Andrew Ng, one of the world’s best-known artificial-intelligence experts, is launching an online effort to create millions more AI experts across a range of industries. Ng, an early pioneer in online learning, hopes his new deep-learning course on Coursera will train people to use the most powerful idea to have emerged in AI in recent years.
AI experts have become some of the most sought-after and well-paid employees in today’s tech economy. Deep learning involves teaching a machine to perform a complex task using large amounts of data along with a large simulated neural network. The technique has typically required deep technical knowledge and expertise to master (see “10 Breakthrough Technologies 2013: Deep Learning”).
Ng left his post as chief scientist at the Chinese Internet company Baidu this March, and there has been widespread speculation about what his next move might be (see “Andrew Ng Is Leaving Baidu in Search of a Big New AI Mission”). He says the deep-learning course is the first of three projects he plans to launch through his new startup, Deeplearning.ai.
Before joining Baidu, Ng was a professor at Stanford and then the founder of Coursera, creating a machine-learning course that has attracted more than two million enrollees over the past few years. Several years ago, Ng was also the founding director of the Google Brain project, an effort to deploy deep learning across the company.
Ng spoke with MIT Technology Review senior editor Will Knight about his mission to rebuild the world using deep learning.
Why did you choose to launch this as your next big endeavor?
The thing that really excites me today is building a new AI-powered society. Even though a lot of the buzz in AI has been around large tech companies—and clearly the large tech companies are creating huge amounts of value with AI through better Web search, online advertising, better maps, better payment systems, and so on—if you look across an entire economy, really any Fortune 500 company can create a lot of value with AI as well.
I don’t think any one company could do all the work that needs to be done, so I think the only way to get there is if we teach millions of people to use these AI tools so they can go and invent the things that no large company, or company I could build, could do.
I think the world will just be better if AI is helping us. It will reduce the cost of goods, giving us good education, changing the way we run hospitals and the health-care system—there’s just a long list of things.
Do we really need so many AI experts?
I don’t think every person on the planet needs to know deep learning. But if AI is the new electricity, look at the number of electrical engineers and electricians there are. There’s a huge workforce that needs to be built up for society to figure out how to do all of the wonderful stuff around us today. So the AI workforce today is so much smaller.
While the university system is shifting in the direction of AI, we can’t rely only on universities. When I learned to code 25 years ago, I learned Basic, and then the world changed and I learned C and then C++, then Java and then Python. And I think the world is changing even faster today. I think the deep-learning transformation is much bigger, and the value creation and the demand for these skills is so clear.
I’ve sat down with CEOs of very large non-tech companies and we’ve brainstormed ideas for new AI products for their industries, but they don’t have anyone with the deep-learning skill set to build it. If you manage tens of thousands of people and you don’t have an AI group, what can you do? I know some of these CEOs intend to send their existing employees to take this deep-learning specialization.
What does your effort say about existing academic institutions?
I think we need lifelong learning. We’re used to this in the tech world. New stuff just gets invented every few weeks. We’re used to the sands constantly shifting under our feet, and I actually like that. I find it makes it more exciting to work in tech. But the AI deep-learning revolution is one thing that’s shifted so fast. Even though universities are ramping up their teaching capacity, there are so many people who are already out of the university system that need to learn these new systems.
I want an AI-powered society because I see so many ways that AI can make human life better. We can make so many decisions more systematically or automate away repetitive tasks and save so much human time. I feel like there are so many good opportunities, and so few people with the skills that are needed to capture these opportunities.
Could deep learning help automate education itself?
I hope that deep learning will play a role in remaking the educational system as well. I think it’s pretty clear it will, actually. Education is one of the industry categories with a big potential for AI. And Coursera is already doing some of this work.
Coursera has a ton of educational data. We know what videos students watch; we know which questions they get wrong. Coursera has an ability to start looking at that data, and we’ve been doing some of that work.
In a lot of industries there are two waves of transformation. One is the IT transformation; then comes the AI transformation. I think education has to do more of the IT transformation, but there are a lot of opportunities for AI as well. I was at an education conference a short while ago, and I was actually surprised by how many AI talks there were.
Is this part of an effort to offset the potential negative impacts of AI?
The biggest negative impact of AI will be job displacement, which is going to happen. But while we’re creating the problem, we should work on creating the solution as well, which I think is going to be better education. It’s not like society is running out of jobs to do; it’s just that the skill sets are mismatched to all the new jobs. I think it’s an educational problem. I do think we need a lot more than Coursera—we need a societal rethinking of our education system.
The East and West Coast quote “elites” have certainly made the East and West Coasts much wealthier. But if we want to create not just a wealthier society but also a fairer one, there is work to do in terms of rebuilding the social contract. There are more pieces of the puzzle. One is just bringing a lot more people into deep learning.
What do you make of China’s recently announced plan to dominate AI?
There’s tremendous interest in AI across the industry in China, and I’m actually very worried about decreases in science and technology funding in the United States. About a decade ago, my work at Stanford was funded by the National Science Foundation and by DARPA, and to be really honest I couldn’t have done it back then without that funding. That also allowed me to propose the Brain project to Google.
The question is which young professor is getting their grant application denied, and so won’t do something five years from now because the NSF didn’t have the funding to fund their proposals.
I find that quite scary. In contrast, other countries, such as China, are investing very heavily in AI research, and the environment for getting that work done is just becoming easier. I see a lot of willingness to adopt new technologies.
A horrifying new AI app swaps women into porn videos with a click
Deepfake researchers have long feared the day this would arrive.
DeepMind’s AI predicts almost exactly when and where it’s going to rain
The firm worked with UK weather forecasters to create a model that was better at making short term predictions than existing systems.
People are hiring out their faces to become deepfake-style marketing clones
AI-powered characters based on real people can star in thousands of videos and say anything, in any language.
What an octopus’s mind can teach us about AI’s ultimate mystery
Machine consciousness has been debated since Turing—and dismissed for being unscientific. Yet it still clouds our thinking about AIs like GPT-3.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.