Google strives to make the search as comfortable as possible, and the result - the most relevant to the user's request. Google representatives report this before each major update and introduction of new technologies.
So, in 2019, BERT appeared, focused on recognizing atypical, complexly formulated queries. Thanks to this innovation, Google has a better understanding of the context of search phrases given by the user's natural language. At the same time, the demand for LSI copywriting increased. When generating content based on it, search prompts and low-frequency “tails” are used, for example, “buy an iPhone 11 for 128 gigabytes red in Kharkov”. As a result, website promotion by keywords has taken on new forms, and the relevance of the content to the user's request has become even more important and significant.
Now, after 2 years, the Google search engine is introducing MUM technology, developed on the basis of the architecture of transformers, but having much more efficiency. According to Google, MUM is 1,000 times more powerful than BERT. The Multitask Unified Model is a multitasking unified model designed to radically change the principles of processing search queries, including those set by voice search. Due to this, Google strives to show the user the most accurate information and the most useful information.
Talking about the new MUM technology, the head of Google's search division Prabhakar Raghavan gave a simple example during his presentation: “I climbed Mount Adams, and next year I want to conquer Mount Fuji - what should I do differently?”. Now, to answer this question, you need to perform several search queries, and analyze 3-4 sources of information. Google hopes that the MUM technology will help the bot understand the user from 1 time and provide him with an immediate answer to his request.
Of course, the technology is currently working in test mode and is not yet capable of handling such complex requests. Moreover, Google has not officially announced the full launch of the new product. But we already know that MUM functions by analogy with human thinking, is able to quickly learn and scale. Moreover, the knowledge gained by the system in a specific language and in a specific area does not require re-studying the topic in other languages.
A representative of Pandu Nayak Corporation said at the presentation that Google MUM was created with an emphasis on the synchronous understanding of different types of information, not only text, but also pictures, audio and video content. From this we understand that website optimization should include not only work with texts, but also with other elements located on the page. Users are expected to be able to receive responses to combined queries, such as uploading a photo of a pair of shoes asking “Is this good for a run?” or "Can this be worn under a dress?" The request, by the way, can be both printed and voice.
The purpose of the first test of the multitasking unified model was to determine the ability of the system to correctly identify the names of vaccines against coronavirus. The result is over 800 variations in 50+ languages, found within seconds. AstraZeneca, Pfizer, CoronaVac known to us have a lot of other names in the world. As it turns out, people enter Google searches for "Coronavaccin Pfizer", "CoVaccine", "mRNA-1273" and others. Without a MUM at its disposal, a search engine would have to process a colossal data set containing all the possible names of a vaccine in the world. This would have taken, according to preliminary estimates, several weeks.
SEO promotion of a website, if it is implemented with an emphasis on high-quality, relevant content and trust backlinks, is unlikely to change in any way due to the release of the Multitask Unified Model technology. After all, you need to initially create a site for people, fill it with expert materials, respond to user requests, and not adapt to the new Google algorithms.
The MUM system is only designed to help people find relevant, reliable and high-quality information in a timely manner. As preliminary testing has shown, the technology has a very promising future, and the introduction of a multitasking unified search model will not only improve the aspects of user interaction with the search engine, but also open up new ways to study and process information.
Google Search has long ceased to be a simple list of links. A quick response block, extended snippets, knowledge graphs have already become familiar aspects of Google's work. MUM technology will only make the search results more accurate, deeper and more relevant. And if your site does not meet the requirements of the search engine, then with each subsequent update of the Google Search system, it will be more and more difficult to break into the TOP. Do you want to get there? Then order the services of the Light Spider marketing agency and take leading positions in the search results.