Google MUM: the new tool that will change search results – the example with Covid | Technique

Google MUM: the new tool that will change search results - the example with Covid |  Technique

Google prepares changes to the search field using an algorithm called MUM. It aims to provide instant and accurate results in complex searches, effectively saving the user from multiple searches.

Google created MUM in an open source architecture. Like the BERT model released in 2019, MUM is a context-based artificial intelligence that can understand and create language. When used, MUM will be able to appear on top of some Google search results to answer questions like an expert. Google says the MUM algorithm can even generate images and create unique articles and explanations like a journalist.

The article continues after the announcement

In a presentation to the media (among them iefimerida.gr) Fellow and Deputy Head of Research, Pandu Nayak, referred to the modus operandi of MUM.

Google MUM: Names of Covid-19 vaccines

Skewer, straw, ball, football, sauce, ointment. There are many things that are said by different names. Sometimes this is due to the function of language, but other times it depends on social trends, different concepts or, simply put, where you are in the world.

A closely related example is Covid-19. As people around the world were searching for information, we had to learn to recognize all the different phrases they used about the new virus to make sure we were providing high-quality, immediate information from trusted health organizations like the World Health Organization and control centers. and prevention of infection. A year later, we face a similar challenge with vaccine names, but only this time we have a new tool at our disposal which is the Multitasking Unified Model (MUM).

READ  Electronic Foundry Points out Why ‘Halo Infinite’ Appears to be So Bizarre

Understand searches for vaccine information

AstraZeneca, CoronaVac, Moderna, Pfizer, Sputnik and all of the widely available vaccines have different names around the world — more than 800, according to our analysis. People looking for information on vaccines can search for keywords like “Coronavaccin Pfizer,” “mRNA-1273,” “CoVaccine” — and more.

The article continues after the announcement

Being able to successfully identify all of these names is essential to providing the world with the latest and most reliable vaccine information. However, identifying all the ways people refer to vaccines around the world is a time-consuming matter that requires hundreds of hours of work.

Using MUM, we were able to identify 800 different vaccine names in over 50 languages ​​in seconds. After confirming the MUM results, we added them to a Google search so users around the world can find instant, high-quality information about Covid-19 vaccines.

Find reliable information about Covid-19 on search

Transfer of knowledge into different languages

MUM was able to complete a project that would normally take a week in just a few seconds, thanks to its ability to impart knowledge. MUM can extract and impart knowledge from the 75 languages ​​in which it is trained. For example, imagine that you are reading a book. If you are a polyglot, you can share its content in the other languages ​​you speak – depending on how well you know them – because you have an understanding of the book that does not depend on language or translation. MUM can and does not impart knowledge in the same way.

Likewise, through its ability to transfer knowledge, MUM does not need to learn new skills in every language – it can simply transfer them, thus helping to make scale improvements quickly, even when there is not enough data to process. This is thanks to MUM’s sampling efficiency – which means that MUM needs less data than previous models to perform the same task. In the case of vaccines, with a small sample of their official names, MUM was able to quickly identify variants in other languages.

Improve your Google search with MUM

This first application of MUM has helped us convey important information to users around the world in a fast way and we look forward to seeing the different ways in which MUM will improve search to be more useful in the future. Initial tests have shown that MUM can not only improve many aspects of our existing systems, but also help us create entirely new ways to search for and explore information.

Follow iefimerida.gr on Google News and be the first to know all the news

Find the latest news from Greece and the world at ifimerida.gr

Leave a Reply

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