Where is Google’s AI Search Engine? How Will it Compare to Perplexity AI?
Google has been one the largest companies to be criticised for a lack innovation in AI by many. I think critics are mainly shareholders watching Nvidia skyrocket off the AI bounce expecting the same for Google.
There has been talk about adding generative AI to their search engine but as you have probably noticed, nothing has materialised which raises further questions.
In this article, we attempt to answer the question of ‘where is it?’. In addition, we hope to shed light on the different ways they could use generative AI to add additional functionality to their search engine to entice users away from competitors like Perplexity AI.
Note: I expect this article to become outdated pretty quickly especially upon the release of the search engine with AI capabilities.
What’s Going On With Google?
Google I/O Event
At the 2024 keynote, Google announced they would be bringing AI overviews to their search engine allowing users to obtain contextual summaries of their queries. This was the most logical implementation of generative AI and supposedly the easiest to incorporate but it has seen limited use.
According to Search Engine Land, the overviews have had a staggered release and only 15% of queries provide the overview. It was apparently 84% but the tool has produced some incorrect or harmful summaries causing its appearance in their main search engine to be cautiously reduced.
Benefits of taking a bath with a toaster
The nonsensical answer to the silly question above is an example of the overviews producing content that is harmful to readers.
This was met with frustration and disappointment from the media as it raised questions about Googles competence when it comes to working with the latest AI technologies. The same technologies that Googles own AI researchers had a massive role in developing in the first place which is probably the most strange realisation from this all.
It was especially disappointing given the promising dispersal of AI announcements of projects we received at their I/O event earlier this year. While Google will likely get it right in the end and in some time it will likely be the best AI search engine out there, its not a good look for a company of their size meant to be in a prime position in this AI revolution.
Right now, it seems like it’s a multifaceted issue not necessarily related to skill or competence but rather to the fact Google is absolutely massive!
Massive Infrastructure with Little Wiggle Room
Google’s reach is massive, encompassing various services used by billions globally:
Despite their recent AI search engine complications, Google still holds a 90% share of the search engine market according to Statista. Processing a staggering 2 trillion searches every year.
YouTube, the biggest video media platform in the world, has over 2 billion active monthly users. Only Facebook has more.
Their advertising business is a direct result of both their search engine dominance and massive digital influence through YouTube. Their advertising revenue was recorded to be $237.86 billion in 2023.
Google maps is probably one of their largest and most impactful offering. It has 1 billion active users. Keeping Google Maps running is feat in itself and it requires a combination of large data centres, robust network infrastructure, AI and machine learning to optimise routes just to name a few things.
If you didn’t know already, Google is huge and their attention is spread out so it’s understandable that they would be slower and less effective than the likes of OpenAI or Perplexity.
However, they should have the resources to add generative AI summaries to their search engine. Surely?
How Do AI Search Engines Work?
AI search engines can work in multiple ways such as providing summaries like Google attempted or even just aggregating search results so there is no one way to describe how their work.
Perplexity AI
Perplexity AI
Using Perplexity as an example, a user is able to enter a query and obtain several search results relating to the query. The only difference to regular search engines is the contextual search capabilities AI can provide.
Prompt engineering can be used to create alternative queries closely resembling and relating to your initial query. Perplexity AI then aggregates all search results pertaining to the list of potential queries.
In addition, Perplexity AI still has the ability to generate text, articles, poems and more similar to what ChatGPT does. It’s effectively an LLM with access to the internet.
Perplexity’s biggest achievement is being able to accurately source search results contextually relevant and factually correct for queries before producing informative responses using an LLM. It’s likely they use a GPT model for text generation but it’s a remarkable achievement anyways and the team of software engineers should be proud because Google couldn’t get it right!
Perplexity is just one example of a company trying to make a dent in Googles search engine dominance. The “smart search” space is only getting bigger and natural language processing advancements are directly driving these improvements.
At Asycd, we are also interested in developing smarter ways to obtain and process information on the web. Our latest tool, the TEV1 AI image generator also leverages web scraping and an LLM to transform image prompts into supremely detailed artwork descriptions.
What We Think?
Google has the potential to have a significant impact on the AI revolution and in fact they already have by developing crucial algorithms and deep learning techniques present in AI tools right now.
BERT – A ground-breaking NLP model able to bi-directionally understand nuances and textual features of a text input. It’s been a revolutionary model helping with tasks such as sentiment analysis or named-entity recognition.
TensorFlow is an open-source machine learning framework that has become synonymous with AI development. It allows researchers and developers to create complex machine learning models with ease, accelerating the pace of innovation in AI. A staple in most data scientists toolkit today.
In computer vision and image recognition, Google has developed algorithms that power image search and enable features like Google Photos’ ability to categorize and search for images based on content. Their work in this area has also contributed to advancements in autonomous driving technology and medical diagnostics.
Google’s DeepMind subsidiary has achieved remarkable feats, such as developing AlphaGo, the first computer program to defeat a world champion in the complex game of Go.
They’ve already done so much in AI so the setback occurring in their search engines should be looked past in our opinion.
It would be sensible for them to continue iterating over their search engine but also exploring different ways they could utilise AI in some of their other platforms. For instance, adding smart voice assistants to Google Maps would be maybe be a good use of time?
Maybe adding semantic search capabilities to YouTube to replace the keyword focused search algorithm they currently employ?
Google definitely has some room to pivot which is nothing new!
