THREAD. You have not read Tolkien's The Lord of the Rings until you've read it out loud.

Seriously, the language *soars* when spoken. Last night I was reading the scene where Gandalf and co. fight off Wargs beneath Caradhras, and the beauty of the language he employs brought me nigh upon tears. TEARS.
That is a minor and otherwise-forgettable scene. You likely don't remember it. They didn't bother to put it in the movie. You read right over it if you're reading silently to yourself.
Reading it out loud though, suddenly these mythic figures of Gandalf, Aragorn, Gimli, and Legolas stride out of legend and onto the field of imagination and strike a pose in the firelight, surrounded by glimmering green eyes, and you see them as if for the first time.
I am not a visual person, but the image is seared into my mind the day after reading it.
Tolkien does this with heightened, poetic language. (Yes, the heightened language that his detractors complain about.)
(It might also help if you took a medieval literature course at some point in your life, and read some of the earliest surviving English poetry, so that you recognize what he's doing.) :D
You also won't notice it as easily until you read it aloud. On the page, it's just funny squiggles that convey meaning. But the funny sentence structure and the alliteration has a real purpose, the same purpose it has in the epic poetry of oral cultures:
Tolkien's use of language, when read aloud, is engaging to the ear and paints a vivid picture of heroic deeds.
You have not read Tolkien's The Lord of the Rings until you've read it aloud.

Bonus points if you read it aloud to an eight-year-old who's hearing the story for the first time. This also heightens the effect. :)

/end

More from All

How can we use language supervision to learn better visual representations for robotics?

Introducing Voltron: Language-Driven Representation Learning for Robotics!

Paper: https://t.co/gIsRPtSjKz
Models: https://t.co/NOB3cpATYG
Evaluation: https://t.co/aOzQu95J8z

🧵👇(1 / 12)


Videos of humans performing everyday tasks (Something-Something-v2, Ego4D) offer a rich and diverse resource for learning representations for robotic manipulation.

Yet, an underused part of these datasets are the rich, natural language annotations accompanying each video. (2/12)

The Voltron framework offers a simple way to use language supervision to shape representation learning, building off of prior work in representations for robotics like MVP (
https://t.co/Pb0mk9hb4i) and R3M (https://t.co/o2Fkc3fP0e).

The secret is *balance* (3/12)

Starting with a masked autoencoder over frames from these video clips, make a choice:

1) Condition on language and improve our ability to reconstruct the scene.

2) Generate language given the visual representation and improve our ability to describe what's happening. (4/12)

By trading off *conditioning* and *generation* we show that we can learn 1) better representations than prior methods, and 2) explicitly shape the balance of low and high-level features captured.

Why is the ability to shape this balance important? (5/12)

You May Also Like