By Daniela Rus, PhD, Director of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL)
Last week I was part of an epic unveiling of a bona fide time capsule here on the campus at MIT. The one that we were lucky enough to have hidden right within our own lab at CSAIL was something technology dreams are made of: a real-life time capsule that could be opened only if someone could solve a complex cryptography puzzle before 2033, the year the creator expected technology to be advanced enough to reveal the mystery.
Lucky for us, Bernard Fabrot, a self-taught programmer, discovered the solution to the puzzle in April, 15 years ahead of schedule. Thanks to his discovery, we were able to open the capsule on May 15, revealing a treasure trove of technology artifacts, including the original 1992 proposal for the World Wide Web, an Altair Basic interpreter from Bill Gates, and a 1979 user manual for VisiCalc. For us technology geeks, it was an amazing thrill. Someone wrote that my colleagues and I were “like a group of giddy schoolchildren opening Christmas presents.” And yet, the truth is that we were more like children opening Christmas presents in July—long before Santa’s sleigh was due to arrive with a bag full of treats! In my eyes, that was the greatest thing about it: Fabrot was able to beat all expectations for solving the puzzle using the combination of a simple algorithm and the unexpected power of today’s computing technology. The implications of that acceleration of computing power reach far beyond the contents of our little time capsule and into nearly every aspect of our lives.
To grasp the scope of this impact, it’s important to first understand exactly what it is that is facilitating this new speed of innovation. What many people think of as AI is actually comprised of three connected, interrelated fields that are suddenly coming together at lightning speed:
1: Robotics (which puts computation in motion)
2: Artificial Intelligence (which enables machines to reason and make complex decisions)
3: Machine Learning (which enables machines to use the power of data to learn, identify patterns, and make decisions with minimal human intervention)
As these three areas continue to progress more rapidly than anyone anticipated, the benefits to every field are expanding greatly—especially when AI is used to help humans do our jobs better and faster than ever. Today, artificial intelligence and robotics are beginning to permeate many parts of our lives, at work and at home, as machines increasingly take on routine tasks so we can focus more on the strategic and fun aspects of our work and life. Some of the most fun examples include:
- Robotic lawnmowers can do your landscaping, giving you time to take your robotic dogs for a walk.
- A remote driver—located at a call center anywhere in the world—will take on the task of navigating your autonomous vehicle through traffic, giving you time to relax after a long work day.
- A new mobile messaging app called Mei uses machine learning to root out subtext, giving you the power to (finally) “get” the sarcasm in your teenager’s text—and even offer a suggested response to help bridge the communication gap.
- A robotic hand named Dextra can challenge you in rock-paper-scissors (but don’t expect to win; Dextra uses a brain-inspired neural network and camera to determine its opponent’s next move and execute a winning symbol 30 times faster than the quickest human).
All AI technologies, from the most amusing to the most transformative, work most effectively when AI is applied in a way that assists humans, rather than attempts to replace them. The reason: people excel at some things, while machines excel at others. It is when our diverse capabilities are combined that the results are most surprising.
For example, in a recent study, an AI system was used to review images of lymph node cells to diagnose cancer. The system’s error rate: 7.5%. That doesn’t seem very impressive when human pathologists in the same study scored much better, with an error rate of just 3.5%. And yet, when the AI system and the pathologists both reviewed the data, the total error rate dropped to just 0.5%—a significant difference that could easily translate into improved outcomes for patients.
In our own lab at CSAIL, we’re working with Toyota on Guardian, an advanced driver assist system designed to avoid crashes by correcting the driver’s mistakes in response to sensor data. For example, if the driver maneuvers to change lanes without noticing a car in its blind spot, the system alerts the driver and steers to avoid the collision. "The human is still the primary driver," says Ryan Eustice, senior vice president of automated driving at the Toyota Research Institute. "It is working with the human and creating a bubble around you of 360-degree situational awareness." That bubble has the potential to save many lives in a world that sees 1.3M fatal crashes on the highway every year.
We’re also working to apply the Guardian technology in the healthcare field, creating a system that provides that 360-degree situational awareness for surgeons, effectively looking over their shoulders to perceive factors that the surgeon can’t focus on while performing a complicated surgery. For instance, the system can identify issues that are not apparent to the naked eye, such as internal bleeding, and alert the surgeon to take action before the problem escalates.
Along those lines, scientists at the Allen Institute for Brain Science are using machine learning to tackle another area of medicine: teaching computers a new way to identify structures inside living cells and enable researchers to see many more cells at one time. This capability has deep implications in areas like cancer biology and regenerative medicine where the ability to view and analyze the 20,000+ different proteins in the human body together could transform our knowledge of healthy cells—and our ability to diagnose and treat diseased cells.
That said, even these advanced machines don’t share the cognitive capabilities of humans. You can teach a machine to identify a “water bottle” by showing it hundreds or even thousands of examples, and yet, without being given additional information, the machine can’t extrapolate that data to understand what a water bottle actually does. For humans, that level of thinking comes easily. For machines, no matter how advanced they may be, their knowledge will always be limited to the information we provide.
Ultimately, AI tools are being created by people, for people. As innovation continues to accelerate, it is becoming exceedingly clear that the most powerful results will be realized when we focus on applying advanced technology in a way that enhances the creativity, communication, and wisdom of humans. Making the most of this division of labor between people and machines depends on understanding what machines are good at—and what they’re not. To put it in the simplest terms, machines have chips and people have hearts. They have speed, and we have wisdom. Robots can move with precision and perform low-level, detailed tasks. AI can learn over time, identify patterns, and generate insights that humans would never be able to do on their own. By bringing together the strengths of humans and machines, with each one filling in the weaknesses of the other, AI is enabling humans to reach new heights in nearly everything we do.
For investors, this is all good news. While there will always be naysayers who fear AI, the evidence is mounting that the more advanced AI becomes, the more applications there will be to put that power into action to help humans do more—better. As a scientist, I have little doubt that the exponential increase in the performance capabilities of AI will continue, and that the speed of change will create opportunities no one among us can even begin to predict. I believe the greatest opportunities—for investors and humanity as a whole—will lie where the physical and computational power of AI is combined with our own cognitive and strategic strengths. It is up to humanity to decide what impact AI will have. It is up to investors to decide how to make the most of that impact.
About the Author
A member of the ROBO Global Strategic Advisory Board, Daniela Rus is the Director of CSAIL at MIT. She serves as the Director of the Toyota-CSAIL Joint Research Center and is a member of the science advisory board of the Toyota Research Institute. Rus’s research interests are in robotics and artificial intelligence. The recipient of the 2017 Engelberger Robotics Award from the Robotics Industries Association, she is also a Class of 2002 MacArthur Fellow, a fellow of ACM, AAAI and IEEE, and a member of the National Academy of Engineering and the American Academy of Arts and Sciences. Daniela earned her PhD in Computer Science from Cornell University.