Imagine you are looking for answers to a marketing problem you are facing at your company. You search the answers on Google to get some ideas from other people who also faced the same problem and have written about it.
After reading multiple articles on the topic, you are starting to get some ideas on what you can do to solve the issues you are facing. Let’s say this takes you 30-40 minutes to search and discover the ideas.
As per Google, it takes on an average of 8 queries to get the answers to complex questions like these ones.
Now imagine the same situation but when you search on Google, it gives you the ideas on a platter without searching in-depth with multiple queries to get the ideas you are looking for.
This is what Google MUM will do.
It will change the way we search for information.
It will change the way we search for ideas, answers, advice..
What is Google MUM?
Google Multitask Unified Model (MUM) is a new technology for answering complex questions that don’t have direct answers.
MUM is a language model built on the same transformer system as BERT.
BERT is a powerful language model released by Google in 2019. However, MUM, as per Google, is 1000X more powerful than BERT.
What makes MUM unique is its multi-tasking capabilities. With this AI-driven approach, it can read text, comprehend it, drive insights about the topic, associate it with multiple content formats like video, images, audio, all at the same time in 75 languages to answer complex questions.
WOAH..Isn’t it something?
It is like talking to a domain expert who has over the years built her knowledge and formed expert insights on a topic.
To understand more about its capability here’s a quick preview of the Google I/O 2021 presentation introducing MUM. Take a look
When will Google add MUM to search results?
Currently, Google is testing the technology and will continue to do so till it is safe to be implemented.
Why is not safe?
As with any AI-driven approach, the challenge is to make the systems bias-proof. Google MUM uses the information it finds on the web and processes it to form insights to answers complex questions.
AI doesn’t understand if the information is false or harmful to someone. It can use information that is racist, or hateful and make that information accessible.
Since there is no way it can use judgement, there are ethical risks involved.
Google does recognize these risks and working to remove the bias.
Will it reduce website traffic for content creators who write insightful articles?
This is another concern about AI-driven search answers where user gets the answers without clicking on the websites that originally published the content which Google MUM uses to serve the answers to complex queries.
As per estimates, almost 65% of the searches produced no clicks in 2020. Source: SearchEngineLand
This is HUGE!
Millions of websites survive on ad revenue coming from website visits that search engines like Google drive to top-ranking websites based on search queries.
If Google starts answering complex queries directly without requiring the user to click on the websites, it raises two questions:
- Who owns the content that Google serves to the users using MUM?
- Why will content creators continue to write when they know website traffic will be negligible?
These are important questions that Google will struggle to answer.
Maybe, Google will need to think about revenue-sharing agreements with publishers. But that is far more complex and difficult to implement considering that it is not easy to prove that Google used a particular website to answer queries.
We can only wait and watch to see what exactly will happen once Google MUM is implemented.
Future of Search
The introduction of Google MUM simply means Google is looking to make search faster and answer complex queries without the user typing multiple search queries.
This AI-powered approach to handle complex queries is good for the users but it raises some deeper questions about content ownership, revenue sharing, and also ethical questions like bias in search due to AI using hateful information to form insights around some queries.
Nevertheless, the future of search looks exciting and less effort intensive.