During watching Lesson "0"

I didn’t want to but couldn’t find Category so I am publishing here with a claim about the site of The Forum:

We need a NEW Category! Part 1 (2021) !!!

while my original one was:
Hi there,
it is June 2021 and I am watching Jeremy’s brilliant, as usual lecture of the new series (? I hope!) of Fast.ai course in 2021!
… and yes, I am one of those constantly failing to finish it students…

“Tenacity” and “Finish a Project” are the keywords of the first part of the Lesson “0” video, and I will definitely will be taking all those advices to heart this time…

My problem is not “too much preparations” b/c I still remember enough of Linear Algebra, Calculus, C++, and keep on using git, editor, etc, and programming every day…
The problem may be that I don’t care about my portfolio… that I don’t look at the course as a leverage for getting a better job…
And yes, I am here b/c I am fascinated with a mystery of Intelligence and Consciousness and it seems like Deep Learning is a widely open road towards understanding them…
, well maybe describing them with existing concepts rather.
And maybe because of impractically and elusiveness of my goals I keep on failing?

But yes, I have a little project on my mind which potentially may be great and which would require a lot of polishing…
It is about making automatic keyword captions of any given lecture video, points of meaningful notions with timestamps, and after applying feedback (constant tuning?) from the user it would gradually make the process sharper and more in line of goals of the user…
Maybe it would lead to tagging concepts with set of keywords and associating different videos (or paragraphs in books or papers) talking about the same things?
Does it look like feasible?

1 Like

Hi Peter

Yes it must be amazing to use Deep Learning to saving humanity where as the rest of us are the Deep Learning for a hobby. So you are proposing a note taker for students. Consider if Jeremy or Rachel says:

There is an assumption about the 1-Lipschitz function but we can effective ignore it by imposing our own ceilings but see the original paper but you can set the ceiling as a hyper parameter to determine the limits before instability.

So you would need to build an audio to text process to identify all the words. You could build a language classifier to to take a sentence and generate multiple categories. So S1 contains 1-Lipschitz function and hyper parameter and instability. S5 in video 9 contains the same categories. So video 1 is similar to video 9 . It perhaps reminds me of finding the latent factors in the Movie Lens lesson.

I think it sounds like a very exciting project for you.

Regards Conwyn