Every key idea is an equation
Attention, backpropagation, cross-entropy and KL divergence are defined in symbols. Skip the symbols and you skip the idea.
Machine learning · deep learning
Study ML on the move, equations and all.
Transformers, optimisation and probability are written in notation. Aethel turns that notation into clear spoken English, keeps the printed equation in front of you, and explains a step when you ask.
Five books and an hour of narration free. No card needed.
The problem
The field moves weekly
New papers arrive faster than anyone can read them at a desk. Listening turns dead time into reading time.
Self-study has no lecturer
When a derivation jumps a step, there is nobody to ask. That is where most self-taught learners stall.
How Aethel helps
Notation spoken clearly
“θ ← θ − η∇L(θ)” becomes “theta is updated to theta minus eta times the gradient of L of theta”.
Ask for the missing step
Stop on a derivation and ask why it holds, or ask for a small numeric example. The answer is read aloud too.
Papers and textbooks together
Import arXiv papers next to free textbooks such as Mathematics for Machine Learning or Dive into Deep Learning (text-based PDFs).
Offline for the commute
Save chapters to your phone and listen underground or on a plane, with the highlighted text still in sync.
What you will hear
These are produced live by the same speech engine the app uses. Press Listen to hear them in your device's voice; in the app you choose a studio or ElevenLabs voice.
Attention(Q, K, V) = softmax(QKᵀ / √dₖ)V
Aethel says
Attention of Q, K, V equals softmax of the quantity QK transpose over the square root of d sub k, times V
Vaswani et al., “Attention Is All You Need”, NeurIPS 2017, equation 1
θ ← θ − η ∇L(θ)
Aethel says
theta is updated to theta minus eta times the gradient of L of theta
X ∼ N(0, 1)
Aethel says
X is distributed as N of 0, 1
Get started in three steps
- 1Create a free account.
- 2Paste the arXiv link of the paper you are reading this week.
- 3Listen to section 3 first: that's usually where the model is defined.
Questions
Can I listen to a whole textbook?
Yes, chapter by chapter. Each plan includes a number of narration hours a month; replaying something you have already heard is free.
What about code listings?
Code is shown on screen as printed. Long listings are better read than heard, so tap the next paragraph to carry on from there.
Compared with
Hear your own book this way.
Free to start: five books and an hour of narration, no card needed.