On the 18th of May, Google’s Senior Vice President, Mr. Prabhakar Raghavan, presented a new Google facility. It is termed as Multitask Unified Model, abbreviated as MUM. This was done in Google’s Annual Developer Conference called Google I/O.
Google’s Bidirectional Encoder Representation from Transformers, popularly known as the BERT algorithm. MUM’s development is based on transformer architecture.
But as per the claims of Prabhakar Raghavan, multitasking will be the major differentiating factor between the two algorithms.
BERT & MUM
Hence making MUM 1000 times better and user-friendly than BERT. According to him, MUM is developed in such a way that it can handle multiple tasks at the same time.
Another eye-catching feature of MUM is that it is trained across 75 different languages. As a result, it can understand user demands to a much greater extent and benefit the user.
It has another very interesting and valuable feature in it. Rather, MUM can accept queries in several forms such as text and images, even in videos.
Raghavan claimed that certain queries can create problems for the present Search Engine algorithms. As a result, they cannot provide the user the data or update which the user wants. But MUM is highly efficient in overcoming this drawback of present Search Engines.
Click Here, to Know about Google’s May Core Update.
To justify his words, Prabhakar Raghavan presented a query in MUM at the Google I/O. It stated that he has successfully hiked Mount Adams and how should he prepare himself differently for hiking Mount Fuji?
In return to this query, the highly efficient Search Engine MUM provided the differences and similarities of the two mountains in terms of climatic condition, geographic resemblance, and risk factors.
At the same time, it may also provide the user with articles regarding the equipment and gear required for hiking Mount Fuji.
MUM accepts queries in multiple forms. According to Prabhakar Raghavan, if a user provides a photo of his pair of shoes and asks whether it is good for hiking mount Fuji, then MUM can understand the image’s content.
For example, this scenario might provide a list of equipment required to hike and mark the gear you queried for in the list if the gear can hike.
Google is at present only running inside test cases programs with MUM. Yet no official date for the launch of Multitask Unified Model is announced in and out by Google.
For the past few months, Google is incorporating various methods together to make a well-coordinated and compatible search engine. If they succeed in their vision with MUM, it will increase the accessibility of information to a much precise and greater extent.
If Google’s MUM functions the same way like it is presented in Google I/O when released for the general, then people will be able to process queries once assumed to be impossible.
Hence it will deal heavy damage to the usage of other search engines in the market. This was all about MUM, BERT & how better than BERT.