Machine learning is one of the coolest topics you can introduce in a Hack Club meeting. The sheer amount of complicated and intimidating theory required to understand machine learning used to make it nearly impossible to teach it to beginners. Thankfully, services like Teachable Machine have changed that.
Jeewoo runs a Hack Club in Vancouver. In this meeting, he runs Hack Club's Teachable Machine workshop. His club normally holds meetings in person, but this meeting was held online over spring break, with about half their normal attendance. Jeewoo's club normally feels like a party, but this meeting feels a lot more like a chill hangout, complete with a chill Spotify playlist, a blue plushie, and two guitars.
What Jeewoo does right
Jeewoo takes advantage of the smaller-than-normal meeting size to create a laid-back, tight-knit environment without it feeling awkward or low energy. There are a few things in particular that he does to make this happen:
He takes things very slow, explaining even basic concepts very thoroughly, stopping to ask if anyone needs help, and making sure everybody is on board before continuing.
A good place where you can see this is from to in the recording.
Another good place where you can see this is from to . He begins by asking generally if everybody's ready, then at asks people individually if they're ready before moving on.
He throws in multiple mini-hacking sections throughout the meeting. Each one lasts 5-10 minutes and everybody in the meeting has made something small by the end of each one. To fill the silence, he plays music from a Spotify playlist.
See to see him introduce the first hacking section, and to see the first music from the playlist play.
As people begin getting their machine learning models to work (see to see a club member get excited when his machine learning model begins working), he starts making conversation and asking people questions about their models (e.g. "How many image samples did you guys use?" at ).
This is a really nice touch: during one of the later hacking sections, at , instead of playing the Spotify playlist, he gets out his guitar and starts playing it. This helps create a super cozy feeling, as if the club is actually a tight-knit group of friends hacking around a campfire.
He stops to ask lots of questions while he's explaining concepts so that his club members stay engaged.
Before the first hacking section, at , he asks everyone to go around and say what two classes they plan on adding to their machine learning model.
He stops to clarify points throughout.
Example: after someone asks a clarifying question, Jeewoo very thoroughly answers, including giving general tips for everyone to make their models better ()
After the meeting, he wrote up a short summary, including key takeaways and even a summary of the workshop instructions. He shared this writeup (included in the "Resources" section below) with his club members.
Wait about 5 minutes for people to come in. Make conversation with people as they start arriving and play music in the background to set the vibe.
Have the Hack Club Teachable Machine workshop opened on a separate device before the meeting starts. You'll be using this workshop as a "script" throughout the meeting.
Introduce Teachable Machine
Introduce the concept of machine learning to your club members.
See to , especially , for a great example.
Introduce Teachable Machine. Explain what it's used for and why you're using it for this meeting. Send your club members the link to Teachable Machine and have them open it in their browsers.
See to for a good example.
Walk your club members through creating an Image project on Teachable Machine. Once everybody has created an Image project, show them an example for how to create two models.
See to for a short example.
Run the first hacking section
Ask your club members to take about 5 minutes to make their own machine learning model. Turn them loose.
Start playing music in the background or otherwise filling the silence with something.
Make yourself available for questions. Don't be afraid to explain "too much" if someone asks a question.
Export the machine learning model and make a website
Once your club members have created their first machine learning model, use the instructions in the workshop to guide them through exporting the model and uploading it to repl.it.
If you have time (and if your club members have made websites before), walk them through customizing the website with a little bit of CSS.
Run the second hacking section
At this point, you've reached the end of the workshop. Congratulate your club members on having made their first ever machine learning model 🎉, and start another hacking section.
Show your club members how to add another class on Teachable Machine. Then turn them loose for another 5 or so minutes to add a third class to their machine learning model
Demo projects
Go through everybody in the meeting (if you have a large attendance, go through 5-10 people) and have them show a demo of their machine learning model. Take as little or as much time as you want.
Have your club members share a screenshot of their projects in some central place (Slack, Discord, whatever).
End the meeting
Once the meeting is over, casually wind down as people leave. Make some final conversation, start playing music again.
Notes
Jeewoo's meeting was optimized for a small club meeting of about 5 or 6 people—but don't worry if you run a larger club! The vibe you create in your meeting is entirely up to you, and all of the steps to run this meeting can be applied just as easily to small meetings as they can for medium and large meetings.
Occasionally, even if you prepared everything beforehand, you may get lost or stuck at some point during a club meeting. This happens to every leader, and whether or not it ruins your club meeting is up to how you respond to it.
If you respond by getting nervous or frustrated, and you spend 30 minutes trying to debug your meeting in front of your club members, they will lose interest in the meeting, they will lose trust in you, and you will run out of time to finish the workshop.
Jeewoo ran into this during this meeting. See the progression between , . Jeewoo responds (at ) by moving on to the next step. Near the end of the meeting, they come back to it, figure out the problem ( to ), and get it working (). This was a great response because the flow of the meeting wasn't very disrupted, and it felt good to come back later and solve it together.