Today we are pleased to feature a guest post from Graham Sabin the CEO and co-founder of Pondr. Founded in Seattle in late 2020, Pondr is helping to legitimize and authenticate consumer reviews for the online retail market.
GPT-3 is at the core of many new startups. Having the reference to the AI on a landing page is a flex in the tech world. The exclusivity around GPT-3 gives credibility to those companies lucky enough to use it. The beta for GPT-3 was released by OpenAI in June, 2020 and has since been game-changing in the tech world.
Back in late 2020 my co-founders and I began solving a problem that companies face when selling physical products: Getting actionable, qualitative feedback from customers in order to improve products. A combination of interviewing companies and brainstorming helped us see a core problem for companies and it revolves around their abundance of product reviews. Thousands of reviews, full of usable data, are essentially left behind on eCommerce sites because they’re impossible to manage at scale. Using natural language processing (NLP), we started helping companies visualize the meaning behind their reviews to draw out consumer sentiment, product trends over time, and specific likes/dislikes on features, all displayed in understandable charts and graphs. To increase offerings to customers, we played around with GPT-2 but were left unimpressed. It’s accuracy and summarization abilities simply did not meet our needs.
Around this time, our acceptance into Microsoft for Startups opened the door to new opportunities and resources, one of those being OpenAI. Our use case convinced OpenAI to give us access to their GPT-3 API and its abilities have been astounding. The revolutionary tech has allowed us to create tools we could only imagine before. Arguably our most anticipated feature lets companies ask questions directly to their reviews. Using GPT-3, we’ve created an AI Q & A model which is trained by a company’s dataset of reviews. Like never before, companies can dive into their reviews to ask questions like “What features should be added?” or “What was the hardest part about using this product?”, receiving actionable answers instantly, based on what real customers have said. We’ve seen the power of these responses prompt companies to add specific new features to products, include additional instructions for setup, or change sizing for increased ergonomics and usability. GPT-3 has also shown its ability when it comes to categorizing extensive datasets. Using supervised as well as unsupervised categories generated by the AI, we’re able to sort thousands of reviews allowing companies to dissect specific aspects of their products based on customer comments.
Within the coming months, Pondr will continue using GPT-3 to guide companies to improve their products, but additionally we are venturing to disrupt the way everyday consumers shop online. Currently, consumers are stuck reading through reviews on Amazon or Shopify in order to get a full picture about what they should buy. By rolling out our AI Q & A to the general public, we’ll be creating a personal shopping assistant for anyone on the web. When you have a question about a product, you’ll ask it to thousands of users at once rather than reading just one review to build a story.
