Understanding natural language for search engines is a complex task, but thanks to Google’s continued investment in artificial intelligence, it has reached a whole new level. In 2021, Google introduced a new Multitask United Model or MUM algorithm. According to Google MUM, this new algorithm is 1000 times more powerful than the BERT algorithm published in 2019. As a result, MUM will enter Google products shortly.
Google MUM Algorithm: What is Multitask Unified Model?
MUM is a language model built on the same BERT transformer system introduced in 2019. BERT is a powerful language model that made significant progress at its release. However, the capabilities of the MUM algorithm are increasing: according to Google, it is 1000 times more powerful than the BERT algorithm.
Much of MUM’s power and capability comes from the fact that it is a multitasking algorithm that can perform several tasks at once. There is no need to do jobs one after the other; MUM is here to do several studies simultaneously. This means that Google’s new algorithm can read the text, understand the meaning, build deep knowledge about the subject, and use video and audio to enhance and enrich it. Google MUM uses over 75 languages and translates these findings into multi-layered content to answer complex user questions. All these things are done at once! Notable, isn’t it?
How robust is the MUM algorithm?
In 2021, Prabhakar Raghavan (one of Google’s specialists) explained how this algorithm works. First, he asked the complicated question, “I’ve conquered Mount Adams, and now I want to climb Mount Fuji next fall; what should I do to prepare?” to show what the MUM algorithm can do. In a typical search session, you must search all the different aspects yourself. Then, once you have everything, you need to combine it to answer all your questions.
Currently, MUM combines insights from many different sources into various aspects of the search, from measuring mountains to recommending rain gear (because autumn is the rainy season on Mt. Fuji) to extracting information from Japanese sources because more content is written in that language about this particular topic.
In complex questions like the above example, it all comes down to the combination of beings, emotions, and the intention to understand the meaning of something. Unfortunately, machines have trouble understanding human language, and algorithms like BERT and MUM can almost approximate themselves to natural language.
MUM goes a step further by processing language and adding video and images because it can do multiple tasks simultaneously. This makes it possible to create a valuable result that answers the user’s question by providing a whole new piece of content. The MUM algorithm is even built into Google Lens, so you can point your camera at your hiking boots and ask if they’re suitable for hiking Mt.Fuji.
Of course, the ultimate goal of all these algorithms is to help you get more information – most likely within Google’s domain – with fewer searches. As a result, we have seen a steady increase in valuable results and detailed answers that are becoming more intuitive and prominent every day and available to us in the shortest time. Many of these developments, both inside and outside of search, paint the picture of Google wanting to answer more of your questions.
MUM Algorithm wholly based on artificial intelligence
Google is very quietly becoming a search engine fully equipped with artificial intelligence. However, even search engine isn’t the right word because it’s more like a knowledge delivery device.
Increasingly, Google opens up ways to enter information from other sources, such as microphones, cameras, televisions, wearables, and smart speakers. As a result, searches and how they are presented must change to achieve all these goals and find a way for machines to behave more logically. For example, a microphone on the fitness tracker should hear your queries and understand them, while the voice assistant should do something with that information and helpfully respond to them.
Understanding language is the key to search engines. Therefore, developing compelling, efficient, and flexible language models that can generate content to provide those answers concisely and naturally will be essential.
Questions that arise about the google MUM algorithm
Of course, we know that all these changes create questions for people. For example, if Google can read, hear and see content in all languages and re-present it in a new format – depending on the context and content generated by artificial intelligence – who owns this content? And who is responsible for these automated results?
And what about bias in artificial intelligence? Bias and ethics are essential issues in artificial intelligence. If we are talking about the fact that artificial intelligence will play a significant role in our future life, we must be sure of its impartiality and reliability. Of course, Google mentions AI bias in an article and is still training this model. AI researchers at Google also talked a bit about KELM: a way to identify false information and harmful content in texts that might eventually bias models.
When will Google launch the MUM algorithm?
Google is testing the MUM algorithm and will continue to do so until it is confident. There is no specific timeline on when it will launch, but it didn’t take long for the BERT algorithm.
The introduction of MUM may mean better search results, but it could also mean a new type of search results. This can affect how you think about content. For example, answering the audience’s questions and solving their problems may not be so crucial because it is probably already done by the system. Instead, improving your product and focusing on building brand preference may be better. People should come across your brand in the best possible way. Think of ways to shine and be seen, and find ways to turn traffic to your site into regular customers.
How to prepare for the MUM algorithm
Google’s algorithm will soon act like a human reading and understanding text. So, how can you prepare for this new language model?
First, by adding schema types to your pages, you can give search engines like Google more information about your content. This helps Google understand what’s on a page and qualifies you for higher rankings on its results page. If you don’t have much programming experience, the Yoast SEO plugin can automatically add the necessary structured data to your site.
In addition, you should have good texts that are easy to read and naturally optimized. To get a better ranking from Google, do not fill your text with keywords; try to write something your audience will be encouraged to read. Something new, engaging, and well structured. Using the possibility of readability analysis in the Yoast plugin can help you. This will give you detailed feedback on your text and what you can improve on it.