Meet MUM: The New Technology being tested by Google
There’s a revolutionary technology for answering intricate search queries and it’s called Multitask Unified Model, otherwise known as “MUM”.
Now Google has launched a brand-new technology called Multitask Unified Model, nicknamed “MUM”. Its goal is to answer questions of the more complicated variety.
At its I/O conference on Tuesday, Google disclosed that it’s currently investigating how MUM can be used as an enhanced way of delivering useful responses and understanding language.
While MUM hasn’t launched yet, Google’s revealed a few sneaky peeks as to how it’ll be developed.
What is Google's 'MUM' technology?
The main idea behind MUM is that it helps get things done in a lot less searches.
According to Google’s latest research, it takes eight searches to complete difficult tasks.
What Google envisions is a world in which MUM generates a response to intricate questions such as: “I’ve climbed Mount Fuji and next, I want to climb Mount Kilimanjaro this summer, what do I need to do to get myself prepared?”
Currently, you’d have to search the internet for every relevant piece of information to get an answer to the question. You’d need to find out how high each mountain is, what the average temperature is in the summer months, how difficult the climb is, the correct equipment you’ll need, and much more.
Google assesses the response MUM would deliver to the answer a mountaineer or mountain professional would give in relation to the question.
MUM will recognise the user is contrasting two mountains, indicating that height and climb info may be applicable. Next, MUM could acknowledge that to “be prepared” could include actions such as getting the right equipment and fitness training.
What’s more, MUM can skim insights connected with a profound understanding of the world. In turn, it may indicate information such as summer is a hot time of year on Mount Kilimanjaro, so the user may need a sun hat and plenty of water in their daypack.
What makes MUM stand out is its ability to search for useful subtopics for a thorough investigation, such as the top training workouts and the best equipment. The search results it brings up would take the user to useful blogs, videos, articles and pictures from around the globe.
The technology behind 'MUM'
MUM is built on a Transformer architecture, like BERT. The difference is that it’s 1,000 times more powerful and boasts the capacity to multitask to get the right information, in the right way.
Google says that MUM can translate info in different formats and languages, meaning it can give the most extensive results:
Google vows to apply this innovative AI to search results responsibly and has begun internal pilot testing of MUM.
As with every update to Google Search, MUM will experience the same strenuous testing before it’s launched. It also states it’ll especially concentrate on patterns that signify bias in machine learning.
Keep an eye out for MUM-driven features and improvements coming to Google products in the next few months and years.