Google MUM to Detect Queries from People In Danger

Google has introduced MUM (Multitask Unified Model), a new AI tool that will help deliver more accurate and precise answers to specific queries.

The company says that Google Multitask Unified Model will be able to automatically and more accurately detect queries from people potentially in danger and try to understand their intentions.

The tool was designed to make the search engine better at handling sensitive queries around personal dangers such as suicide, sexual assault, substance abuse and domestic violence. According to Google, the tool will attempt to understand whether people are in danger while asking certain questions and therefore offer more relevant and precise help.

Currently, people seeking information on sensitive queries such as suicide, drug abuse and domestic violence are likely to come across general but relevant national hotlines which may not necessarily be helpful to their immediate needs.
Google says that MUM will now be used to give a better understanding of the alternative queries that people in danger are likely to key in if and when they don’t type an obvious cry-for-help statement.

How does MUM work?
If, for instance, a user searches “Sydney Suicide hot spots,’’ Google systems would ordinarily give general information such as a list of places regarded as “suicide hot spots.” However, with MUM, the query may be translated to be potentially from someone in crisis and would give more helpful information such as suicide hotlines.

The AI-powered tool is also meant to give more accurate answers to longer search queries. Google’s system hasn’t been very articulate in answering long natural queries asked in a human tone, but MUM could turn this around. This means, for instance, if a person types “why did he attack me when I said I don’t love him,” the search engine could treat this as a case of potential domestic violence and show therefore display relevant and actionable results.

MUM has been trained in 75 languages, making it more helpful to a wide user case. This will also help the machine-learning AI-powered tool to display information from trusted partners such as local hotlines.

Previous articleXiaomi launches 5G editions of the Redmi Note 11 and Redmi 10
Next articleNigeria approves implementation of NIN-SIM policy
Avatar of Gbenga Ajao
Gbenga Ajao is a journalist. He has garnered experience in print, electronic and new media. He is passionate about mobile and web technology.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.