Nick Frosst, co-founder of Cohere, believes everyone needs to be more realistic about what AI can and can’t do

AI companies are swallowing up investor money and achieving astronomical valuations early in their lifecycles. This dynamic has led many analysts to call the AI ​​sector a bubble.

Nick Frosst, co-founder of Cohere, a company that builds custom AI models for enterprise clients, recently told TechCrunch’s Found podcast that he doesn’t think the AI ​​industry is in a bubble. While he acknowledges the excitement in the space, he thinks calling it a bubble discredits companies, like his own company Cohere, that are building features that are actually useful for their customers.

“I often see someone using our model and they’ve enabled a completely new feature that wasn’t possible before or they’ve automated a process that was really slowing them down,” Frosst said. “And that’s tangible value. It’s hard to create a complete bubble when you have something that’s that useful.”

But that doesn’t mean Frosst is optimistic about everything the industry is building. He doesn’t think AI will ever reach artificial general intelligence, defined as human-level intelligence, which is a significantly different narrative than some of his AI peers like Mark Zuckerberg and Jensen Huang. He added that if the industry does get there, it won’t be for long.

“I don’t think we’re going to have digital gods anywhere anytime soon,” Frosst said. “And I think more and more people are coming to that conclusion and saying this technology is amazing. It’s super powerful, super useful. It’s not a digital god. And that requires adjusting the way we think about technology.”

Frosst said Cohere is trying to be realistic about what AI technology can and can’t do and what types of neural networks can provide the most value. Cohere’s approach to building its business model is based on the research of Cohere co-founder and CEO Aidan Gomez when he was at Google Brain. Gomez is, of course, known for his extensive research in AI. He’s best known for co-authoring a paper that helped AI acquire the transformer model that ushered in this era of generative AI. But he also co-authored a 2017 paper called One Model to Learn Them All. That research concluded that a large, global language model is more useful than small models trained for a specific task or on data from a specific industry, Frosst said.

Today, Cohere uses this core model as a basis for creating custom models for enterprise customers.

“We specialize as individuals. We orient ourselves toward particular domains. But the first part of our education is about how to use language in general,” Frosst said. “We spent a lot of time learning to read and write. It’s only very late that we specialize in a particular subdomain of language. So there’s something similar happening with neural networks.”

But even as he thinks bigger, the fundamental models that will prevail in his market – among those who create such services – he doesn’t think companies should ask their own unique models to do everything: consumer tasks, B2B tasks, product tasks.

Frosst says companies that want to successfully use AI technology need to be focused and also aware of what AI technology can and cannot do.

“We’re pretty clear-eyed about the utility of this technology and the value it can bring, and to be clear, insane value,” Frosst said. “But I don’t think it’s going to result in the death of all humans. So we’re able to take a realistic approach that maybe spares us some of the extreme rhetoric from either side.”