Tag: Meta

Posted on

A Universal Translator Comes

The idea of a Babel Fish comes from the legendary novel or series of novels “The Hitchhiker’s Guide to the Galaxy”. Douglas Adams alluded to the Tower of Babel. In 1997, Yahoo launched a web service for the automatic translation of texts under this name. Various attempts to implement the Babel Fish in hardware and software followed. Meta’s SeamlessM4T software can handle almost a hundred languages. In a blog post, the American company refers to the work of Douglas Adams. “M4T” stands for “Massively Multilingual and Multimodal Machine Translation”. Again, it is a language model that makes spectacular things possible. It has been trained on four million hours of raw audio. A demo is available at seamless.metademolab.com/demo. The first step is to record a sentence. The sentence is displayed as text. Then select the language you want to translate into, for example Japanese. The sentence is displayed again in text form and, if desired, in spoken language. A synthetic voice is used. You can also use your own voice, but this is not yet integrated into the application. A paper by Meta AI and UC Berkeley can be downloaded here.

Posted on

A New Language AI

“Meta’s AI lab has created a massive new language model that shares both the remarkable abilities and the harmful flaws of OpenAI’s pioneering neural network GPT-3. And in an unprecedented move for Big Tech, it is giving it away to researchers – together with details about how it was built and trained.” (MIT Technology Review, May 3, 2022) This was reported by MIT Technology Review on May 3, 2022. GPT-3 (Generative Pre-trained Transformer 3) is an autoregressive language model that uses deep learning to generate natural language. Not only web-based systems, but also voice assistants and social robots can be equipped with it. Amazing texts emerge, and long meaningful conversations are possible – almost like between two real people. “Meta’s move is the first time that a fully trained large language model will be made available to any researcher who wants to study it. The news has been welcomed by many concerned about the way this powerful technology is being built by small teams behind closed doors.” (MIT Technology Review, May 3, 2022)