Meta issue for structure and overall approach

[davanstrien]

Possible structure from discussion with @mark-bell-tna

Based on some of the feedback, we want to update the structure to address feedback and also aim to pack the content that most people will be interested in upfront.

• Start with applications and uses of machine learning right at the start. What can machine learning do? Focus on tasks, i.e. something a bit like https://huggingface.co/tasks
  • Maybe start with generative AI? even how you might use it for emails, but what about GLAM uses
  • Then more specific tasks: creating search queries, i.e. using chat GPT to help create advanced queries
  • More GLAM-specific examples (more about projects)
• How does this work:
  • Impact of data?
• Then move to ethics and limitations
• Machine learning projects at the end: add more about APIs, ecosystems, commercial relationships, etc.

The community-led materials section, i.e. for demos, could be helpful, but we don't want to include it in the main lesson because otherwise, content will be out of date quickly.

[davanstrien]

Other tasks for lesson

• template slides for delivery
• resources for further learning
• how to get started
• review objectives
  • How to engage with companies
    • How can we work with commercial companies
  • thinking about where AI might be embedded in other services

[MikeTrizna]

Yes! I absolutely agree with getting right into "unstructured data" Tasks! Although do we have to caveat the title or the introduction?

I wish there was a better, beginner-friendly term to differentiate between tabular structured data vs. unstructured text, images, audio, etc. When I brought up the term at the workshop last week, there was immediate pushback that their image of text datasets were organized and standardized, and thus structured.

[leighphan]

@davanstrien @mark-bell-tna I agree this is a great direction to updating the lesson structure. I included some feedback on the general structure of the lesson that can apply here as we consider restructuring in https://github.com/carpentries-incubator/machine-learning-librarians-archivists/issues/82#issuecomment-1683079126, including:

• Subdivide the episodes into smaller "bite-size portions"
• Reduce prerequisite knowledge of existing AI/ML terms
• Standardize use of vocabulary throughout the lesson.

During the sprint I believe @mark-bell-tna mentioned an earlier discussion about including the environmental impacts of using generative AI in the lesson, which I think would make sense to include in the "How does this work: Impact of data?" section.