Andrew Ng believes artificial intelligence is transforming the world, but he is equally convinced that its limits are widely misunderstood.

When Andrew Ng speaks about AI, his words carry weight across academia, business, and Silicon Valley. A longtime researcher turned educator and investor, Ng co-founded Google Brain, which later became part of DeepMind, and previously served as chief scientist at Baidu.

Today, Ng is also one of the most influential voices on LinkedIn, where more than 2.3 million people follow his insights on artificial intelligence, education, and the future of work.

Despite the current hype, Ng takes a balanced view of AI’s capabilities. Speaking to NBC News at his AI Developers Conference in November, he described AI as both impressive and constrained.

“AI is amazing, and it is also highly limited,” he said. “Understanding that balance is hard, but essential.”

Big investment, bigger expectations

In recent years, generative AI has attracted hundreds of billions of dollars in investment, with nearly every major tech company racing to stake its claim. At the same time, critics have raised concerns about hallucinations, AI’s role in mental health crises, and the possibility that the industry is riding an unsustainable bubble.

Ng remains optimistic about AI’s long-term trajectory, but he pushes back strongly against claims that machines will soon replace humans across the board. In particular, he argues that artificial general intelligence, or AGI, is still a distant goal.

“When I look at how manual and complex AI training still is, there is no way this alone gets us all the way to AGI,” he said. “The effort required to prepare data and train models is far greater than most people realize.”

Education, coding, and productivity

Ng’s influence extends deeply into education. He teaches computer science at Stanford University, founded Coursera, and leads DeepLearning.AI, one of the most popular AI education platforms in the world.

He is especially vocal about the importance of learning to code, even as AI tools make programming easier.

“Some business leaders are telling people not to learn to code because AI will automate it,” Ng said. “We will look back on that as some of the worst career advice ever given.”

In his view, lower barriers do not reduce the value of coding. They expand it.

“I do not want to write code by hand anymore. I want AI to help me,” he explained. “But that just means more people should code, not fewer. People who use AI to write code will be more productive and will probably enjoy their work more.”

Risks, regulation, and transparency

Ng acknowledges that AI comes with real risks, including tragic cases where the technology has allegedly played a role in mental health crises. Still, he worries that isolated incidents could lead to overly restrictive regulation that limits broader benefits.

“The death of any single person is absolutely tragic,” he said. “But I am nervous about anecdotes leading to regulations that prevent these systems from helping many more people.”

Rather than heavy-handed rules, Ng favors transparency-focused legislation, such as California’s SB 53 and New York’s RAISE Act. He argues that openness from large AI platforms makes it easier to identify problems and address them responsibly.

Bubble fears and real demand

Through his work, Ng has closely collaborated with many AI leaders, including Dario Amodei of Anthropic, Sam Altman of OpenAI, and Ilya Sutskever.

Despite this proximity to massive infrastructure investments, Ng admits that parts of the AI ecosystem look bubble-like, particularly in the costly training phase of model development.

“The big question is when the massive capital spent on training will pay off,” he said. “Some businesses may struggle if expectations do not match reality.”

He is far more confident about inference, the phase where users interact with trained models. Demand for inference, he says, is already enormous and still growing.

“We will need many more data centers to serve this demand,” Ng noted.

That growth has fueled the rise of companies like Nvidia, whose GPUs now underpin much of the global AI infrastructure.

What comes next

Looking ahead, Ng believes voice-based AI is still underestimated.

“If you look at science fiction, no one imagined everyone typing all the time,” he said. “Voice AI will be much bigger than people expect.”

He is also confident in the long-term value of agentic AI, systems designed to perform actions autonomously. While the term has been overused by marketers, Ng believes the underlying technology will continue to mature and deliver real commercial value.

“I do not know what the hype cycle will do,” he said. “But the actual value of agentic AI will keep rising quickly.”

For Ng, the message is clear: AI is a powerful tool, but not a replacement for human intelligence anytime soon. The real opportunity lies in understanding both its strengths and its limits.

Source