Google’s New AI Model: Who is MUM?
Google LLC is an American multinational technology company which provides search engine technology, online advertising, cloud computing, computer software, quantum computing, e-commerce, and artificial intelligence. Its current headquarters are in Mountain View, a city in Santa Clara County, California which is located in the heart of Silicon Valley.
Over the years, Google has updated many of its features. From the occasional themed interface to the search algorithm, Google gets better every year. As another step towards improving the search algorithm, Google launched a new AI update.
MUM is short for Multitask Unified Model and it was introduced in May 2021 at the Google IO Conference as a Google Search update to improve user experience.
Why the Update?
Google has a feature called “snippets” which highlight or “top” the response to a search through describing a page which comes before a page link. The results that are displayed in this way are called featured “snippets”.
Snippets are concise results that Google packages and delivers to the top of the page for different searches. While snippets are inherently helpful and efficient, they are not that trustworthy. After all, humans have not fact-check or curated these snippets. They are algorithmically driven, like most searches.
There are common problems that are associated with snippets. Snippets can have issues, for example, a reasonable retort to a flawed query. Snippets don’t always result in a correct and factual response to a user’s search. Google has bridged this gap through the MUM update.
More about MUM
A better way to answer everyday questions, MUM is an algorithm to search the Internet across different languages, text, images and media and helps give unified results for more complex searches. It not only provides exact information but also suggests related information.
MUM has a self-checking mechanism through cross-referencing. It can check snippets against an established high-quality search result. In this way, the AI can determine the similarities and coherence of information. The result should be consistent even if the query is in different wording or language. Along with understanding language, MUM can also generate it.
The new AI model will also solve Google’s “reasonable-seeming” problem. The new ranking system can help the AI determine the level of appropriateness of the result depending on the premise of the query. If the question is structurally false, then a snippet won’t appear. This is helpful for questions that do not have answers or at least a factual one.
The AI is trained across 75 different languages and bids farewell to Google’s BERT that was used to understand user’s search queries and the contents indexed by the search engine.
Google will also expand the use of warning messages whenever the search engine fails to procure a result if the overall search results do not meet a specific quality standard.
- Google has a new AI search innovation called the “Multitask Unified Model.”
- The new AI empowers the search engine’s capacity to deliver quality results to users.
- MUM is 1000 times more powerful than BERT and also removes language barriers because it delivers answers from search engines that are originally in other languages.
- This AI update will improve the quality of snippets at the top of the page and address the “reasonable-seeming answers” to queries that inherently have no answers.
Our Trademark Lawyers in Sydney are Experts when it comes to Registering Trademark or Trademark Opposition Process in Sydney
Bianca “Bianx” Ysabel, Digital Administrator
Disclaimer. The material in this post represents general information only and should not be taken to be legal advice.