Blog Post

Prmagazine > News > News > Switzerland launches its own open-source AI model
Switzerland launches its own open-source AI model

Switzerland launches its own open-source AI model

There is a new player in the AI ​​game, and it’s the whole country. Switzerland only Apertus was releasedits open source national big speech model (LLM) hopes to replace models provided by companies such as OpenAI. Apertus is the “open” Latin language of the world and was developed by Lausanne (EPFL), Eth Zurich in Zurich and the Swiss National Supercomputing Centre (CSCS), all of which are public institutions.

“Average, Apertus is the primary AI model for public institutions: models built for the public good. This is our best evidence that AI can become a public infrastructure such as highways, water or electricity.” Public Infrastructure.

The Swiss agency designed Apertus is completely open, allowing users to check any part of their training process. In addition to the model itself, they also published a comprehensive documentation and source code for their training process and the data sets used. They established Apertus to comply with Swiss data protection and copyright laws, which could make it one of the better choices for companies that want to comply with European regulations. The Swiss Bankers Association has previously said that the local LLM will have “long-term potential” as it will be able to better comply with Switzerland’s strict local data protection and bank confidentiality rules. Currently, Swiss banks are already using other AI models for their needs, so it remains to be seen whether they will switch to Apertus.

The new model can be used by anyone: researchers, hobbyists and even companies are welcome to tailor it to their needs. For example, they could use it to create chatbots, translators, and even educational or training tools. Apertus is trained in over 1,000 languages ​​in 15 trillion tokens, with 40% of the data being in other languages ​​other than English, including Swiss German and Romans. The Swiss announcement said the model trained only on public data, its crawlers respected machine-readable requests when they encountered machine-readable requests on the website. It should be noted that AI companies like Confusion have been accused of scratching websites and Bypass the protocol The creepers who intend to stop them. Some AI companies have been Prosecution News organizations and creatives use their content to train models without permission.

Apertus currently has two sizes, with 8 billion and 70 billion parameters. Currently available via Swiss information and communications technology company Swisscom or via Hug the face.

Source link

Leave a comment

Your email address will not be published. Required fields are marked *

star360feedback Recruitgo