Fintech Chime is launching an in-house ChatGPT AI tool this year

• Chime builds its own version of ChatGPT trained on its internal code base and data.
• ChatGPT and generative AI are transformative technologies, but there are IP and security issues.
• Chime’s CTO walks us through his strategy and how he navigates risk.

Chime is in the early stages of building its own private version of ChatGPT to launch this year, Insider has learned.

Neobanken is working with a stealth startup to train its own internal large-scale language models on its codebase to help Chime’s software engineers do their jobs faster and better, CTO James Barrese told Insider. If done successfully, Chime will unlock a competitive advantage by launching new products and features faster and more cost-effectively.

Large language models are the underlying algorithms behind ChatGPT, the souped-up AI chatbot that uses massive amounts of data and computing power to produce human-like responses. The bot uses generative AI, which refers to artificial intelligence that can produce content, from creating art to coding software for engineers. On Wall Street, technology has the potential to transform everything from sourcing investment deals to fighting fraud.

But questions around intellectual property, privacy and regulatory compliance have also emerged as the technology becomes more mainstream. Such uncertainties have made some companies hesitant to fully embrace the technology, with companies such as Amazon, JPMorgan, Goldman Sachs and Citibank discouraging employee use and blocking access to ChatGPT.

“The tools that are already out there are very good at showing the art of the possible, but there are still a lot of questions that need to be answered around IP and ownership,” Barrese said.

To try to get around that, Chime is working with a third party to use their technology to train AI on Chime’s code base in its own private cloud.

“It’s a way for us to go ahead and use it in real-world situations without worrying about any kind of intellectual property or security issues,” Barrese added.

Chime has been working on the stealth startup, which was only founded in January, for just one week, Barrese said. Chime is still in the exploratory phase and has yet to deploy any major language models to help engineers do their jobs faster.

He expects to be up and running with the startup, which he declined to name, within a few weeks and that the digital bank is “weeks and months” away from doing internal pilots that will last a quarter or two, Barrese said. “Definitely this year” Chime will use the technology in real-world applications, he added.

“This is a reflection of how quickly this space is moving,” he said.

For Chime, AI is about engineering productivity

As Chime expands its use of artificial intelligence — the fintech has used artificial intelligence to identify fraud and measure credit risk with its SpotMe product — it will hire more data scientists to build the models and engineers to build tools on top of the artificial intelligence, Barrese said. He declined to disclose hiring targets and the size of Chime’s technology organization.

In the near future, Chime’s generative AI efforts will be focused on engineering productivity. Barrese said he sees the technology as an assistant to the engineers and said AI will not replace engineers.

“You can’t trust it at this point, can you? It has too many errors, what they call hallucinations,” Barrese said. But it can be useful to give developers a starting point or framework to work within, in the same way that someone building a presentation might start with a predefined template.

Barrese acknowledged that beyond the initial concerns about security and ownership, the broad use of AI will bring to the fore many philosophical and ethical questions around how the technology is used and misused “that used to be theoretical and now are real.”

But for now, the fear of missing out is enough to keep Chime in the game.

“I’ve seen this play out time and time again, right? Where there’s a new innovation, and then everybody in the industry broadly recognizes that this disruption is coming, but then they turn around and refuse to engage with it,” Barrese said. who was previously CTO of PayPal and VP of Technology at eBay.

“I want to make sure that we understand where the technology is useful and where it is not useful, so that as soon as possible we can pivot and start doing things in a new way, in a new architecture, at least so that the technology matures , we can take advantage of that,” Barrese said.