by Martin Ford
MARTIN FORD: Let’s talk about progress in AI toward human-level artificial intelligence or AGI. What does that path look like, and how close are we?
DANIELA RUS: We have been working on AI problems for over 60 years, and if the founders of the field were able to see what we tout as great advances today, they would be very disappointed because it appears we have not made much progress. I don’t think that AGI is in the near future for us at all.
I think that there is a great misunderstanding in the popular press about what artificial intelligence is and what it isn’t. I think that today, most people who say “AI,” actually mean machine learning, and more than that, they mean deep learning within machine learning.
I think that most people who talk about AI today tend to anthropomorphize what these terms mean. Someone who is not an expert says the word “intelligence” and only has one association with intelligence, and that is the intelligence of people.
When people say “machine learning,” they imagine that the machine learned just like a human has learned. Yet these terms mean such different things in the technical context. If you think about what machine learning can do today, it’s absolutely extraordinary. Machine learning is a process that starts with millions of usually manually labeled data points, and the system aims to learn a pattern that is prevalent in the data, or to make a prediction based on that data.
These systems can do this much better than humans because these systems can assimilate and correlate many more data points then humans are able to. However, when a system learns, for example, that there is a coffee mug in a photograph, what it is actually doing is it’s saying that the pixels that form this blob that represents the coffee mug in the current photo are the same as other blobs that humans have labeled in images as coffee mugs. The system has no real idea what that coffee mug represents.
The system has no idea what to do with it, it doesn’t know if you drink it, eat it, or if you throw it. If I told you that there is a coffee mug on my desk, you don’t need to see that coffee mug in order to know what it is because you have the kind of reasoning and experience that machines today simply do not have.
To me, the gap between this and human-level intelligence is extraordinary, and it will take us a long time to get there. We have no idea of the processes that define our own intelligence, and no idea how our brain works. We have no idea how children work. We know a little bit about the brain, but that amount is insignificant to how much there is to know. The understanding of intelligence is one of the most profound questions in science today. We see progress at the intersection between neuroscience, cognitive science, and computer science.
MARTIN FORD: Is it possible that there might be an extraordinary breakthrough that really moves things along?
DANIELA RUS: That’s possible. In our lab, we’re very interested in figuring out whether we can make robots that will adapt to people. We started looking at whether we can detect and classify brain activity, which is a challenging problem.
We are mostly able to classify whether a person detects that something is wrong because of the “you are wrong” signal—called the “error-related potential.” This is a signal that everyone makes, independent of their native tongue and independent of their circumstances. With the external sensors we have today, which are called EEG caps, we are fairly reliably able to detect the “you are wrong” signal. That’s interesting because if we can do that, then we can imagine applications where workers could work side by side with robots, and they could observe the robots from a distance and correct their mistakes when a mistake is detected. In fact, we have a project that addresses this question.
What’s interesting, though, is that these EEG caps are made up of 48 electrodes placed on your head—it’s a very sparse, mechanical setup that reminds you of when computers were made up of levers. On the other hand, we have the ability to do invasive procedures to tap into neurons at the level of the neural cell, so you could actually stick probes into the human brain, and you could detect neural-level activity very precisely. There’s a big gap between what we can do externally and what we can do invasively, and I wonder whether at some point we will have some kind of Moore’s law improvement on sensing brain activity and observing brainwave activity at a much higher resolution.
MARTIN FORD: What about the risks and the downsides of all of this technology? One aspect is the potential impact on jobs. Are we looking at a big disruption that could eliminate a lot of work, and is that something we have to think about adapting to?
DANIELA RUS: Absolutely! Jobs are changing: jobs are going away, and jobs are being created. The McKinsey Global Institute published a study that gives some really important views. They looked at a number of professions and observed that there are certain tasks that can be automated with the level of machine capability today, and others that cannot.
If you do an analysis of how people spend time in various professions, there are certain categories of work. People spend time applying expertise; interacting with others; managing; doing data processing; doing data entry; doing predictable physical work; doing unpredictable physical work. Ultimately, there are tasks that can be automated and tasks that can’t. The predictable physical work and the data tasks are routine tasks that can be automated with today’s technologies, but the other tasks can’t.
I’m actually very inspired by this because what I see is that technology can relieve us of routine work in order to give us time to focus on the more interesting parts of our work. Let’s go through an example in healthcare. We have an autonomous wheelchair, and we have been talking with physical therapists about using this wheelchair. They are very excited about it because, at the moment, the physical therapist works with patients in the hospital in the following way:
For every new patient, the physical therapist has to go to the patient’s bed where they have to put the patient in a wheelchair, push the patient to the gym where they’ll work together in the gym and at the end of the hour, the physical therapist has to take the patient back to the patient’s hospital bed. A significant amount of time is spent moving the patient around and not on patient care.
Now imagine if the physical therapist didn’t have to do this. Imagine if the physical therapist could stay in the gym, and the patient would show up delivered by an autonomous wheelchair. Then both the patient and the physical therapist would have a much better experience. The patient would get more help from the physical therapist, and the physical therapist would focus on applying their expertise. I’m very excited about the possibility of enhancing the quality of time that we spend in our jobs and increasing our efficiency in our jobs.
A second observation is that in general, it is much easier for us to analyze what might go away than to imagine what might come back. For instance, in the 20th century, agricultural employment dropped from 40% to 2% in the United States. Nobody in the 20th century guessed that this would happen. Just consider, then, that only 10 years ago, when the computer industry was booming, nobody predicted the level of employment in social media; in app stores; in cloud computing; and even in other things like college counseling. There are so many jobs that employ a lot of people today that did not exist 10 years ago, and that people did not anticipate would exist. I think that it’s exciting to think about the possibilities for the future and the new kinds of jobs that will be created as a result of technology.
MARTIN FORD: So, you think the jobs destroyed by technology and the new jobs created will balance out?
DANIELA RUS: Well, I do also have concerns. One concern is in the quality of jobs. Sometimes, when you introduce technology, the technology levels the playing field. For instance, it used to be that taxi drivers had to have a lot of expertise—they had to have great spatial reasoning, and they had to memorize large maps. With the advent of GPS, that level of skill is no longer needed. What that does is open the field for many more people to join the driving market, and that tends to lower the wages.
Another concern is that I wonder if people are
going to be trained well enough for the good jobs that will be created as a result of technology. I think that there are only two ways to approach this challenge. In the short term, we have to figure out how to help people retrain themselves, how to help people gain the skills that are needed in order to fulfill some of the jobs that exist. I can’t tell you how many times a day I hear, “We want your AI students. Can you send us any AI students?” Everyone wants experts in artificial intelligence and machine learning, so there are a lot of jobs, and there are also a lot of people who are looking for jobs. However, the skills that are in demand are not necessarily the skills that people have, so we need retraining programs to help people acquire those skills.
I’m a big believer in the fact that actually anybody can learn technology. My favorite example is a company called BitSource. BitSource was launched a couple of years back in Kentucky, and this company is retraining coal miners into data miners and has been a huge success. This company has trained a lot of the miners who lost their jobs and who are now in a position to get much better, much safer and much more enjoyable jobs. It’s an example that actually tells us that with the right programs and the right support, we can actually help people in this transition period.
MARTIN FORD: Is that just in terms of retraining workers, or do we need to fundamentally change our entire educational system?
DANIELA RUS: In the 20th century we had reading, writing, and arithmetic that defined literacy. In the 21st century, we should expand what literacy means, and we should add computational thinking. If we teach in schools how to make things and how to breathe life into them by programming, we will empower our students. We can get them to the point where they can imagine anything and make it happen, and they will have the tools to make it happen. More importantly, by the time they finish high school, these students will have the technical skills that will be required in the future, and they will be exposed to a different way of learning that will enable them to help themselves for the future.
The final thing I want to say about the future of work is that our attitude toward learning will also have to change. Today, we operate with a sequential model of learning and working. What I mean by this is that most people spend some chunk of their lives studying and at some point, they say, “OK, we’re done studying, now we’re going to start working.” With technology accelerating and bringing in new types of capabilities, though, I think it’s very important to reconsider the sequential approach to learning. We should consider a more parallel approach to learning and working, where we will be open to acquiring new skills and applying those skills as a lifelong learning process.
MARTIN FORD: Some countries are making AI a strategic focus or adopting an explicit industrial policy geared toward AI and robotics. China, in particular, is investing massively in this area. Do you think that there is a race toward advanced AI, and is the US at risk of falling behind?
DANIELA RUS: When I look at what is happening in AI around the world, I think it is amazing. You have China, Canada, France, and the UK, among dozens of others, hugely investing in AI. Many countries are betting their future on AI, and I think we in the US should do too. I think we should consider the potential for AI, and we should increase the support and the funding of AI.
DANIELA RUS is the Andrew (1956) and Erna Viterbi Professor of Electrical Engineering and Computer Science and Director of the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT. Daniela’s research interests are in robotics, artificial intelligence, and data science.
The focus of her work is developing the science and engineering of autonomy, toward the long-term objective of enabling a future with machines pervasively integrated into the fabric of life, supporting people with cognitive and physical tasks. Her research addresses some of the gaps between where robots are today and the promise of pervasive robots: increasing the ability of machines to reason, learn, and adapt to complex tasks in human-centered environments, developing intuitive interfaces between robots and people, and creating the tools for designing and fabricating new robots quickly and efficiently. The applications of this work are broad and include transportation, manufacturing, agriculture, construction, monitoring the environment, underwater exploration, smart cities, medicine, and in-home tasks such as cooking.
Daniela serves as the Associate Director of MIT’s Quest for Intelligence Core, and as Director of the Toyota-CSAIL Joint Research Center, whose focus is the advancement of AI research and its applications to intelligent vehicles. She is a member of the Toyota Research Institute advisory board.
Daniela is 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. She is the recipient of the 2017 Engelberger Robotics Award from the Robotics Industries Association. She earned her PhD in Computer Science from Cornell University.
Daniela has also worked on two collaborative projects with the Pilobolus Dance company at the intersection of technology and art. Seraph, a pastoral story about human-machine friendship, was choreographed in 2010 and performed in 2010-2011 in Boston and New York City. The Umbrella Project, a participatory performance exploring group behavior, was choreographed in 2012 and performed at PopTech 2012, in Cambridge, Baltimore, and Singapore.
Chapter 13. JAMES MANYIKA
Somebody should be thinking about what the regulation of AI should look like. But I think the regulation shouldn’t start with the view that its goal is to stop AI and put back the lid on a Pandora’s box, or hold back the deployment of these technologies and try and turn the clock back.
CHAIRMAN AND DIRECTOR OF MCKINSEY GLOBAL INSTITUTE
James is a senior partner at McKinsey and Chairman of the McKinsey Global Institute, researching global economic and technology trends. James consults with the chief executives and founders of many of the world’s leading technology companies. He leads research on AI and digital technologies and their impact on organizations, work, and the global economy. James was appointed by President Obama as vice chair of the Global Development Council at the White House and by US Commerce Secretaries to the Digital Economy Board and National Innovation Board. He is on the boards of the Oxford Internet Institute, MIT’s Initiative on the Digital Economy, the Stanford-based 100-Year Study on AI, and he is a fellow at DeepMind.
MARTIN FORD: I thought we could start by having you trace your academic and career trajectory. I know you came from Zimbabwe. How did you get interested in robotics and artificial intelligence and then end up in your current role at McKinsey?
JAMES MANYIKA: I grew up in a segregated black township in what was then Rhodesia, before it became Zimbabwe. I was always inspired by the idea of science, partly because my father had been the first black Fulbright scholar from Zimbabwe to come to the United States of America in the early 1960s. While there, my father visited NASA at Cape Canaveral, where he watched rockets soar up into the sky. And in my early childhood after he came back from America, my father filled my head with the idea of science, space, and technology. So, I grew up in this segregated township, thinking about science and space, building model planes and machines out of whatever I could find.
When I got to university after the country had become Zimbabwe, my undergraduate degree was in electrical engineering with heavy doses of mathematics and computer science. And while there a visiting researcher from the University of Toronto got me involved in a project on neural networks. That’s when I learned about Rumelhart Backpropagation and the use of logisti sigmoid functions in neural network algorithms.
Fast forward, I did well enough to get a Rhodes scholarship to go to Oxford University, where I was in the Programming Research Group, working under Tony Hoare, who is best known for inventing Quicksort and for his obsession with formal methods and axiomatic specifications of programming languages. I studied for a master’s degree in mathematics and computer science and worked a lot on mathematical proofs and the development and verification of algorithms. B
y this time, I’d given up on the idea that I would be an astronaut, but I thought that at least if I worked on robotics and AI, I might get close to science related to space exploration.
I wound up in the Robotics Research Group at Oxford, where they were actually working on AI, but not many people called it that in those days because AI had a negative connotation at the time, after what had recently been a kind of “AI Winter” or a series of winters, where AI had underdelivered on its hype and expectations. So, they called their work everything but AI—it was machine perception, machine learning, it was robotics or just plain neural networks; but no-one in those days was comfortable calling their work AI. Now we have the opposite problem, everyone wants to call everything AI.
MARTIN FORD: When was this?
JAMES MANYIKA: This was in 1991, when I started my PhD at the Robotics Research Group at Oxford. This part of my career really opened me to working with a number of different people in the robotics and AI fields. So, I met people like Andrew Blake and Lionel Tarassenko, who were working on neural networks; Michael Brady, now Sir Michael, who was working on machine vision; and I met Hugh Durrant-Whyte, who was working on distributed intelligence and robotic systems. He became my PhD advisor. We built a few autonomous vehicles together and we also wrote a book together drawing on the research and intelligence systems we were developing.
Through the research I was doing, I wound up collaborating with a team at the NASA Jet Propulsion Laboratory that was working on the Mars rover vehicle. NASA was interested in applying the machine perception systems and algorithms that they were developing to the Mars rover vehicle project. I figured that this was as close as I’m ever going to get to go into space!
MARTIN FORD: So, there was actually some code that you wrote running on the rover, on Mars?
JAMES MANYIKA: Yes, I was working with the Man Machine Systems group at JPL in Pasadena, California. I was one of several visiting scientists there working on these machine perception and navigation algorithms, and some of them found their way onto the modular and autonomous vehicle systems and other places.