Technological advances have the power to alleviate suffering and improve living standards, but we still have much work to do before we can realize their benefits broadly, on a global scale. We cannot take innovation, or the benefits it brings, for granted.
Basic research by governments, corporations, and universities has led to new technologies as diverse as GPS, the internet, vaccines, and, most recently, artificial intelligence. To build a more prosperous future—and to make sure that prosperity is shared widely—basic scientific research must remain among our highest priorities.
I’ve spent my life working with AI and related technologies, and the progress has been breathtaking. Today, we’re rapidly moving toward a world where computers routinely make decisions for humans. This is a massive shift, and I believe it will make the world a better place.
If we want to build a future of shared prosperity, enabling fundamental research must be among our very highest global priorities.
To realize the full promise of AI, however, we need significantly greater understanding—of how it works and exactly where it falls short. Gaining this understanding will take an enormous amount of good, old-fashioned scientific research, just as it has for advances in medicine, agriculture, and transportation.
In particular, I believe that to build that AI-powered future safely and sanely, we must focus on four key areas.
1. Complexity and Reliability
Years ago, algorithms were pretty simple, but today we rely on extraordinarily complicated computer systems. Managing the complexity of these systems has become a real engineering challenge. It’s one thing if they are performing a simple and harmless task, but handling complexity reliably is critical when they perform more important tasks like driving your car or plane.
We are making software that is far more elaborate than anything humans have ever built before, and we have to learn how to build it in a consistently safe and dependable way. Doing so will require fundamental advances that only basic research can bring about.
2. Security
Related to this challenge is security. As our dependence on decision-making algorithms grows, we need to make sure all our systems are well protected. Despite some important advances in this area, this problem will become even more acute as decision-making algorithms proliferate.
Any society has bad actors, and we have to assume that such people will try even harder to exploit the security flaws of increasingly critical software. People didn’t pay much attention to security when the internet was first invented, and today we continue to pay the price. So let’s learn from our experience and devote the resources necessary to learn how to truly safeguard our systems.
3. Data Integrity and Bias
Substantial problems in data privacy, data ownership, and data bias remain unaddressed. Without solutions, the algorithms that increasingly drive our world are at high risk of becoming “data-compromised.”
Legitimate concerns about data use hinder the development of new and beneficial algorithms. This means we are missing the opportunity to tackle pressing problems because of the lack of accessible data. And part of the reason is that we haven’t solved data privacy and ownership issues.
Again, significant fundamental research into data integrity and bias is urgent.
4. Explainability
Finally, the notion of explainability remains a central problem in machine learning, a type of artificial intelligence that’s finding widespread application across industries.
Unfortunately, machine learning algorithms can’t provide an explanation for their outputs—at least not in the way people normally understand the word “explanation.” This is obviously a problem not just for researchers, but also for society, which increasingly uses such algorithms in the course of everyday life.
We must be able to understand the world we’re interacting with to be comfortable living in it, and as AI becomes more ubiquitous, people eventually won’t tolerate its inability to explain its decisions.
To realize the full promise of technological innovation, we must focus on solving fundamental scientific problems.
The way researchers approach these four issues will have huge implications for the world we live in tomorrow. And similar challenges affect nearly every other area of revolutionary innovation, be it gene editing, robotics, or distributed ledger technology.
If we want to build a future of shared prosperity, enabling fundamental research must be among our very highest global priorities.