Having spent many years working on AI research and building AI products, I’m fortunate to have participated in a few innovations that made an impact, like using reinforcement learning to fly helicopter drones at Stanford, starting and leading Google Brain to drive large-scale deep learning, and creating online courses that led to the founding of Coursera. I’d like to share some thoughts about how to do it well, sidestep some of the pitfalls, and avoid building things that lead to serious harm along the way.
AI is a dominant driver of innovation today
As I have said before, I believe AI is the new electricity. Electricity revolutionized all industries and changed our way of life, and AI is doing the same. It’s reaching into every industry and discipline, and it’s yielding advances that help multitudes of people.
AI—like electricity—is a general-purpose technology. Many innovations, such as a medical treatment, space rocket, or battery design, are fit for one purpose. In contrast, AI is useful for generating art, serving web pages that are relevant to a search query, optimizing shipping routes to save fuel, helping cars avoid collisions, and much more.
The advance of AI creates opportunities for everyone in all corners of the economy to explore whether or how it applies to their area. Thus, learning about AI creates disproportionately many opportunities to do something that no one else has ever done before.
For instance, at AI Fund, a venture studio that I lead, I’ve been privileged to participate in projects that apply AI to maritime shipping, relationship coaching, talent management, education, and other areas. Because many AI technologies are new, their application to most domains has not yet been explored. In this way, knowing how to take advantage of AI gives you numerous opportunities to collaborate with others.
Looking ahead, a few developments are especially exciting.
- Prompting: While ChatGPT has popularized the ability to prompt an AI model to write, say, an email or a poem, software developers are just beginning to understand that prompting enables them to build in minutes the types of powerful AI applications that used to take months. A massive wave of AI applications will be built this way.
- Vision transformers: Text transformers—language models based on the transformer neural network architecture, which was invented in 2017 by Google Brain and collaborators—have revolutionized writing. Vision transformers, which adapt transformers to computer vision tasks such as recognizing objects in images, were introduced in 2020 and quickly gained widespread attention. The buzz around vision transformers in the technical community today reminds me of the buzz around text transformers a couple of years before ChatGPT. A similar revolution is coming to image processing. Visual prompting, in which the prompt is an image rather than a string of text, will be part of this change.
- AI applications: The press has given a lot of attention to AI’s hardware and software infrastructure and developer tools. But this emerging AI infrastructure won’t succeed unless even more valuable AI businesses are built on top of it. So even though a lot of media attention is on the AI infrastructure layer, there will be even more growth in the AI application layer.
These areas offer rich opportunities for innovators. Moreover, many of them are within reach of broadly tech-savvy people, not just people already in AI. Online courses, open-source software, software as a service, and online research papers give everyone tools to learn and start innovating. But even if these technologies aren’t yet within your grasp, many other paths to innovation are wide open.
Be optimistic, but dare to fail
That said, a lot of ideas that initially seem promising turn out to be duds. Duds are unavoidable if you take innovation seriously. Here are some projects of mine that you probably haven’t heard of, because they were duds: