooms and settled down to break my fast (I don’t like to eat before a talk) and peoplewatch.

It was nice to watch the flow of people going past, singly and in groups. Friends who saw me stopped for a chat. Others who had seen my talk also stopped by, said thank you, asked questions, and moved on. Then I noticed the posts and comments start to appear on LinkedIn. I’m still humbled by all the positive comments about the talk, and by how many people showed up for it.

The rest of my day was spent with friends, old and new, chatting and sharing, having dinner at El Cholo, then going for ice cream at Wanderlust, a short walk down Colorado Boulevard. Finally, I was spent and decided to head for home. Fun as it had been, the day had been long and I’m not as spry as I used to be.

For me, the key takeaways of the day were about hope and community. The entire day had been about people accepting each other where they are, sharing what they know, freely and without hesitation, and with a sense of joy in the giving and receiving of knowledge and encouragement to continue to learn and strive. I know others had different experiences, but I hope most of them were positive, too.

So long as people can come together as they did this weekend, sharing knowledge and empowering each other to grow and learn and achieve great things, I can hold on to the hope that we will survive as a species. In these troubled times, filled with greed and hate coming from so many sources, I needed that, and I will do my best to pay that hope forward to others.