Questions about sports trading cards? Ask CollX’s new AI feature

CollX, a collectables valuation app for items such as sports trading cards, announced the feature CollX AI last week, which is available on CollX Pro, the platform’s paid premium tier.

Cofounder and CEO Ted Mann said the South Jersey-founded company started working on an artificial intelligence feature earlier this year and released a private beta version five months ago, only available to a select group of “power users” on the app.

The new feature came after Mann had been playing around with ChatGPT and became interested in what the tool knew about his startup’s industry.

“We got impressed enough that we thought that if we extended the experience and really supercharged it with a lot of knowledge about trading cards, we could make it into an incredibly powerful utility for all card collectors,” Mann told Technical.ly.

The launch comes alongside a slew of other tech companies adding AI capabilities, from Guru’s Answers to Google’s Workspace, and even social science researchers are using ChatGPT in their work, as the technology remains in a very bright spotlight.

CollX already had a chat interface on its app where users could talk to each other, but Mann said he originally didn’t expect to integrate large language models. He and CTO Kostas Nasis took a class to learn how to create an AI proof of concept before building the actual feature.

CollX AI was built using GPT4, but the company also made custom indexes using Llamaindex.

ChatGPT only knows about events up until 2021, so it didn’t have any knowledge about athletes or trading cards past that year. Nasis said CollX supplemented the tool with its own data about cards that it collected from users.

And because CollX launched after 2021, the founders needed to add information about the company and its history to the tool as well.

CollX AI can be used to ask questions about specific trading cards, for example, if you wanted to learn more about the history of a card.

However, users weren’t asking other types of questions, and Mann and his team realized they needed to prompt their users to do so, such as what are the best cards from a certain year or season.

“What we learned from that period is that it’s really important with this technology, not only that it work[s], but that you encourage the users and almost coach them into asking it different kinds of questions and using it different ways to understand where they’re gonna find the most utility,” Mann said.

There’s also an opportunity for AI to assist with listing cards and pushing them out to the CollX Marketplace and CollX’s Card Dealer Pro site, Nasis said.

OpenAI recently announced features where users can talk to ChatGPT like Alexa or Siri or send images to the tool and ask questions. Nasis said he sees an opportunity to integrate those features into CollX AI as well.

“So, being able to have people take pictures of cards and having the AI tell them about that card and what they should do with that card,” Nasis said. “Which cards to hang on to, which ones you can sell. A lot of different things around the value of their collection, strategies for collecting. There’s a lot of room for growth. We’ve really just scratched the surface.”

Sarah Huffman is a 2022-2024 corps member for Report for America, an initiative of The Groundtruth Project that pairs young journalists with local newsrooms. This position is supported by the Lenfest Institute for Journalism.