Google MUM (Multitask United Model) is the new AI centred all-rounder algorithm, whose purpose is to get you the most relevant information from a variety of sources, all at the same time. But isn’t that what Google is now? Well, kind of. Google is always trying to understand search queries, the language used, and how we talk conversationally. The more understanding of the question, including its context, the better the results.
In 2019, Google Released BERT (Bidirectional Encoder Representations from Transformers), which used natural language processing (NLP) to accurately assess a query as a whole, rather than individual words. For example, a hypothetical search for something like “taxi” could have previously produced the result “The History of Taxicabs”. With BERT, the top results would be local taxi services.
BERT seems pretty good. However, Google has upped the ante by introducing MUM, promising to be 1000x more powerful than BERT. What sets MUM apart from BERT? MUM can successfully multitask simultaneously from a variety of technology sources including audio, video, text in over 75 languages. Meaning you only have to ask one question, instead of ten.
MUM was introduced at I/O 2021 by Google’s Prabhaker Raghavan, who demonstrated MUM using a complex search query “I’ve hiked Mt. Adams and now want to hike Mt. Fuji next fall, what should I do differently to prepare?” Previously, the user would have to search multiple queries and piece together information for themselves to get their answers. BERT, even though pretty good at understanding language and nuance, is limited by text. Google MUM goes to the next level and incorporates images and text within the search, making it multi-modal.
Google wants to give you all the answers you’ll need themselves. Presenting you with rich results in an even quicker time. Google is still testing MUM at this stage but if BERT is anything to go by, it won’t be long before MUM is rolled out. Google is putting their eggs in the AI basket, creating faster and more advanced ways of consuming content and finding rich results, answering questions we didn’t before we’ve even asked them.
Google’s first real-world application involved using it to understand variations of searches made for COVID-19 vaccine information. MUM was able to identify over 800 variations of how vaccine names were being searched in over 50 languages, saving hundreds of human hours and providing information-rich results. We’re looking forward to seeing how MUM will change the face of Google, SERPs and SEO.