New Google Technology Set to Transform Search
| August 28th 2023
The world of search technology is changing all the time, and these days it’s literally a case of blink and you’ll miss it. Or miss out!
When we first wrote this article in 2021, we were introducing Google MUM, an at-the-time pioneering artificial intelligence-powered technology designed to answer complex questions that don’t have straight answers. And indeed it still is pioneering, with its multimodal functionality.
Only a few months before, we’d been singing the praises of Google BERT and how it had evolved into the DeepRank algorithm, setting the pace of search to high with its machine learning and natural language processing capabilities.
Whilst AI search technologies have burgeoned since the introduction of MUM, we can still consider it pioneering. So let’s revisit what it’s all about, before delving into the newer kids on the Google AI block.
What is Google MUM and how does it work?
Google Multitask Unified Model (MUM) is a pioneering artificial intelligence-powered technology designed to answer complex questions that don’t have straight answers.
Google MUM is a thousand times more powerful than BERT. And BERT is pretty clever. So what’s behind it?
Google itself has said that it is constantly innovating to make the search engine work better for its users. One issue it has identified is that sometimes, when we’re looking for answers, we have to enter multiple queries and carry out lots of searches to collate the information we need.
Here’s the scenario Google uses to explain the issue…
You’ve hiked Mt. Adams. Next, you fancy tackling Mt. Fuji. You have plenty of questions you need answered, namely what you need to do differently this time to prepare.
Google can of course help you with this. But you’d need to do a few searches. The elevation of each mountain. The average temperature when you’re going. The difficulty of the hiking trails. The right gear to use, and more. So, a fair amount of time and searches later, you get your answer.
But wouldn’t it be dandy if you could speed this process up, and get all the information you need in one search, just like you would if you asked a mountaineering expert the same question?
We have lots of these types of queries in everyday life that need multiple steps with Google. In fact, the search engine says that people use eight separate searches on average for more complex queries.
There’s nothing out there at the moment that’s sophisticated enough to deliver direct, rich answers at the same sort of level that an expert would. And that’s where Google MUM comes in. This is Google getting closer to helping users with more complex search needs. The overall aim is to help us get the information we need, in fewer searches. And we all love a bit of time saving.
Google MUM: The solution for when there’s no straight answer
MUM is built on a Transformer architecture, just like BERT. But as we’ve said, it’s a thousand times more powerful.
Google MUM is engineered not just to understand language, but also generate it. It is trained across an impressive 75 different languages, and has the ability to multitask. This means it can develop a more comprehensive understanding of world knowledge and rich information than BERT and other previous models.
MUM is also multimodal. This means it comprehends information not just across text, but also images, with video and audio to follow in future developments.
Going back to the Mt. Fuji query. MUM would be able to understand, just like a mountaineering expert would, that you are comparing two mountains. It would therefore know to return elevation and trail information. It would also understand, in the context of hiking, that the word ‘prepare’ could mean things such as sourcing the right gear, and fitness training.
Also, because Google MUM has an in-depth knowledge of the world, it could provide useful insights, such as whilst both Mt. Adams and Mt Fuji have similar elevations, if you’re heading to Fuji in autumn, you may need waterproofs because it’s the rainy season. It could even bring up helpful offshoot topics to help you with deeper research, such as top rated gear, best fitness training exercises and the like, presenting you with helpful content from across the web.
Bringing down language barriers
When you’re researching global topics, it could be that the best information is not in your native language. This has traditionally been a considerable barrier to accessing information. But with MUM, there is the potential to bring down these barriers, because it has the power to transfer knowledge across languages. MUM has the ability to learn from sources that are outside your native language, so widening the scope of information that’s available to you.
Say for example there was some hugely useful information about Mt. Fuji, but it was in Japanese. Today, that information probably wouldn’t surface if you carried out your search in English. But MUM could transfer this knowledge into a format you can understand, and it could be rich information you wouldn’t otherwise have had access to.
So much more than just text search
Imagine if you could snap a shot of your hiking books and ask Google, “can I use these to hike Mt. Fuji?” Now that’s clever. And MUM is eventually make that possible with its multimodal functionality.
MUM can understand information from different formats such as websites and images, all at the same time. It could understand your hiking boots image and link it to your question, giving you a helpful yes or no, so you know whether you need to invest in a new pair of boots or not.
What are the latest Google search tools?
Unsurprisingly, much of Google’s latest search technology revolves around artificial intelligence.
Google says that it wants to ‘bring the benefits of AI to everyone’. It is excited about the ‘transformational power of AI’ and the helpful new ways it can be applied, and says that it is committed to ‘responsible innovation and technologies that benefit all of humanity’.
“We want to use AI to augment the abilities of people, to enable us to accomplish more and to allow us to spend more time on our creative endeavours.”
Jeff Dean – Google Senior Fellow
Here’s our round-up of Google’s AI-powered technologies ones to watch, and why…
Google revealed the next generation of its Pathways Language Model (PaLM 2) in May 2023 at Google I/O 2023. It’s a large language model (LLM) that’s a step up from its predecessor PaLM, and that is reckoned to be able to take on its biggest rival, OpenAI’s GPT-4, the technology behind ChatGPT.
PaLM 2 is said to excel at reasoning tasks such as mathematics, code, translation, answering questions and natural language generation.
Immersive View for routes in Google Maps
Since Google Street View was introduced, AI has been used to stitch together billions of panoramic images, allowing people to explore the world from their device. Immersive View is a relatively new technology which uses AI to create a high-fidelity representation of a place, so you get to experience it before you visit.
The tech is also being expanded to help users get to where they want to go. Google Maps provides billions of miles of directions every day. But how about if you could view your entire trip ahead of setting off? Well that’s now possible with Immersive View for routes, whether you’re driving, cycling or walking.
Say you’re in Central London and you want to explore walking routes in some of its green spaces. Hyde Park looks enticing, but so does Greenwich Park. You want to compare them so you can decide which to take, so you click on Immersive View for routes, and you are met with an amazing bird’s eye view of the walk.
You can also tap into all sorts of useful information, such as air quality, traffic and weather, and discover how they may alter during the course of your walk.
To use Immersive View in Google Maps, search for a point of interest, city or route for which Immersive View is available. Tap the icon for the route or landmark, and select the Immersive View card below it.
Google Maps will then load the Immersive View for that particular location, allowing you to interact with it and explore the location. You can zoom in and out, and double-tap to move the map around.
Bard is an artificial intelligence chatbot technology powered by Google’s Language Model for Dialogue Applications (LaMDA).
Much like ChatGPT, Bard is a conversational AI platform, drawing on information from the web to provide fresh, high-quality responses. Also like ChatGPT, Bard has a human-like ability to produce content in seconds.
“Bard can be an outlet for creativity, and a launchpad for curiosity, helping you to explain new discoveries from NASA’s James Webb Space Telescope to a 9-year-old, or learn more about the best strikers in football right now, and then get drills to build your skills,” says Google.
One of the biggest issues with ChatGPT is that, by OpenAI’s own admission, it does not always provide factual information. However, Google Bard has a head start with its established background of using AI to improve the search experience for users.
BERT was able to understand the intricacies of the human language. And then as we know came MUM, which was a thousand times more powerful that BERT with its next level and multilingual understanding, as we explored earlier in this post.
Google Bard allows you to pose questions across a variety of subjects, from history to current events, and from science to literature. You can also use it to generate content, from letters and songs to scripts and code. It’s also a language translation tool.
As with ChatGPT, Bard thrives on feedback. Once you’ve interacted with it, you get the opportunity to give a thumbs up or down. Bard will then use this feedback to refine its responses in the future. Whilst you are encouraged to converse naturally, it’s important to be as precise as possible with your questioning. And if you don’t get the answer you were expecting, or you need more detail, you can go back and ask for an improved response.
Do bear in mind though Google’s caveat that ‘Bard is an experiment and may give inaccurate or inappropriate responses.’
Genesis is an AI tool designed specifically to assist journalists in creating news content. Whilst it is still in the early stages of development, Google has said that it could be used to help journalists with tasks such as writing headlines, generating leads and summarising articles, as well as writing in a particular style.
Google has developed Genesis in partnership with news publishers with the goal of giving journalists a choice of using emerging AI technologies to boost their productivity. Google is very particular about the fact that Genesis is not intended to, nor will it be able to, replace the essential role journalists play in researching, reporting, writing and fact-checking their work.
In summary, Google is continuing to push the boundaries of artificial intelligence and is strongly focused on making its products more helpful and useful for people, businesses, and communities.
Looking for SEO expertise that’s bang on the latest industry trends?
Here at Figment, we know the importance of staying on top of changing search trends. It’s something we’ve been doing since Google first came onto the scene. It’s vital to respond to changes as they happen, and prepare for future shifts, so we can ensure our clients maintain the top ranking positions and online visibility they need to drive sales a and achieve their business growth goals.
Looking for a trusted, top rated SEO agency in London or Surrey? Talk to Figment to learn how we can make a difference to your sales and help you grow your business with SEO expertise, please do get in touch.