AI companies are gobbling down investor capital and reaching sky-high valuations early in their lifecycles, leading many to say the industry is a bubble.
Nick Frosst, co-founder of Cohere, which builds custom AI models for enterprise customers, recently said on TechCrunch's Found podcast that he doesn't think the AI industry is in a bubble. He acknowledges that the industry is in a bubble, but he believes calling it one discredits companies like Cohere that are building features that are truly useful to customers.
“We frequently see people using our model, enabling entirely new capabilities that weren't possible before, or automating processes that were really tedious and slowing everything down,” Frost says. “And that's the measurable value. It's hard to get into a complete bubble when you have all this usefulness.”
But that doesn't mean Frost is optimistic about everything the industry is building. He doesn't think AI will ever reach artificial general intelligence, defined as human-level intelligence, a notable difference in opinion from AI colleagues like Mark Zuckerberg and Jensen Huang. He added that it will be a long time before the industry gets there, if at all.
“I don't think we're going to see digital gods popping up anywhere in the near future,” Frost said, “and I think more and more people are waking up to it and saying, this technology is great. This is very powerful, this is very useful. It's not digital gods. It requires a shift in how we think about technology.”
Frost said Cohere tries to be realistic about what AI technology can and can't do, and what kind of neural networks can provide the most value. Cohere's approach to building its business model is based on research done by Aidan Gomez, Cohere's co-founder and CEO, from his time at Google Brain. Gomez is, of course, known for his extensive AI research. He's best known as a co-author of the paper that brought AI the Transformer model that ushered in this era of generative AI, but he also co-authored a paper in 2017 called “One Model to Learn Them All.” The research concluded that comprehensive large-scale language models are more useful than smaller models trained on specific tasks or data from specific industries, Frost said.
Cohere now uses its main model as a base to build custom models for enterprise clients.
“As humans, we specialize. We go into specific areas, but the first part of our education is learning how to use language in general,” Frost says. “We spent a long time learning how to read and write. It wasn't until much later that we identified specific subfields of language. Something similar is happening with neural nets.”
But on a larger scale, while he believes the foundational model will win in his market (among the companies building such services), he doesn't believe enterprise companies should demand that their single model do all the consumer tasks, B2B tasks, product tasks, etc.
Frost said companies that want to successfully leverage AI technology need to focus and be aware of what it can and can't do.
“We're pretty calm about how useful this technology is and what value it could bring — and frankly, it's tremendous value,” Frost said, “but we don't think it would result in the deaths of all humanity, so we're able to take a practical approach that gets away from the extreme rhetoric on either side.”