Skip to content

Latest commit

 

History

History
76 lines (49 loc) · 6.08 KB

roadmap.md

File metadata and controls

76 lines (49 loc) · 6.08 KB

Roadmap (subject to perennial revision)

The learning plan is split into multiple courses. The goal is for each stage in each course to have unique objectives, a clearly defined set of terms and concepts, and 1-3 assignments.

As this AI-learning knowledge graph grows, it will become part of a custom AI tutor that can lead learners through each course and stage. I'll provide the knowledge graph here too, eventually, so that you can see how it grows and develops.

Future iterations of this project will extend it to new courses; a few I'm plotting are intro-level humanities, a degree program in PPE (Philosophy, Politics, & Economics), classics, Shakespeare, & more. I'd love, for example, to integrate classic syllabi from great teachers into this format, so that we could participate, in some form, in the best courses from the past.

Current planned major revision: rebuild this roadmap based on the Dreyfus model of learning & mastery. To achieve this, I'll have to add additional sections toware the end of the course that will provide some additional levels of learning, along with more specific ways to track progress using the categories the Dreyfus model provides.

Learning AI 001: Prereqs

In this course, you'll learn the basics. If you're a coder, you've probably already done most of this. If not, don't worry—you can get through these objectives relatively quickly, and they'll lay the foundation for everything we'll do in the future.

Do these steps on your own. Once you've done them, come back here and follow the rest of the program.

  1. Create your learning map.
  2. Create your Github account.
  3. Make a blog.
  4. Learn how to use github & the pull/commit/branch model. Extra credit: get comfortable with the command line & learn to use github through its command-line interface.
  5. Learn the basic concepts. I learned a lot from Michael Woolridge's book A Brief History of Artificial Intelligence and Andrew Ng's AI for Everyone course. Write an essay articulating everything you've learned. Try to integrate as much as possible. It won't be perfect. That's ok. You can make it better as you learn more.

Learning AI 002: Working on AI, with AI

It's always beneficial to integrate your tools, objectives, and learning resources as much as possible. If you're learning knowledge graphs, build a knowledge graph about what you're learning. If you're learning AI, try using AI. Figure out its current strengths and weaknesses. Explore the amazing new tools popping up every day. At the very least, this process will help you understand where AI is weakest and needs improvement. At best, you'll find that many of these AI tools can be very helpful in your learning process, if you use them productively.

A cautionary note, though: you won't learn much if you let AI do everything for you. Especially if you let it draft your essays. When it comes to learning, the obstacle is the way. Essay writing is a profoundly helpful way to learn. But it's helpful because it's so counter-cultural: slow, difficult, even boring at times. Force yourself to do it and you'll learn a lot more—not just AI, but also about writing and, most importantly, about yourself.

In this course, we'll explore the different ways to use AI to assist learning. Objectives & assignments coming soon.

  1. Read the intro.
  2. Learn the basics.
  3. Deploy a standard open-source LLM, such as one from HuggingFace or Llama 3.2.
  4. [in progress] Train & deploy an LLM. Then test it and find its weaknesses. This section is long and should be subdivided.
  5. Train your first RAG-based LLM & deploy it; then find its weaknesses.
  6. Write an essay about what you learned in this unit.
  7. Assess where you're at and where you're heading. What are your weak spots so far? Work on those. What are you enjoying most? Keep pursuing that.

Learning AI 102: Developing knowledge graphs

  1. Learn taxonomy
  2. Learn ontology
  3. Learn knowledge graphs by building one based on the ontology & taxonomy you developed previously.
  4. Write an essay on everything you learned about information architecture in this unit.
  5. Assess where you're at and where you're heading. What are your weak spots so far? Work on those. What are you enjoying most? Keep pursuing that.

Learning KG+AI 201: Mastering GraphRAG

  1. Learn the basic patterns for integrating knowledge graphs & LLMs.
  2. Build a knowledge graph w/Neo4j or Protege.
  3. Train your first KG+RAG LLM & deploy it, then find its weaknesses.
  4. Learn about AI ops: testing, maintaining, & updating the tools you've built.
  5. You got it: write the essay.
  6. Assess where you're at and where you're heading. What are your weak spots so far? Work on those. What are you enjoying most? Keep pursuing that.

Learning KG+AI 202: GraphRAG Patterns

This course will explore many different design patterns for graph RAG applications, considering the advantages and disadvantages of each.

Learning AI 301: Graduate Seminar / Dissertation

OK, when you make it this far—(let's be honest, when I make it this far, too)—you're going to be very good at building AI and using it for all kinds of different purposes that will help you reach your goals and fulfill your potential. I think the next step is to go do whatever it is that led you here in the first place.

  • Start a company.
  • Get a new job.
  • Build something amazing.

Now go learn something else—and build an AI+KG tutor to help yourself & others!


Reference Materials

Bibliography | Glossary