Multitask Unified Model – What Is MUM Technology?

Multitask Unified Model – What Is MUM Technology?
17 September 2021
According to unofficial sources, Google implements several hundred algorithm updates every year. Most of them introduce only minor changes unnoticeable to the vast majority of SEO specialists and search engine users. During Google I/O 2021, one of the most important events in the technology industry, Google announced the launch of a new technology known as Multitask Unified Model (MUM). The solution is supposed to initiate significant changes in the search results. MUM - what is it and what modifications can you expect in the SERPs once the technology is implemented?
Table of Contents:

What Is MUM Technology? Google Multitask Unified Model

At the very beginning, note that the conference wasn’t about a new algorithm but about technology. Of course, at this stage, it's hard to predict what solution will eventually be delivered, but it's likely that MUM (Multitask Unified Model) will consist of smaller algorithms and, at least in theory, can significantly alter the search results. How does the MUM technology work? Or actually, how is it supposed to work? Google has prepared an extensive blog entry: MUM: A new AI milestone for understanding information. This blog post provides an example of a query typed by a user who has climbed Mount Adams, a mountain in the USA, in the state of Washington. His next destination was Mount Fuji in Japan. The user searched Google for information on how to prepare for the expedition, what should be done differently and how to approach things in general. Currently, getting appropriate answers requires sending multiple queries, whereas finding a result that would cover all the aspects is rather infeasible. [caption id="attachment_42875" align="aligncenter" width="604"]What is MUM technology Source:[/caption] Now, it’s impossible to ask the search engine for multiple issues and obtain answers to these questions right away. However, thanks to the MUM technology, Google is likely to become a real expert and provide answers to very complicated queries. All this without having to type in numerous key phrases. The results? A user entering a query similar to the one in the screenshot below will get a comprehensive answer and learn how to plan a trip. [caption id="attachment_42877" align="aligncenter" width="605"]mum technology how does it work source:[/caption]

How is MUM Going To Work?

With MUM, the search engine is supposed to understand your queries better and give you a response as if you asked multiple questions (such as precipitation on Fuji, what to take to Fuji, trails on Fuji, or differences between climbing Mt. Adams and Fuji) at once. These are only some of the examples. After entering such phrases, you should immediately get tips on how to prepare for climbing these two different mountains. Direct Answers can be considered a makeshift of the new solution  - the results that appear in the so-called zero position and, according to Google, answer a given query best. But are these search results going to be somehow related to each other? We’re still talking about the technology that is supposed to be implemented and probably even Google specialists aren’t familiar with its final version. When reacting to the query, Google is expected to analyze multiple sources and find answers to all the questions that are of interest to users. Thanks to MUM, robots are supposed to scrutinize content not only in the source language which should enable using multiple sources and providing even more accurate answers. With MUM, the search engine will generate content in 75 languages! Moreover, multimedia will be analyzed together with the text. After the BERT update, the MUM technology is the next step towards a better understanding of queries. Back then, in 2019, Google started to better comprehend words used to ask questions. MUM is expected to be another milestone. It's supposed to be based on the so-called "transformer architecture", just like BERT. increase your visibility


Actually, it’s hard to predict what the SERPs will look like once the MUM technology is implemented. So far, we can only guess that since the user will automatically receive an answer for a complex query, he/she will not visit many pages that would be entered during a normal search. If information is displayed similarly to the abovementioned Direct Answers, it’ll definitely have a negative impact on CTR. On the other hand, we can’t determine the form of presenting the answers. It’s hard to imagine responses for complex queries - after all, they would include lots of content. It's worth noting that thanks to the fact that the Google MUM technology is going to be based on multilingual sources, your website is likely to reach Internet users from all over the world, not only those who enter queries in your native language. If the new technology appeals to users and they get used to typing in complex queries, it may cause a dramatic change in the methods of browsing the net. Moreover, we can expect a profound modification of search volumes. There will be fewer phrases with very high search volumes and queries will become more diverse. This means that SEO will need to focus primarily on long-tail phrases.

How is MUM Going To Change Content?

Nowadays, content creators have different approaches to writing articles. Both extensive texts that cover multiple topics and shorter ones that address selected elements of a particular field are highly popular. The fact that Google will be able to respond to expanded queries may motivate authors to create elaborate entries showing the subject from different perspectives. Perhaps it’ll be a step that will allow you to benefit from the new technology.
Logo Delante

Delante is an online marketing agency specialized in generating website traffic from search engines and paid ads (SEO / SEM). Over 80% of our clients are from rapidly growing e-commerce industry.

Leave a comment


Get a
Free Quote


european search awards finalist

Check our Free SEO and Marketing Tools!

Free SEO and Marketing Tools