` – Google is about to launch a new search engine. A technology that, thanks to artificial intelligence, will be able to give precise answers to increasingly complex questions. For Mountain View ‘Mum’, this is the name of the software, an acronym that stands for ‘multitask unified model’, is destined to radically change the way we are used to using the search engine, because this will be able to better understand the wishes of users, returning more and more precise results.


Mum, which in Italian we can render as a ‘unified model of operations carried out simultaneously’, is designed to behave a bit like the human brain: it learns from the type of research we do, from the words we use, tries to understand our ‘deep intentions’ ‘to return results that are increasingly consistent with our desires. But, like any innovation, it will bring benefits that hide a dark side. And for critics it is that it could become a way to shut down more and more users in the Google galaxy. Much more than what is already happening today.

“We think we are close to a big new milestone,” Pandu Nayak, head of MUM, told the Financial Times. Who says he is sure of one thing: this technology will answer “thousands of confused and ambiguous questions that people ask themselves every day”, but which at the moment they have not formulated in a direct, clear and distinct way.


An example? What kind of school is best for their children, he explains, or how to identify the feeling of having to start a new diet, or training: “Someone might be wondering what is the best fitness regimen for my level of fitness and age” , explains Nayak. Currently, to find answers to such precise questions “it is necessary to ask the search engine several questions, and try to cross-reference the information received to get the answer you want from Google”, the manager reasoned. But the future of ‘queries’ with Mum is destined for answers to needs that are “more and more complex to formulate, and more and more realistic”.


Mum is the result of a decade of development of knoledge graph (knowledge graph), a function introduced by Google in 2012 on the search engine, which allows you to define the relationship between different concepts, different objects, giving algorithms a certain degree of semantic understanding of research, in other words ‘humanizing’ its understanding.

Semantic understanding that has been gradually improved, until it is included in another project of the Silicon Valley giant: RankBrain, a machine learning program launched in 2015 designed to improve search engine results.

The evolution of this technology first led to the launch of Bert (Bidirectional encoder representations for transormer), which arrived in Italy in 2019, which essentially allowed the algorithm to ‘guess’ what a user was looking for even when the search engine was formulate whole sentences, or searches for inexact words. Mum, presented last May during the annual developer conference, is its natural evolution. But about 1,000 times more powerful.


Text search is just the beginning. Mum, explains Google, will be able to cross searches involving text, audio, photos and videos: a completely different way of understanding online research. Objective: to find a rationality in the indistinct desires, a logic in the vague.

Mum will put our desires on Cartesian axes, and make them consistent in our eyes. With enormous potential also on the advertising market, the gasoline that powers the whole engine of Mountain View.

And whoever supplies it will ultimately determine the success or otherwise of this technology: some doubt that more precise and direct searches could lead to a decrease in searches themselves, therefore a decrease in advertising. But for Google there is no doubt: more precise searches mean more targeted advertising. History has so far proved him right.


By Editor

Leave a Reply