The conventional approach to internet search has changed forever due to the recent democratization of access to language models. So there is a high possibility that for many internet users googling will soon become binging, youing, or perplexiting. If none of these verbs ring a bell, don’t worry, it’s true that most eyes are on two tech giants, Microsoft and Google. However, when digging just a little deeper, you can find a range of equally intriguing and innovative AI chatbot startups that play important roles in shaping the future of how we interact with technology and get information. Their research teams and machine learning experts are busy creating new-age AI-powered search engines that offer much more personalized solutions than traditional search engines.
The Recap of Search Technology Breakthrough
At the end of the year, OpenAI took over the news by announcing public access to its chatbot that can answer questions in full sentences instead of a list of links that may or may not be relevant to your query. Essentially, with the release of ChatGPT, the stereotype of an omniscient robot that has an answer for every question became true. Being a long-standing OpenAI investor, Microsoft managed to quickly capitalize on the buzz and adopted OpenAI’s language model for Bing. In February, Google revealed its own chatbot Bard to a group of trusted testers, which will also make the headlines the minute it’s released to the public.
Search Tools Beyond Google and Microsoft
While Microsoft and Google will most likely remain the promoters of AI-powered chatbots, we must consider other lesser-known industry players to get the full picture. The thing is that other companies can also access the GPT-3 language model, the core technology behind ChatGPT, through the OpenAI API. In other words, they also now have a shot at coming up with their unique conversational AI service without spending millions of dollars on training machine learning models.
Now let’s see what other companies do differently.
You.com
You.com was launched in 2020 and began offering AI chatbot in December 2022, shortly after the ChatGPT release. A distinct advantage of You.com is that it has separate engines for copywriting, coding, generating images, chatting, and summarizing web pages. A few months ago, You.com also announced multimodal search, which allows a chatbot to answer questions using images or embedded apps.
For example, when asked a question about stock prices, the chatbot will provide a snippet of You.com’s stock widget with a short description instead of just raw numbers. On top of that, You.com’s brand is built around privacy, which can be appealing to increasingly more aware internet users. The search engine doesn’t capture cookies, provide ads, or keep personal data.
Metaphor
Nearly all AI-based chatbots tailored for internet search use OpenAI’s large language model to predict the next likely word in a sentence based on text input. However, what if the engine completed prompts with links instead of words? This is exactly what metaphor. systems platform by an innovative startup called Metaphor does. The idea may sound strange at first sight, but Metaphor’s approach is an exceptionally effective alternative to conventional Google search. To get a list of links for your prompt, you need to write a sentence that can be ended with a link. So, instead of writing “knitting tutorials” in Google, you would write “here is the best knitting tutorial” in Metaphor.
The advantage of Metaphor is that it prioritizes links based on their likelihood to complete the sentence. This means that when a prompt is worded even slightly differently, it will produce different results. This allows Metaphor to dig deeper into the internet, helping users to discover content that is truly relevant to their query but otherwise buried on the 70th page of Google search.
Perplexity
Perplexity was launched by former employees from OpenAI, Meta, Nvidia, and Quora. The platform provides answers to prompts based on the current data and supports them with sources after almost every sentence. ChatGPT and Jasper, on the other hand, have been trained on search data up to somewhere between 2020 and 2022. However, this peculiarity can come as a trade-off for information accuracy.
One of Perplexity’s standout features is its Chrome extension which allows users to summarize a page they’re currently viewing or answer queries based on the data from a specific website or page. This capability can be exceptionally valuable for users that need to get detailed information about certain subjects in a specific domain. For example, instead of using the built-in Reddit search engine to find threads and comments of interest, analyze them, and make conclusions, users can order Perplexity to scour for info across specific Reddit communities and provide answers in an easily-digestible format.
An even playfield
There is no doubt that with the rapid advancements in language models and deep learning in general, our approach to acquiring knowledge, generating content, and getting answers to simple questions will continuously transform. Loosely speaking, with OpenAI API available to tech startups, the competition among search engines comes down to finding the right approach to sourcing and presenting information rather than figuring out the right technology to do so. The playfield has been somewhat evened out, and the challenge now lies in offering an intuitive, transparent, and meaningful experience to users.
Tune in to MTC Podcast for visionary Martech Trends.
ABOUT THE AUTHOR
Andrey Koptelov, Innovation Analyst at Itransition
Andrey Koptelov is an Innovation Analyst at Itransition software development company headquartered in Denver. With a profound experience in IT, he writes about new disruptive technologies and innovations.