Google announced MUM in 2021, and its search level now allows you to identify emotional crises using AI.
Google understands that this pandemic process has caused the emotional health of the planet to deal with various situations.
It is not uncommon to see reactions on social networks, text messages in groups or ephemeral publications that, in some way, show the lack of support in cases of anxiety and depression. Given this, which becomes critical, the Mountain View company has proposed an option that seeks to assist these cadres from searches using Artificial Intelligence.
With the development of MUM (Unified Multitasking Model), the new AI model that complements BERT – the natural language interpretation AI system -, Google has a tool that is up to 1000 times better for interpreting searches carried out by the user on topics of emotional crisis, anxiety, depression and suicide.
“Now, we can automatically and more accurately detect a broader range of searches about personal crises,” Google added in a statement. “MUM is able to better understand the intent behind people’s questions to detect when a person is in need. This helps us more reliably display reliable and actionable information at the right time.”
Google and assistance in cases of anxiety and depression
“One of the critical components to displaying the right information is at the right time. That shows our ability to understand people’s queries and the meaning of their intentions. Let’s remember that human language is complex and nuanced ”, mentions Anne Merrit, Product Manager at Google and in charge of adding resources on mental health in the search engine. “This is where our AI can be incredibly helpful, as our older systems were limited in understanding the ways people search for personal information.”
In the case of MUM, it is able to recognize up to 75 different languages and determine the peculiarities that develop in topics of sexual assault, domestic violence, substance abuse and crisis situations.
“People in crisis need actionable information immediately. In many of these situations, they try to find the correct information. Speed is essential because the window of time they have to seek help can be limited in a significant way, “ says Merritt,
SafeSearch: Google’s safe browsing
Along with this implementation, the defense mechanisms against shocking or unexpected content continue to improve their abilities to defend the navigation of minors, even when the options are disabled. Google will now include its BERT system in this section.
We spoke with Tulsee Doshi, head of human-centric AI at Google, who noted that Google is training its BERT system to identify explicit queries and remove them from results. “With BERT, we can more reliably downgrade graphic content by up to 30%. This, in addition, has had a positive impact on the reduction of explicit content associated with ethnic origin, sexual orientation and gender”.
The arrival of BERT in SafeSeaerch implies a more complete way of analyzing results in real time, due to the vast understanding of the language from the AI engine and its multitasking capacity.
“Many of these models can only perform one query classification at a time, but with MUM’s multitasking we can do this in 75 different languages and scale globally,” says Doshi, linking this potential to MUM’s language understanding capabilities. BERT, the engine can act quickly to dismiss results with explicit or unexpected content.
Google told NIUSGEEK that it hopes to see this personal crisis assistance model replicated to other situations on the Internet, such as empathy and bullying in children and adolescents. “Both issues are incredibly important. We hope to implement those improvements in the coming months; And as we continue to grow and expand our impact in this area, we want to think about a lot of these different topics . ”