It is tough to convince users to switch search engines. This is one of the reasons why public search startups rarely succeed. Another is that indexing huge numbers of websites is pricey (Google has a so-called estimated tens of billions of pages indexed), but one company backed by Y Combinator, And Iis undeterred – he continues to work on creating an AI assistant that will provide answers instead of links when searching online.
Andi was founded by Angela Hoover, who signed up for YC’s Startup school after dropping out of college and making it to YC’s 2022 winter class. After working abroad in construction and at Microsoft as a data center project administrator, Hoover met Andi’s co-founder Jed White at the Denver airport after she returned to the U.S.
Hoover and White – who have backgrounds in artificial intelligence and search, particularly content quality ranking, querying and classification – talked about how bad online search has become for travel and what it would take to build a fresh type of search engine from scratch .
“Gen Z hates Google. For us, search is broken. We are on our phones using messaging apps with visual feeds like TikTok and Instagram,” Hoover told TechCrunch in an email interview. It’s anyone’s guess – Google executives admitted this. “I keep hearing my friends say Google sucks. Search results are filled with ads, SEO spam, and clutter. Gen Z is so desperate for an alternative that we use TikTok as a search engine. We hate invasive, scary ads and how Google is Big Brother and monitors everything.”
Hoover offers the artificial intelligence-based Andi assistant as an alternative. The general-purpose system tries to find and extract answers to questions by combining gigantic language models, similar to OpenAI’s GPT-3, with live internet data.
Behind the scenes, Andi extracts information from Internet search results, ranked by relevance to the question being asked, as well as overall quality (though it’s unclear how Andi defines “quality”). Depending on the topic, the platform uses various artificial intelligence systems tailored to specific industries (e.g. factual knowledge, programming or consumer health) and language models that generate answers by combining knowledge from many sources (e.g. Wolfram Alpha, Forbes, The Fresh York Times , etc.).
This is a step further than Google highlighted fragments, which pull text from websites to answer frequently asked questions, and are closer to so-called cognitive search engines like Amazon Kendra and Microsoft SharePoint Syntex, which draw from knowledge bases to combine answers. Startups like Hebbia, Kagi and You.com also utilize artificial intelligence to return specific content from the Internet in response to queries, rather than basic lists of results.
So what sets Andi apart? Unlike some competitors, Hoover says it does not charge for its services or log personal information. Andi does not record or store searches or the results that people read or click on, and only uses approximate location data to improve the relevance of search results.
“Even if we add the option of user accounts in the future, we will only collect and store data that will help our customers use the service effectively when they want to create an account or be remembered across devices and sessions, and to improve the service we provide,” Hoover said. “Users tell us that Andi can save them 15 or 20 minutes of searching and are asking us to let them use it with their team and personal data… As question answering technology improves and we add support for connecting to private data sources, We think it has huge potential.”
To filter out information that may be misleading – or clearly false – Hoover says Andi uses techniques including blocklists and ranking metrics. Disinformation is, of course, an evolving problem – just like Google’s fought z. Hoover, however, expressed confidence in the technical steps taken by Andi to mitigate the effects.
“Every new search startup creates another weaker copy of Google with the same cluttered blue-link page facing the web browser, with more or less differences in advertising and privacy practices,” she said. “The content you see [Andi’s] results are pulled from a live feed when possible, not from a stale index. Answering questions is improving rapidly and is already excellent in many areas.”
As a quick experiment, I submitted a few controversial queries to Andy and found that the search engine handled them quite well, consistently pointing to factual sources. Searching for “Who Really Won the 2020 Election?” gave the answer “Joe Biden”, while the question “Are Covid-19 vaccines fake?” downloaded an article from Forbes refuting conspiracy theories about the pandemic.
Andi is still largely in the alpha phase and intends to stay lean as it iterates based on feedback from early users, Hoover says. The startup will have to make tough decisions. As a Fresh Yorker notessearch algorithms are prone to various errors, for example, they prioritize only sites that utilize newfangled web technologies. They also open the door to bad actors. In 2020 on Pinterest took advantage quirks of the Google image search algorithm that result in more content being displayed in Google image search results.
While struggling with these issues, Andi’s team continues to work on their business model. While the basic service will remain free, Hoover says Andi will eventually offer paid professional and business plans with premium features and API access, allowing customers to utilize Andi’s search functionality and answer questions across paid content, personal information, and internal company and team data .
Paid features are probably the way to go, given that Google’s share of the global search market has remained steady at over 90% for most of the last decade. Bing with 3.4%, second place is taken by Yahoo! (full disclosure: parent company TechCrunch) at 1.34%, According to to Statcounter.
To fund the development of these features and potential partnerships with alternative search engines, Andi recently raised $2.5 million, which includes support from YC, Gaingels, GoodWater Capital, K20 Fund, Acacia Venture Capital Partners, Fepo Capital and BBQ Capital, as well as a compact fund family and friends around.
“We kept fuel consumption low by working as digital nomads outside Mexico to extend our runway and staying fuel efficient. Even with the addition of AI developers and increased model training costs, we have a runway well beyond two years,” Hoover said. “We are using these funds to enhance our proprietary generative AI for answering complex questions and the “vertical search and API search” technology that Andi uses to combine gigantic language models with live data, specifically: developing and training AI models , adding some AI developers to our team, and hosting and inference costs as we start to raise usage as we get closer to product-market fit… In this early stage, we’re focused on building really great search results that our users will love, before generating revenue. “
Andi doesn’t collect detailed data, but Hoover estimates that about 5,000 users currently utilize the search engine. Andi plans to hire a full-time AI developer before the end of the year, bringing the total number of employees to three, including Hoover and White.