[FSFLA] Google TPU Tensor Processing Unit initiative

[willi uebelherr]

(español detrás)

Dear friends,

most of us are working in the direction of free technology. One of our 
important project is the global interconnection of all language spaces. 
The principle is to use a inter-lingua for exchange. The best way is a 
plan-language, a syntetical language like Esperanto.

In the space of Sanskrit based languages like latin and indo-germanic it 
is clear. For slavic and east asian languages it is open.

Google in this time use englisch as his intermedia language. Therefore, 
all what we can't transport in english, don't exist there.

But here we find some interesting documentation about neuronal networks 
combined with a specific chip. The principle: Massivly parallel.

TPU Beat Intel, Nvidia, says Google
http://www.eetimes.com/document.asp?doc_id=1331560

Quantifying the performance of the TPU, our first machine learning chip
https://cloudplatform.googleblog.com/2017/04/quantifying-the-performance-of-the-TPU-our-first-machine-learning-chip.html

Google supercharges machine learning tasks with TPU custom chip
https://cloudplatform.googleblog.com/2016/05/Google-supercharges-machine-learning-tasks-with-custom-chip.html

TensorFlow: An open-source software library for Machine Intelligence
https://www.tensorflow.org/

The study with pdf download:

In-Datacenter Performance Analysis of a Tensor Processing Unit​
https://drive.google.com/file/d/0Bx4hafXDDq2EMzRNcy1vSUxtcEk/view

The implementation in machine translation

A Neural Network for Machine Translation, at Production Scale
https://research.googleblog.com/2016/09/a-neural-network-for-machine.html

many greetings, willi
Asuncion, Paraguay


QueridAs amigAs,

la mayoría de nosotros estamos trabajando en la dirección de la 
tecnología libre. Uno de nuestros importantes proyectos es la 
interconexión global de todos los espacios lingüísticos. El principio es 
usar un inter-lingua para el intercambio. La mejor manera es un lenguaje 
de plan, un lenguaje sintético como el Esperanto.

En el espacio de los idiomas basados en sánscrito como el latín y 
indo-germánico está claro. Para los idiomas eslavos y asiáticos 
orientales está abierto.

Google en este tiempo usa englisch como su idioma intermedia. Por lo 
tanto, todo lo que no podemos transportar en inglés, no existen allí.

Pero aquí encontramos una documentación interesante sobre las redes 
neuronales combinadas con un chip específico. El principio: Masivamente 
paralelo.

TPU Beat Intel, Nvidia, says Google
http://www.eetimes.com/document.asp?doc_id=1331560

Quantifying the performance of the TPU, our first machine learning chip
https://cloudplatform.googleblog.com/2017/04/quantifying-the-performance-of-the-TPU-our-first-machine-learning-chip.html

Google supercharges machine learning tasks with TPU custom chip
https://cloudplatform.googleblog.com/2016/05/Google-supercharges-machine-learning-tasks-with-custom-chip.html

TensorFlow: An open-source software library for Machine Intelligence
https://www.tensorflow.org/

El estudio con descarga de pdf:

In-Datacenter Performance Analysis of a Tensor Processing Unit​
https://drive.google.com/file/d/0Bx4hafXDDq2EMzRNcy1vSUxtcEk/view

La implementación en traducción automática

A Neural Network for Machine Translation, at Production Scale
https://research.googleblog.com/2016/09/a-neural-network-for-machine.html

muchos saludos, willi
Asuncion, Paraguay