Generative AI search is the next iteration for online search as we know it from Yahoo.com, Google.com, Bing.com, and more.
Artificial intelligence (AI) has influenced our lives since the launch of AI personal assistants on mobile devices, like Cortana, Siri, Google Assistant, Alexa, etc. Using AI on Internet of Things (IoT) devices; you can control many things at home and the workplace.
Next, AI entered the multimedia production industry to produce content like images, video, audio, and text from keywords or instructions. These days, advanced AIs can edit content like video and audio to perfection. In a nutshell, AI is everywhere. Hence, that day is not far away when you will see AI influencing content searching from the world wide web.
What Is Generative AI Search?
To understand the generative AI search concept, you must first understand what a generative AI is. Such an AI simply produces content like text, images, audio, program code, etc., from sample content.
Developers train a generative AI to understand natural content in machine language using machine learning (ML) models. Such AIs could be supervised, unsupervised, or semi-supervised.
Generative AI Models
Generative AI uses various ML models to train an AI program, chatbot, or virtual assistant. Some of these models and their outcomes are outlined below:
Generative Vs. Discriminative Model
In the discriminative model for AI training, a human supervisor trains the AI to learn the differences between two or more objects in an input sample. For example, if you supply 10 images of 10 different animals to the AI as input, the underlying discriminative model will help it to distinguish all the animals successfully.
On the other hand, the generative model helps an AI to create objects referring to sample data with semi-supervision or no supervision. A generative machine learning model helps an AI understand the input data and preserve the understanding in its neural network memory so that it can call this experience in the future if a similar challenge arises.
Generative Adversarial Networks (GANs)
This machine learning algorithm combines both generative and discriminative models for AI training. Here, the generative model creates samples from the input vectors like keywords, questions, etc.
Then, the discriminative model must identify if the created sample is fake or original input. If it is fake, then the generative model reworks on the task to create another output for the discriminative model. It goes on in iterations until the generative model can create fakes that the discriminative model can not distinguish anymore from the original input.
Transformer-Based Models
Transformer models for ML are deep neural networks that analyze the input vectors sequence by sequence. Then it predicts what the output could be. For example, if you present a transformer with a series of unrelated words, it will analyze the words and try to predict preceding or successive words that could fill in the blanks and convert unrelated words into meaningful sentences.
In a transformer, an encoder extracts all the features or data points from the input sequence and converts them into input vectors. Then, the decoder analyzes the input vectors, creates a context from the data, and produces an output sequence.
There are many successful transformer-based AI models like the ones outlined below:
- Generative pre-trained transformer model 3, aka ChatGPT
- Language Model for Dialogue Applications, aka LaMDA, built on Google Transformer
Using all the above models, AI developers have successfully created many functional generative AI programs that can do or produce the following from simple inputs like images, texts, descriptions, audio, etc.:
- Generating images of human beings that do not exist by referencing inputs from websites, magazines, Google Image Search, etc.
- Generate real images from sketches
- Transfer artistic or creative style from one art to another
- Synthesizing a CT scan from an MRI as an input
- OpenAI’s Dall-e AI can create outstanding images just from texts
- DeepMind, Amazon Polly, etc., AIs can create human speeches from texts
- AI Music, acquired by Apple, can transform copyright-free public music into soundtracks
Now, generative AI search is a combination of all these tools and technologies of AI to present you with accurate content from the web. With such an AI-powered search capability, you do not need to search through millions of suggestions created by Google, Bing, Yahoo, etc., search engines.
The generative AI search will present you ready to publish or consume content from online sources with various supported content like images, videos, and text on one screen, just like the ChatGPT works.
How Is Generative AI Search Different From Regular Online Search?
Website search, as we know it from the launch of the Archie search engine back on 10 September 1990, will change completely if generative AI search becomes popular and easily accessible.
The regular world wide web search is a manual online research process. Here, you need to type your question or keyword into a search box of the search engine. Search engine providers like Google, Yahoo, Bing, etc., weigh the resulting websites containing the content you want according to some proprietary logic.
For example, the website’s authority on the relevant niche, readership, website page quality, etc. Then, the search engine assigns ranks to each website against your search and shows all websites according to their rank. For instance, a website with rank one will show up on the top of the search engine result pages.
In a nutshell, regular online search engines do not create content. They just syndicate content from several websites. When you click on a search result, you go to a specific website directly.
However, with the launch of generative search, you will get limited content. The underlying AI will analyze all the search results, generate custom content, and show that to you using a web browser. There could be links to the sources the generative AI used to craft the content it shows you.
If generative AI search becomes the new normal of online search, then you can expect the following additional differences:
- The output content from your search query will depend a lot on the polarization of the company that created the generative AI search model
- Some schools of thought will prefer XYZ generative AI search tool over the ABC tool. Thus there will be rising anomaly in online search-based works
- Such search tools may sometimes come up with similar content, and the publisher may risk uploading plagiarized content on their website
- The search result will become intuitive and filled with related content in different forms like texts, images, video, audio, and so on
- You will stop visiting websites and interacting with web ads if you get content in a ChatGPT-like interface where there are no distractions
- Your online research effort will decrease drastically. You no longer need to read through several web pages and compose your own content
- AI developers will come up with new AI-based online advertising and other revenue models to increase their operational profits
- There will be less distraction in web search, so the quality of search may decrease a lot
- You must employ talented and expert online research professionals and data analysts to analyze the AI-produced content before you can use the content for commercial purposes
- There is no clear guideline on how such an AI-based website search will be linked back to the source website and give those websites some credit. Because these AIs can not generate content without relying on reference content
Next, we will explore the impact of generative AI search on search engines.
Impacts of Generative AI Search on Search Engines
Find below how the progress of generative AI search can impact conventional search engines:
- The popularity of the search giants like Google, Yahoo, DuckDuckGo, and Brave will go down a lot
- The ad revenue collected from search engines will also decrease substantially
- The free and fair web search results will be impacted, and a new revenue stream will spring up where website owners will pay the generative AI search providers to show content from their pages
- Footfall on websites will decrease a lot because the users will get the content they need on a different web page
Now, we will explore some of the search engines that use generative AI search.
Search Engines That Use Generative AI Search
Search engine companies already know that AI-based generative search is the future. Hence, various search giants have started rolling out prototypes and beta testing for AI search engines. Find below some AI-based search engines that you can use now:
#1. Bing
Microsoft did not just stop by acquiring the ChatGPT developer OpenAI. It used the proprietary technologies and licenses from OpenAI to enhance Bing Search with AI capabilities. The revamped product is popularly known as the new Bing.
The search engine helps you with complete answers to real questions—no more keyword-based ranking of websites and manual scrubbing of data from top-ranking websites. You can also chat with the search engine, just like texting and messaging with an expert in the domain where you work or do business.
The search engine chat will allow you to ask up to five follow-up questions to fine-tune the search result that the underlying generative AI model provides. The new Bing is not just to get online web results. It can also help you with the following:
- Get tips on various topics and niche domains
- Produce creative content with the help of a generative AI like ChatGPT
- Get intuitive and accurate search results so you can get to your work quickly without any distractions from ads and call-to-action pop-ups
#2. Google
Google Search has been leveraging AI search tools for many years. RankBrain is the first AI tool Google used in 2015 to rank websites. This AI interprets the search results and ranks relevant websites at the top of the ranking hierarchy.
Other AI programs that Google uses in its search engine are as follows:
- Neural matching helps search engine figure out how queries are relevant to pages
- Bidirectional Encoder Representations from Transformers or BERT for natural language processing pre-training
- Google Lens for object searching using a mobile phone or tablet camera
- Multitask Unified Model or MUM for COVID-19 vaccine information on web results
#3. You
You is a production-ready AI search engine tool. Users can use it to get rich search results as outlined below:
- It shows a count of apps, tools, and results at the top of the SERP
- The People Also Ask results pop up in the right-side panel
- Get a suggestion for YouChat
- It shows top discussion cards from high-authority social media like Reddit
- Users can also add more queries to the same search
Currently, it offers the following AI Search products:
- YouChat
- YouCode
- YouWrite
Author’s Note
Generative AI-based search engines can become a problem for online research. As soon as these companies start planning revenue generation by getting payments to promote content from certain websites or publishers, the online search will become highly biased.
The AI search engine developers must form a consortium to draft an ethical practice to ensure free and fair search practices.
So far, you went through a detailed discussion on the definition of generative AI search, its difference from the conventional website search concept, and its impact. Also, you explored novel examples of generative AI search tools that can get you outstanding content in minimum time spent on website research.
This article helps you to understand whether you should go with an AI-based search or not. However, generative machine learning-based AI search of the world wide web shall become the new trend.
Next, check out Artificial Narrow Intelligence (ANI).