homeaboutarchivenewslettermembership!
aboutarchivemembership!
aboutarchivemembers!

kottke.org posts about translation

Google’s MusicLM Generates Music from Text

posted by Tim Carmody   Feb 01, 2023

A screenshot of Google's Music LM's examples of Painting Captioning Conditioning -- Dali's the Persistence of Memory, a portrait of Napoleon, and Henri Matisse's Dance are all converted to captions and then music is created from the captions

Google Research has released a new generative AI tool called MusicLM. MusicLM can generate new musical compositions from text prompts, either describing the music to be played (e.g., “The main soundtrack of an arcade game. It is fast-paced and upbeat, with a catchy electric guitar riff. The music is repetitive and easy to remember, but with unexpected sounds, like cymbal crashes or drum rolls”) or more emotional and evocative (“Made early in his career, Matisse’s Dance, 1910, shows a group of red dancers caught in a collective moment of innocent freedom and joy, holding hands as they whirl around in space. Simple and direct, the painting speaks volumes about our deep-rooted, primal human desire for connection, movement, rhythm and music”).

As the last example suggests, since music can be generated from just about any text, anything that can be translated/captioned/captured in text, from poetry to paintings, can be turned into music.

It may seem strange that so many AI tools are coming to fruition in public all at once, but at Ars Technica, investor Haomiao Huang argues that once the basic AI toolkit reached a certain level of sophistication, a confluence of new products taking advantage of those research breakthroughs was inevitable:

To sum up, the breakthrough with generative image models is a combination of two AI advances. First, there’s deep learning’s ability to learn a “language” for representing images via latent representations. Second, models can use the “translation” ability of transformers via a foundation model to shift between the world of text and the world of images (via that latent representation).

This is a powerful technique that goes far beyond images. As long as there’s a way to represent something with a structure that looks a bit like a language, together with the data sets to train on, transformers can learn the rules and then translate between languages. Github’s Copilot has learned to translate between English and various programming languages, and Google’s Alphafold can translate between the language of DNA and protein sequences. Other companies and researchers are working on things like training AIs to generate automations to do simple tasks on a computer, like creating a spreadsheet. Each of these are just ordered sequences.

The other thing that’s different about the new wave of AI advances, Huang says, is that they’re not especially dependent on huge computing power at the edge. So AI is rapidly becoming much more ubiquitous than it’s been… even if MusicLM’s sample set of tunes still crashes my web browser.

Lydia Davis on Translation and Learning Languages

posted by Tim Carmody   Dec 10, 2021

Lydia Davis.jpeg

My favorite contemporary writer is probably Lydia Davis, in no small part because I don’t know if anyone takes a finer care for the language they use, as a writer and reader.

Davis also does double duty as both an original writer of fiction and essays, and a translator of other people’s writings, in multiple languages. In her new collection, Essays 2, she describes her unusual technique:

Although she learned German by immersion, Davis’s preferred method of language acquisition is quite different, and, to an outside observer, demonically challenging: She finds a book published in a language that she does not fully or even partially understand and then tries to figure out what it means.

To improve her Spanish, she digs into a copy of “Las Aventuras de Tom Sawyer.” In some cases the decryption proves easy. Words like “plan” are the same in English and Spanish. In other cases she inductively reasons the meaning of a word after noticing it in different contexts. Hoja initially stumps her when it pops up in the phrase hoja de papel — “hoja of paper.” Later in the book, it occurs in the context of a tree. Finally, Huck wraps a dry hoja around something to make a cigarette, and Davis realizes that only one meaning would work as well with paper as with a tree or a cigarette: “leaf.” Of course, it would be possible to solve the hoja enigma in two seconds by plugging the word into Google, but that would destroy the fun.

I’m (re)learning Italian right now — I sort of learned it backwards the first time, starting with Dante and Petrarch and only now learning how to ask where the bathroom is (dove el gabinetto?) and the difference between coat (cappotto) and hat (cappello). But what remains exciting are the little associations you learn, the conjunctions of phrasing, the possible substitutions of one term for another, the way a question and an answer can reflect the same structure — a map of phonemic possibilities that is also a way of seeing the world. Davis’s method might be impractical for learning a second language, but for a gifted language learner, it seems to put a premium on finding those connections. Which is, indeed, a big part of the fun.

Translating Homer in public

posted by Tim Carmody   Mar 16, 2018

siren vase 2.jpg

I can’t claim to have finished Emily Wilson’s translation of The Odyssey by Homer — epic poems are, well, epic — but I’m a huge fan of everything I’ve read, and especially Wilson’s Twitter feed, which is often devoted to explicating some small bit of Homeric text and comparing her approach to that of other translators.

Here, for example, she takes on the depiction of the Sirens. I’m going to pick and choose a few tweets, but you should read as much of the thread as you can.


This last observation prompted a haunting distillation by Lev Mirov of Odysseus’s journey and his encounter with the Sirens:

Back to Wilson, who translates the brutally short passage of the sirens this way:

She explains:

Translation is hard, but translation in public is harder and better. There’s a richness in the commentary, and also a reckoning with the accretion of meanings that have come down through past readings, that you don’t often get without diving into scholarly apparatus. It’s not just peeling back the plaster; it’s trying to understand the work that plaster did in holding the whole structure together. Just remarkable.

Update: Dan Chiasson wrote about Wilson’s use of Twitter for the New Yorker.