Jeremy, firstly, thank you very much for all your valuable lessons. It has definitely been phenomenal!
Please can you share advices with us about what your approach or your journey was into getting into machine / deep learning? And how you stay on track with everything?
Thank you in advance. Much appreciated.
Do you plan to update the Fastai book and release a second edition at some point in the future?
Will you continue working on medical research/application, something like WAMRI in the near future?
Best place to discuss fastai development would be the Discord server, here is an invite:
Jeremy has mentioned that he does have plans for this in the future…
I’d be pretty curious to hear Jeremy’s thoughts on geometric algebra in general…
I have explored it a bit and it does seem like a very interesting way of looking at math and physics, and I am also curious what kinds of applications it may have in deep learning… I think @drscotthawley may have explored it a bit too? 
My question for Jeremy: What do you think about the increasing rhetoric and discourse regarding AGI (artificial general intelligence)? Especially as we have seen increasingly more capable and general models from research labs like DeepMind and OpenAI (ex: DALL-E 2, Gato, PaLM, etc.)…
As a follow-up to this, what should any student do, specifically to prepare for part 2 of the course?
TL;DR: learn how to drive the car before trying to be a mechanic.
My advice is to initially resist to look under the hood (otherwise you’ll get overwhelmed pretty quickly 
)
So use the library and pre-made notebooks, change them, try to get the “feeling of how the different hyper parameters affects the result”.
Moreover, try to focus on single, narrow scoped problem that works out of the box with fastai (ie: image classification).
You’ll gradually get a good understading of what is happening, that’s the moment to gradually open the hood.
Oooo! I’ll try to keep this brief and save the thread for Jeremy’s response because I’m very interested in his insights as well.
…but here’s my $0.02: You’ve touched on a topic for which my enthusiasm so far exceeds my proficiency. I was inspired to begin learning about GA from Eleonora Grassucci’s 2021 deep learning paper in the signal processing domain, and have since assembled a blog post of resources (YouTube videos, articles, etc) on the topic:
(There was a 2017 book chapter entitled “Outlook for Clifford Algebra Based Feature and Deep Learning AI Architectures” that may be of interest to you in particular, because its first author, Xiao Xia Yin, does a lot of medical imagery work!)
The growing popularity and utility of GA in computer graphics and signal processing, I think, also suggest great utility for deep learning, particularly in developing fast operations for multidimensional embeddings (e.g. CJ of dadabots advocated implementing some rotation operations for our generative audio model and I intend to…at some point…write this using GA [or PGA or CGA] notations). Steven de Kenick’s ganja.js JavaScript library for computer graphics is blindingly fast, and allows for computations involving points, lines, planes, circles, etc, using the same code for geometric primitives for arbitrary numbers of dimensions, offering advantages for readability & maintainability, and the fact that these code also tend to work in arbitrary number of dimensions (without having to rewrite matrix operations) I find…very appealing.
…it is still however a bit of a small but enthusiastic and growing community, so I’m not aware of any GA-based Deep Learning computation libraries yet… but probably one place to check – since we’re all Discord users — would be the Bivector Discord: cf. https://bivector.net/
PS- Actually, Eleonara has a (fairly) new paper on medical imagery too!: Apr 2022: “Multi-View Breast Cancer Classification via Hypercomplex Neural Networks”. (PyTorch code on GitHub)
I have nothing intelligent to say about this – other than that I’d be interested to learn about it.
So please feel free to be verbose!
You may find geometric algebra beneficial for your daughter’s schooling. My physicist friend tells me it’s a much better gate way into physics, giving deeper intuition as to what’s happening and why.
Jeremy did a talk about his Journey to Deep Learning .
Thanks very much for this. Much appreciated a lot.
Hi Jeremy, I can see people become professional deep learning practitioners (who are building and training great models mostly) through fastai courses, what do you think of the importance of learning to build datasets professionally (i.e., I mean building good kaggle-competition level datasets) for the people who eventually want to define and solve their own problems? Do you think you will teach us to do that someday?
In addition to Daniel’s question about home school, I am curious if you have any recommendations on home school curriculums, or other apps other than the dragonbox algebra 5+(it really helped my son too)? I would love to learn how you do your home school math lectures. I have tried to teach my son more advanced math than his grade level, but I have struggled with how and what I should teach.
I would be willing to pay way more and the average home school curriculum so that can teach my kids the advanced math concepts that you have. I am pretty sure I would learn from the lessons as well.
I started composing a reply that segued into considering teaching kids programming (for its algebraic benefits), but then ended thinking it might deserve a thread of its own… Teaching functional programming (to kids)
p.s. DragonBox blew my mind in the innovative way it teaches algebra to 4yrs olds.
Would you have the link the app, I wanted to get that for my 6yo … I think Roblox is not doing him any good 
infinite flexible function