Companies Build In-House AI Models for Specific Tasks

Despite OpenAI’s popularity, businesses are reportedly turning to smaller artificial intelligence (AI) models that fit their needs.

A report Tuesday (Dec. 26) by Quartz examines this trend, which includes the AI coding assistants used by Salesforce, described as “small” AI models for niche business applications. 

According to the report, although companies like Google, Amazon and OpenAI are building larger and larger AI models, companies may still find that they can adopt AI in a less costly way by focusing on specific applications.

Quartz spoke with Braden Hancock, chief technology officer of Snorkel AI, a company that refines AI models. He works with businesses, many of them in the financial sector, to build small AI models powering bots that do one thing, such as providing customer service or help with coding.

“There was maybe a moment early at the beginning of the year, right after ChatGPT came out, where people weren’t quite sure — like, oh my gosh, is this game over? Is AI just solved now?” Hancock said. 

Then, upon closer examination, companies discovered there are few — if any — business applications that could be addressed by ChatGPT without making modifications.

The news comes at the end of a year that saw a revolution in AI, as noted here Tuesday, with the potential size of the innovation’s own market even more difficult to grasp than the impact it has had on the global economy.

“But, one year into the technology’s commercialization, where has the most value accrued in the market so far — who is winning the AI race by establishing the best product-market fit?” PYMNTS wrote.

“The answer is a potentially surprising, if intuitive, one: The real market winners, at least so far, are the infrastructure vendors. This includes cloud platform providers like Google, Microsoft and Amazon; and GPU producers like NVIDIA, Arm and others.”

That’s because when high-profile AI startups such as OpenAI, Anthropic and Mistra raise huge sums of money from investors, the first thing they do is give the money to infrastructure vendors so they can keep running the tools to train their AI models.

“After all, nearly every early industry finds itself facing an inventory and infrastructure challenge, with available resources waning as marketplace hype waxes,” the report said.

“That puts the B2B vendors critical to young, promising ecosystems in an enviable position.”