The architecture, Pathways, could have a significant impact on how businesses can be found on Google Search.
The company began working on Pathways(PaLM) about two years ago, when it realized that many of its complex AI algorithms could be simplified. The final product was released in March of this year, and according to Google, it is making search easier and more accurate. Pathways is a neural network architecture that allows AI models to perform multiple tasks independently. Instead, the Pathways approach allows models to be trained and optimized for a specific task
Pathways could enable multimodal models that simultaneously span visual, auditory, and language comprehension. So whether the model is processing the word “elephant,” the sound of someone saying “elephant,” or a video of an elephant bathing, the same response is triggered internally: the concept of an elephant. The result is a model that is more insightful and less prone to error and bias. It might even infer the cultural context of the concept, referencing “Dumbo,” for example.
Pathways could also have a profound impact on Google Search, which is powered by artificial intelligence. Pathways could make Google Search understand multiple data sources on any topic people search for and yield a more complete set of results. In addition, Pathways will likely make Google Search more useful and accurate for the different ways people search: text queries, voice search, and visual search. Google was already heading in this direction with the use of Multitask Unified Model (MUM) technology, which makes it possible for Google to provide more complete answers to a variety of queries and search types (including visual search). Although we’re still in the early days, Pathways looks like a more powerful way for Google Search to take advantage of technologies like MUM.
With Pathways, Google Search will be able to search deeper into a site’s content to find answers to queries, and this includes images, videos, text and auditory content. It also means that Schema markup will be more valuable. Schema is a markup system and a form of code that webmasters place on their websites (via HTML) to make it easier for search engines to understand their content. A site that uses Schema (also known as Schema.org) can improve the way search engines read and represent your content on search engine results pages (SERPs) with rich snippets.
Schema is also a critical component of voice search. Because Pathways is powered by AI, having more robust support for voice search could result in branded content gaining visibility for a wider range of searches. Seth Nickerson also noted that because Google Pathways implements machine learning, search results theoretically should improve over time.
Google Pathways will also make visual content more important as MUM has already done. Being able to analyze a broader set of images and videos (multitasking better with Pathways) will help Google Search display search results that contain visual content, including, of course, content on a brand’s website.
And of course, Pathways need not be restricted to SEO, it could handle more abstract forms of data, helping to find useful patterns that have eluded humans in complex systems: such as climate dynamics, helping to design better drugs by infusing these models with knowledge of chemistry and physics, or contributing to advancement in educational systems.