AI Tinkerers Ottawa v2.5.0 trip report
Published on , 686 words, 3 minutes to read
I love this meetup so much
I live in Ottawa, the capital of Canada, and there's surprisingly very few meetups there. Like, it's somehow a G7 capital that claims to be a "tech hub", but there's basically no meetups worth going to.
There is one glaring exception though, AI Tinkerers Ottawa. It's a meetup that's been running since last year, and it's quickly becoming my favorite tech meetup in general. It's an AI meetup, but not that kind of AI meetup. It's not a bunch of people trying to shill you their startups. It's not one of those low-signal high-noise environments where people are thinking it's gonna be the next coming of sliced bread. It's really about people talking with eachother about what they can do with AI models and then working together to make it happen.
AI Tinkerers is a federation of meetup groups that are all focused on having builders come together to learn about new things and then use them to make something cool. It's a breath of fresh air in an industry that is so dominated by hype cycles. It's a place where you can genuinely come in with some skills and an idea, then find community to help you make it so.
This month I got to sponsor the food for the event. I've never sponsored an event like this, and it was fun to be able to do that. In terms of the DevRel stuff I do, I tend to wait for people to come up and talk with me (to make the more transactional bits of the conversation feel more natural and less forced), but when you're sponsoring the food people will come up to you a lot more. It made me a bit worried that things would be lower signal/higher noise, but it all turned out great.
The person who usually does the recordings of the talks was at another conference this time, so I stepped in and was able to use my portable recording studio setup to save the day. This is one of the first times I've recorded this for other people and I'm not sure if the mic was peaking too much, but I'm sure there's something salvageable in there.
I love this meetup. It's so much networking, so little hype, and so much signal. I can't wait for the next one. Maybe I won't choose Pizza Pizza for the food vendor next time, they didn't really have good options for plates. Only these sad little cardboard squares that failed to hold the weight of the pizza slices, causing us to lose brave solders more than I'd want. I'll have to think about that for next time.
Things to look into
Every time I go to one of these meetups, I make a list of "things to look into". These are things that are brought up in the opening address, the talks, or conversations I have. Here's the list from this time:
- ChromaDB - local vector database, python ecosystem
- Confidence score for vector database matches as a way to test "strength" of results
- Archie / 8base - UI creation from images/mockups
- Prompting foundation models is easier than you think, may need to write something on it
- Mixture of agents to beat GPT-4
- Refusal in Language Models Is Mediated by a Single Direction
- Langgraph - langchain but a graph
- Facebook released a bunch of new models, the one where they generate four tokens at once for the rate of generating three tokens traditionally is interesting as hell
Facts and circumstances may have changed since publication. Please contact me before jumping to conclusions if something seems wrong or unclear.
Tags: