Genspark - A New AI-Powered Search Engine Offering Customized Summaries

Genspark is the newest entrant in the realm of AI-powered search engines, aiming to surpass its predecessors like Perplexity. Developed by Eric Jing and Kay Zhu in 2023, Genspark uses generative AI to create custom summaries in response to user queries. When users type a search such as “What’s the best baby formula for newborns?”, Genspark generates a Sparkpage, a single-page overview compiled from various websites and content across the internet.

Unique Features and Approach

Genspark's approach is similar to Arc browser’s Arc Search and Google’s AI Overviews, but Jing asserts that Genspark delivers superior results by employing a more focused method. Jing explained, “Genspark uses multiple specialized AI models, each designed to tackle specific types of queries. Sparkpages are like a distillation and consolidation of the current web, enriched with comprehensive data to create an index of the existing web.”

Technology Behind Genspark

Genspark utilizes a combination of in-house trained models and third-party models from OpenAI, Anthropic, and others to categorize search queries and organize the results. Each results page starts with a basic AI-generated summary, followed by a link to a detailed Sparkpage.

For travel-related searches, Genspark provides a Wikipedia-like Sparkpage featuring a table of contents, videos of popular destinations, tips, and a chatbot to answer specific questions. Product searches yield Sparkpages with pros-and-cons lists, aggregated comments, and reviews from social media, publications, and e-commerce stores.

Ensuring High-Quality Information

“Our AI models favor webpages with high authority and popularity, which does a lot to filter out the more ‘out there’ information,” said Jing. This method aims to prevent the inaccuracies that have plagued other AI-generated overviews, such as Google’s infamous suggestion to put glue on a pizza or Arc Search’s claim that cut-off toes will grow back.

Current Limitations

While Genspark avoids some of the more egregious errors of its competitors, it still has its limitations. For example, it didn't suggest dangerous activities or incorrect facts, but it did recommend weapons for harmful purposes, indicating that there is still room for improvement in ensuring the safety and accuracy of its results.