SEO News You Can Use: Google Introduces MUM, a Language Model 1000x More Powerful Than BERT |

SEO News You Can Use: Google Introduces MUM, a Language Model 1000x More Powerful Than BERT

SEO News You Can Use

Google has been hard at work perfecting its understanding of language through various models. In the past few months alone, we were introduced to BERT, a machine learning technique that understands the nuances and context of words. Then along came SMITH, a model introduced in a Google research paper, that not only understands the relationship between words but also the context of paragraphs and how they relate to one another.

But forget about SMITH for a second. Also, BERT who? Now, MUM’s the word – Multitask Unified Model, that is. At Google’s I/O conference on Tuesday, Senior VP Prabhakar Raghavan showcased a new technology that has been described as “1000 times more powerful” than BERT in answering more complex queries. Dubbed MUM, the technology is trained on multiple languages and can perform numerous tasks all at the same time (hence the term “multitasking.”) 

Some of the things MUM is able to do simultaneously include:

  • Understanding and generating language
  • Acquiring knowledge of the world
  • Training across 75 languages
  • Gaining an understanding of multiple modalities, for example, images, text and video

Raghavan provided an impressive example of MUM in action, using the simulated Search query: “I’ve hiked Mt. Adams and now want to hike Mt. Fuji next fall, what should I do differently to prepare?” 

MUM enabled Google Search to show the differences and similarities between Mt. Adams and Mt. Fuji. It also pulled up articles covering the equipment needed to hike the latter. 

Because MUM is capable of developing an understanding of images and video, too, Google envisions a future where a user could take a photo of a pair of hiking boots and ask Google, “Can I use these to hike Mt. Fuji?” MUM would understand this type of query and answer questions that were previously thought too complex for a machine to comprehend. In other words, MUM could provide the same information any hiking expert would.

These advancements are, in a word, incredible. Google’s understanding of language has grown in leaps and bounds, and MUM is on the next level of sophistication. And as the possibilities of what we can search expand, so, too, will SEO. MUM is not ready to roll out just yet – Google is still in the process of rigorous testing and running internal pilot programs. But it seems that SEO could enter another stage of transformation sooner than we think.  

More SEO News You Can Use

Google Page Experience Update Will Also Apply to Desktop Search Results: If you thought Google’s Page Experience update only applied to mobile sites, you’re not alone. When Google announced last November that the update would affect Search results, the understanding was that only mobile Search would be impacted. Turns out this is not the case. During a Page Experience preparation session from Google I/O, Jeffrey Jose, a product manager on Google Search, said that plans have changed: The update is making its way to desktop Search too. Jose said Google was working hard to make the Page Experience live on desktop because the search engine believes these ranking signals are “critical” regardless of how a user is browsing the internet. As yet, it’s unclear if the desktop update will roll out at the same time as the mobile update, but Google has promised more information to come. There’s no way to avoid this anymore, so be sure to consult a reliable Google Page Experience guide to help you prepare.

Paid Search Campaigns Have No Impact on Site Rankings: In the latest Search Central SEO office-hours hangout, a viewer told John Mueller that their site rankings disappeared immediately from the search engine results pages (SERPs) after they began a Google Ads campaign. The moment they stopped the campaign, their site returned to the SERPs in the same position it had been before. This viewer has not been the first to report ranking changes – whether up or down – after launching a paid search campaign. Suspicious. But Mueller maintained these ranking changes would have been 100 percent unrelated to the paid search campaign. These two events, he said, would have no causal relationship at all. Mueller highlighted the fact that there is a strict separation between Search and Google Ads teams at Google, and these both exist on separate systems. In other words, this type of scenario, though common, is nothing more than a coincidence.

Google I/O Reveals Exciting New eCommerce Possibilities With Google and Shopify: The powerful partnership between Google and Shopify is set to expand even more. Google’s I/O conference on Tuesday revealed a wealth of new and upcoming eCommerce features that will change the online shopping experience significantly. All Shopify merchants will soon be able to list their products on various locations across Google with just a few clicks. Google is also improving the eCommerce experience by introducing brand-new ways to shop. One of the most exciting inclusions is the ability to purchase products directly from screenshots. Users will soon be able to take screenshots of items they would potentially like to purchase later and look up these products within any screenshot using Google Photos. Clearly, we’re only at the tip of the iceberg when it comes to eCommerce possibilities.

Don’t Be Too Concerned With Your Anchor Text Ratios: Anchor text is one of those aspects of SEO that the community keeps circling back to. Tons of research has been done on the topic, and some SEOs claim to have found common anchor text ratios among top-ranking sites. Many web owners and marketers believe in a golden ratio that makes ranking in Google easier. And it’s true that back in the olden days (2005), Google’s PageRank did, in fact, consider ratios. But today? Not so much. The argument is that Google’s algorithms can pick up on unnatural linking patterns and that these patterns are best avoided to “look natural” to Google. While we don’t know for sure whether anchor text ratios can result in a site being flagged for manual review, if all your links are natural and your SEO approach is white-hat, you have nothing to worry about. Not convinced? To put yourself at ease even more, check out Roger Montti’s Search Engine Journal (SEJ) analysis

In the Battle Between Quality Content and Quality Backlinks, Content Wins: We all know that the quality of website content and the quality of the backlinks pointing to your site are of critical SEO importance. They’re excellent partners in crime, but if they had to be pit against each other, content still takes the crown. This is what SEJ’s Adam Riemer said in response to the latest “Ask an SEO” question sent through by a reader. It’s a chicken-and-egg situation. First, before anything else, you need to provide a great experience on your site if you want to earn equally great backlinks. Even if you get tons of traffic from high-quality backlinks on other sites, if the content doesn’t satisfy what your users expect, they’re going to bounce. Focus on your content first, and the brilliant backlinks will follow.

Editor’s Note: “SEO News You Can Use” is a weekly blog post posted every Monday morning only on, rounding up all the top SEO news from around the world. Our goal is to make a one-stop-shop for everyone looking for SEO news, education and for hiring an SEO expert with our comprehensive SEO agency directory.

Leave a Reply

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

Related Posts

SEO News You Can Use: OpenAI CEO Offers Insights Into AI Oversight

May 22, 2023 by Robynn Farrell

It seems government regulators found an ally with OpenAI CEO and ChatG...

SEO News You Can Use: Google Admits Defeat to Open Source AI in a Leaked Internal Memo

May 15, 2023 by Arrabon Toribio

The tech world is intrigued by a recently leaked memo by one of Google...

SEO News You Can Use: Google Discourages Using Canonical Tags for Syndicated Content

May 08, 2023 by Arrabon Toribio

Google has already removed their recommendation to use canonical tags ...