Switzerland launches open source artificial intelligence model

Photo of author

By [email protected]


There is a new player in the artificial intelligence race, which is an entire country. Switzerland only has I came upusThe Open National Language Model (LLM), which hopes to be an alternative to the models offered by companies such as Openai. Apertus, Latin for the world “Open”, was developed by the Swiss Federal Technology Institute in Lausanne (EPFL), ETH Zurich and the Swiss National Computing Center (CSCS), all of which are public institutions.

“Currently, Apertus is a major model for the general year: a model built by public institutions, for the public interest. It is the best evidence that artificial intelligence can be a form of public infrastructure such as highways, water or electricity,” said Joshua Tan, a leading supporter in making artificial intelligence such as highways, water or electricity. Amnesty International is a general infrastructure.

The Swiss institutions are designed to be completely open, allowing users to examine any part of their training process. In addition to the same model, they issued comprehensive documents and the source symbol of their training process, as well as the data sets they used. They built Apertus to comply with Swiss data protection laws and copyright laws, which may be one of the best options for companies that want to adhere to European regulations. The Swiss Banker Association previously stated that the local LLM will get “long -term potential”, as it will be able to better comply with the rules of strict local data protection in Switzerland and banking secrecy rules. Currently, Swiss banks are already using other models of artificial intelligence to meet their needs, so it remains to see whether they will turn into Apertus.

Anyone can use the new model: researchers, amateurs and even companies are called to build on it and allocate it to meet their needs. They can use it to create chat tools, translators and even educational or training tools, for example. Apertus has been trained on 15 trillion symbol across more than 1,000 languages, with 40 percent of data in languages ​​other than English, including German Swiss and Romanch. The Switzerland advertisement says that the model was trained only on the data available to the public, and the crawling requests that are respected by reading reading of the machine when they chanted it on web sites. It should be noted that artificial intelligence companies, such as confusion, were previously accused of stripping websites Cross the protocols It means preventing their crawling. Some artificial intelligence companies were also prosecution By news organizations and creativity to use their content to train their models without permission.

Apertus is currently available in two sizes with 8 billion and 70 billion teachers. It is currently available via Swisscom, which is the Swiss Information and Communications Technology Company, or via Embroidery.



https://s.yimg.com/ny/api/res/1.2/V21jyZsUwd0yZMOSDhf56Q–/YXBwaWQ9aGlnaGxhbmRlcjt3PTEyMDA7aD03OTk-/https://s.yimg.com/os/creatr-images/2020-06/c7354350-b6f9-11ea-bbfb-dde9a5673b07

Source link

Leave a Comment