Using AI to learn about Zynthian

Using Zynthian
1

Hi

I did an experiment in generating material with NotebookLM to improve my understanding.

NotebookLM (Google NotebookLM; LM short for “Language Model”) is a research and note-taking online tool developed by Google Labs that uses artificial intelligence (AI), specifically Google Gemini, to assist users in interacting with their documents.

Screenshot_20260305-152804
Screenshot_20260305-1528041080×751 64.9 KB

WARNING: All content below is hallucinated by AI, don’t trust it for wiring up your Zynthian!

Generated audio overview “podcast” about the topic.

Slideware and visuals

Zynthian_Hardware_Blueprint.pdf (17.6 MB)

Building_Zynthian_Controls.pdf (16.9 MB)

unnamed
unnamed1920×1072 456 KB

What other ways have you used AI to help with your Zynthian journey?

Regards

1 Like
2

I’m relatively sceptical about AI in these matters. I think AI showed a typical result here: The broad “general” informations were maybe 65% correct, the more specific informations (including everything you might want to know how to connect things to things) were utterly wrong. This bot connected “clock” to encoders and put the I2C lines to 5V and this was enough to know for me.

But from a design POV it’s not bad!

3

Hi Hannes, agreed, we need to be sceptical. On the other hand it did manage to distill quite a few gotchas in the visuals and audio. Also, this was based on generating without extra human input, which could have produced better results.

4

I’d be honestly interested in the gotchas. But I’d weight them against the almost certain fatal damage to the machine when sticking to these suggestions.

5

Well, I am glad to be recognised, even if it is by some atoms in the web! We take our reward where we can find it. :wink:

1 Like
6

I particularly like what it-they did with the multi-meter leads:

image
image606×617 94 KB

6 Likes
7

It’s learning a bloomin sight quicker than my kids did….

1 Like
8

This is a very interesting approach to user familarisation and training. I just finished listening to the 20 minute episode and it is mostly very accurate, provides some great info and is fairly easy to listen to. One must try to avoid being distracted by knowing the two voices are not really human and worse… ignore how much priase @wyleu gets! (Much soup for you tonight.)

I think we could possibly use this to create some curated introduction tutorials. This one has a great intro about what zynthian is and a section on building plus some sections on customisation and diy ethos. I would like to see those in seperate podcasts. @jawn can you provide some info on how you prompted NotebookLM to produce this? Maybe we could use it (or similar) to create specific, targetted content. Are we able to adjust the output, e.g. correct errors or misconceptions?

Does anyone know the licensing aroung the output of such a technology? Are we free to create media and distribute it?

There may also be a way to hook in a LLM to zynthian’s workflow recorder to generate a description of some demonstration.

1 Like
9

I’ll add that when I was writing a wall of text in the better out of box experience issue,I thought I’d liked to a gemini chat that was helpful, but not correct enough to get me to an “aha” moment. I think some of the questions and (correct or incorrect) answers could help to distill what difficulties a newb like me is having when getting started.

10

I set the notebook to public, see if you can check the sources etc.

All I had to do was

  1. Create a notebook with the prompt from my original post
  2. Accept the suggested sources (can be tweaked afterwards)
  3. Generated audio overview (and other material)
1 Like
11

Thanks! Very interesting this AI virtuality…

Cheers,
Maarten

12

Interesting, just listened to it, and granted: the overview was more accurate. I peeviously just looked into the pdfs above, which were awful.