.png){kind=logo}
Nearly two months after announcing the Multitask Unified Model (MUM), Google began testing out its first application of the technology and is reporting some pretty impressive results. According to a blog post published on July 6, Google was able to identify more than 800 variations of vaccine names in over 50 languages in just seconds during its first-ever application of MUM. The blog post stated that without MUM, this would have taken the search engine weeks to complete.
After validating MUM’s findings, Google applied them to Search so that people could “find timely, high-quality information about COVID-19 vaccines worldwide.”
“Our ability to correctly identify all these names is critical to bringing people the latest trustworthy information about the vaccine,”said Google. “But identifying the different ways people refer to the vaccines all over the world is hugely time-intensive, taking hundreds of human hours.”
What is MUM and How Does It Work?
Announced back in May at Google I/O, MUM is a technology developed by Google that is built on a transformer architecture. It has been compared to the BERT algorithm that rolled out in 2019, however, Google is reporting that MUM is 1,000 times more powerful and has incredible multitasking capabilities.
Google explained that MUM can learn from and transfer knowledge across 75+ languages.
Here is a great example provided by Google that demonstrates what this looks like:
“Imagine reading a book; if you’re multilingual, you’d be able to share the major takeaways of the book in the other languages you speak — depending on your fluency — because you have an understanding of the book that isn’t language- or translation-dependent. MUM transfers knowledge across languages much like this. Similarly, with its knowledge transfer abilities, MUM doesn’t have to learn a new capability or skill in every new language — it can transfer learnings across them, helping us quickly scale improvements even when there isn’t much training data to work with.”
This means that MUM requires far fewer data inputs than previous models to accomplish the same task.
While its first application focused on identifying vaccines, going forward, Google will be looking at ways to use MUM to dramatically improve Search as a whole.
“Our early testing indicates that not only will MUM be able to improve many aspects of our existing systems, but will also help us create completely new ways to search and explore information,” said Google.
