NEW YORK — When Andrew Ng talks about AI, people listen — in classrooms, boardrooms and in Silicon Valley.
THE researcher–become an educator–become an investor has become something of an AI statesman, co-founding Google Brain, which became part of Google’s flagship division DeepMind, which now produces some of the world’s best AI systems, and as chief scientist at Chinese tech titan Baidu.
In today’s influencer-obsessed news landscape, Ng’s biggest claim to fame might be her title as a “Top Voice” on LinkedIn, an honor the platform bestows on a select few experts with more than 2.3 million followers.
With decades of experience in AI, Ng says he remains clear-eyed about AI’s capabilities. “The problem with AI is that it’s amazing and it’s also very limited,” Ng told NBC News in an interview on the sidelines of its AI developers conference in November. “And understanding that balance between how amazing and limited it is, it’s difficult.”
Over the past few years, generative AI has attracted hundreds of billions of dollars of investment, as nearly every major tech company has turned its attention to the industry’s hottest topic. But in recent months, many have wondered whether the investment boom has created a bubble that is now at risk of bursting due to lingering problems. like hallucinations, AI’s involvement in mental health crises And increased regulatory oversight.
Ng is generally optimistic about AI’s upward trajectory, although he is quick to question the potential for AI systems to largely supplant humans in the near future. He has argued repeatedly that artificial general intelligence (AGI), defined roughly as AI systems capable of matching human performance on all meaningful tasks, is a distant possibility – unlike other AI luminaries who envision AGI’s emergence in the next few years.
“I look at how complex the training recipes are and how manual AI training and development is today, and there’s no chance that it’s going to get us all the way to AGI on its own,” Ng said.
“When someone uses AI and the system knows a language, it takes a lot more work to prepare the data, train the AI, learn that set of things than is widely appreciated,” he added.
Ng also enjoys an excellent reputation in the world of education. In addition to teaching computer science at Stanford University, Ng founded Coursera, one of the the world’s largest online learning platforms – and oversees one of the most popular AI-driven educational platforms, DeepLearning.AI.
With over a decade of success in the AI-meets-education ecosystem, Ng embraces a Chef Gusteau’s approach to AI and coding education in particular, arguing that everyone should code given advances in coding tools.
“Some top business executives have recently advised others not to learn to code, on the grounds that AI will automate coding,” Ng said. “We’ll consider this some of the worst career advice ever given. Because as coding gets easier, as it has for decades, as technology improves, more people should be coding, not fewer.”
Many experts recently stated this coding is “epicenter of AI advancements” and that the shocking capabilities of AI only become evident when people use AI tools to code. These developments have led some to theorize that coding jobs will disappear with the rise of AI, and early evidence supports these claims.
“It’s true that I don’t want to write code by hand anymore. I want AI to do it for me. But as the barriers get lower and lower, more and more people should do it. For example, my best recruiters don’t screen resumes by hand. They write prompts or write code to screen resumes.”
“People who use AI to write code will just be more productive and, I think, have more fun than people who don’t use it. There will be a big societal shift in favor of people who code,” Ng added.
As AI systems become more powerful, Ng is aware that real downsides are emerging – but he thinks the current risks pale in comparison to the potential benefits of AI.
“I think for many AI models, the benefits are much greater than the harms,” he said.
“The death of just one person is absolutely tragic,” Ng added, referring to recent suicides which would have involved the use of AI. “At the same time, I’m nervous about one or two anecdotes that lead to stifling regulations. That means it doesn’t save 10 lives, right? It’s a very difficult calculation for the number of people who get good mental health support from these systems.”
Instead of what he describes as stifling regulation, Ng is a strong supporter of laws that require transparency from major AI companies, like the recently passed law. SB 53 in California And RAISE law in New York.
“If I had a choice, if I were a regulator, I would push for transparency from the big platforms, because it gives us a much better chance to clearly see what the problems are, if any, and then work towards their solution,” Ng said.
Ng is also closely linked to many current private sector AI leaders: He supervised Dario Amodeico-founder and CEO of Anthropic, when Amodei worked at Baidu; briefly taught Sam Altman, co-founder and CEO of OpenAI, at Stanford; And was a postdoctoral advisor to Ilya Sutskever, co-founder of OpenAI who left to create a competing early-stage organization called Safe superintelligence.
Despite his ties to this Silicon Valley executive who announced billions of dollars investments in AI infrastructure, Ng does not hesitate to assert that part of today’s A.I. the landscape looks like a bubble.
Ng said the early stages of creating AI models, called “training” or “pre-training” stages, “is where a lot of the questions are, where the very real questions are. When will all the capital expenditures that go into that training be rewarded, when will they pay off?”
“Whatever happens will be good for the industry, but some companies may perform poorly,” Ng said, referring to the possibility of a bubble collapse.
Instead, Ng sees a steady and growing demand for the later stages of AI computing, called the “inference” stage, when users query pre-built AI systems. “The demand for inference is enormous and I am confident it will continue to grow.”
“We need to build many more data centers to meet this demand,” Ng added.
Ng was one of the first defenders of the most widely adopted approach today to creating the most advanced AI models, in which AI companies use powerful computing hardware called graphics processing units (GPUs), once used primarily by video game enthusiasts. Nvidia’s status as the world’s most valuable company stems almost entirely from its cutting-edge GPUs that now power many of the world’s best data centers.
As for other areas of AI, Ng said the public should pay more attention to voice-related AI. “I think people underestimate how big voice AI will become. If you look at the ‘Star Trek’ movies, no one imagined that everyone was typing, right?”
Ng also said the public should expect more and faster advances in “agentic AI,” a term he helped popularize that refers to AI systems that can perform many actions autonomously.
“Over the summer of last year, a group of marketers took the term ‘Agentic AI’ and slapped it like a sticker on everything in sight, causing the hype to take off incredibly quickly.”
“I am confident that the field of agentic AI will continue to grow and increase in value,” Ng said. “I don’t know what the hype will do. It’s hard to predict. But the actual business value will continue to increase rapidly.”