7 Tech Trends in AI and Search for 2022

7 Tech Trends in AI and Search for 2022

The evolution of “smarter search” continues to dominate search applications and search engine technologies. In the last 3 years they have gone from hype to a number of practical applications of AI and Search. We expect this trend to continue in 2022 with a continued better understanding by the market of how to deploy the most practical elements of AI in search, even while some of the more far-reaching claims or predictions remain too immature or impractical to deploy. Here are the top tech trends in AI and Search that you should pay attention to in 2022.

Tech Trends in AI and Search 2022

1. Question Answering Systems

Many predictions of the evolution of search applications focus on machine learning to personalize results and anticipate user needs, especially in ecommerce search. In enterprise search and broader search applications, however, we believe turning intranet and other application search bars into Question Answering Systems presents a lower, much more achievable goal. The internet search experience really frames users’ most basic expectations.

2. Spread of Knowledge Graphs in Search

Search using knowledge graphs used to live only in the academic domain, or at some knowledge graph tech shows. This year, KMWorld featured more sessions than ever – six – that discussed knowledge graph and search applications. The technology is mature enough that we expect more and more search applications to leverage knowledge graphs in their architecture. By the way, this goes hand in hand with the emergence of more Question Answering Systems.

3. No HAL 2000 … Yet

As more and more people type full natural language questions into search bars or ask complex questions of their digital assistants like Alexa, Siri and Google Assistant, the expectations will increase on the complexity of the responses. In the near term, however, expect results to be limited to predictable answers derived from knowledge graphs, fact extraction from text, and keyword search results. Anything that seems “smarter” than that is just engineers and their Turing Test tricks to humanize digital assistants. We are still a long way off from wide deployment of true “thinking” AI that can deduce human intent like HAL locking out astronaut Dave Bowman in 2001: A Space Odyssey for fear that Bowman was about to shut him down.

4. Better BERTs

BERT paved the way for using machine learning models to improve natural language processing, thanks in part to Google’s support and an eager developer community. But BERT is now several years old and there are now 40 or 50 variants that have emerged (GPT3, RoBERTa, AlBERT, etc.), either as complementary or alternative tools. While this may be daunting for developers new to NLP and search, we think that this means exciting improvements in speed, accuracy and multi-language capabilities for NLP and search-based applications. The evolution of these advanced NLP tools remains one of the more enduring tech trends in AI and Search, and Pureinsights can certainly help you navigate the sea of choices.

5. OpenSearch is Legit

Our blog has already covered a lot about Elasticsearch vs OpenSearch. We still believe Elasticsearch and ELK are the best open-source (ish) search engine and analytics stacks out there. But as long as it OpenSearch has the support of AWS, and because its roots well established with Elasticsearch, we think you will see continued a broader deployment footprint for OpenSearch in 2022. Expect more news like this from AWS and Roche.

6. The Magnitude of a Vector Product

As far as emerging technologies go Vector Search will have the most impact. But it will be a quiet disruption as the technology infuses into numerous applications. With Vector Search, a text query (or even an image) is encoded into a vector and then a nearest-neighbour search performed to find the closest query related vectors in a space. It enhances peoples’ search experience making it a lot more natural and intuitive because you search by concept rather than keywords. It also enables question and answer type functionality. Today, Google Search, YouTube Recommendations and Spotify Discovery are already using vectors.

If you want to explore further check-out Pinecone who offer “a fully managed vector database that makes it easy to add semantic search to production applications.” Also Weaviate and Milvus an open-source vector database. If you are new to Vector Search, this article in InfoWorld is a nice, albeit somewhat technical summary.

7. Change Is the Only Constant

Google has identified three changes to the way people are interacting with web search. They talk about the notion of “information” expanding from documents to richer formats including images and video. That people are increasingly using mobile devices to search and that they are looking for direct answers to their questions. Microsoft envisages other challenges for workplace search such as providing a universal search experience, summarization and interaction with search results, and conversational information retrieval.

I don’t think that there is any doubt that we will see advancements from these organisations, and other search vendors, in the use of AI technologies to meet these challenges including graph neural networks, Natural Language Processing and Machine Learning algorithms such as BERT and GPT.

Here are some links that give you a sneak peek at what Google, Microsoft, and AWS are working on in their research labs here:

I hope you enjoyed my attempt at practical prognostication of key tech trends in AI and Search for 2022. The bottom line is that AI will continue to be pragmatically deployed to enhance search applications and it’s a more exciting time than ever to be working on search. If you have any questions drop, contact us or drop me a note at info@pureinsights.com.

Best regards and Happy Holidays to all.
– Phil

Share on twitter
Share on linkedin

Stay up to date with our latest insights!