Andrew Ng warns AI hype is distorting its true value 

With over two decades of experience in artificial intelligence and machine learning, Andrew Ng has a unique perspective on its evolution. Speaking at the CrewAI AI Agent Summit in New York last week, he criticized companies for hyping AI’s capabilities, while emphasizing its genuine potential. (Note that both boldstart and Ng have invested in Crew AI.)

Without naming names, he called out companies whose PR machines are overemphasizing how powerful they are for their own gain. “There's a certain set of views that have been disproportionately amplified, such as how technology is so dangerous you could wipe out humanity, which I don't think is true, but it makes them look more powerful,” Ng said.

He emphasized the importance of identifying practical use cases for agentic AI today, beyond the foundational models themselves. “I think the technology is working when you identify the right use cases and build to it, there is a very significant ROI,” Ng said at the Crew event. “I think a lot of the ROI of AI stories have been muddied by the fact that training foundation models has cost many billions of dollars.”

Ng sees numerous opportunities to achieve faster returns on investment once you move beyond the hype surrounding model costs. “But for people that are not training these giant models, but building agentic applications on top of these models, we can access them inexpensively when you find the right use cases. Then it becomes much more capital efficient to build an application and get ROI,” he said.

There's a certain set of views that have been disproportionately amplified, such as how technology is so dangerous you could wipe out humanity, which I don't think is true, but it makes them look more powerful.

And the beauty of these agents, especially coding assistants, is they open up a world of possibility that simply wasn’t there before because of time, personnel or cost limitations. Many projects that traditionally fell off priority lists can now be completed thanks to the efficiencies brought by agentic coding.

Ng sees the potential here, perhaps as well as anyone, but he also recognizes that it is an evolving process. “Just as human work is complicated, and when we hire a new employee, it takes a while to train them, it’s taking everyone a while to look at the very complex workflows and to build AI agentic workflows,” he said. 

That said, if we can get past the hype and continue building, Ng sees plenty of potential to get work done today with the current wave of technology. “The PR hype is disconnected from the business reality of progress. We frankly [need to] ignore most of the hype and just keep doing the work that creates value,” he said.

Photo by Steve Johnson on Unsplash