IBM, MIT Sloan School of Management and Harvard Business School today are sponsoring a symposium at the the two universities. The morning topic: How advances in information technology can help improve productivity, and improve incomes and create jobs for the 99%. It’s being followed this afternoon by a mock Jeopardy! match between Watson, IBM’s very smart computer, and teams from MIT and HBS.
Teams of three students from MIT/Sloan and HBS take on IBM’s Watson. (This is only the second contest matching Watson against collegians. In the previous contest, Watson beat teams from Carnegie Mellon University and the University of Pittsburgh. Pitt came in second, much to the chagrin of rival CMU!)
Harvard wins the first question, with “What is Belize?” Answering: countries in central America, ending with “e”
But then Watson takes over, running the category.
The machine picks “Who’s Your Daddy Company?” as the next category, eliciting a huge hook of laughter from the audience.
They finished the Jeopardy! round, with Watson, $8600; Harvard, $5200 ; and MIT, $-200 .
(I got disconnected from HBS’s Wi-Fi at a crucial moment, destroying the coverage of the second round. Grrrrr)
Clue: Finding the spot for this memorial caused its creator to say “Americans will march across that skyline.”
The question: Mt. Rushmore.
Harvard and Watson answer correctly. MIT does not.
Final score: Watson, $53,601; Harvard, $42,399; MIT, $100.
Erik Brynjolfson, MIT Sloan School of Management, kicks off by talking about a concern these days about technology and its role in society. Some people are saying that innovation has been stagnating, and that’s contributing to the slowness of the economic recovery.
“The issue isn’t that technology is stagnating, but that we haven’t been keeping up with technology. Societies, institutions and structures haven’t advanced rapidly enough to keep up with the advances. We’re creating a lot of wealth through technology, but the benefits aren’t going to regular people in the middle of the income distribution.”
This has been a great decade for productivity growth, even better the 1990s. It has contributed to income growth per capita. Yet median income growth has not improved much. “A lot of wealth has been created that goes to the people at the very top of the income brackets.”
The context: Here’s the new book by Erik and Andrew McAfee, Race Against the Machine.
IBM Fellow David Ferrucci talks about the making of Watson, IBM’s question-and-answer machine, which in February beat the top past champions on the TV game show Jeopardy! (He’s speaking again this afternoon, so I’m going to go into detail on that.) For now, here’s another IBM Fellow, Bernard Meyerson, talking about the importance of Watson.
Question to IBM’s David Ferrucci about Watson: How long until Watson can program itself?
It already does that, but will do so more in the future.
“I can imagine a situation where you’re searching for different models, different weights to answers, and it automatically programs itself to do that.”
He wouldn’t predict when.
10:45 a.m. Panel: What Can Technology Do Today and in 2020?
Andrew McAfee of MIT asks the question: Why are we seeing these astonishing advances now?
Afred Spector of Google: The Web makes it possible to combine a lot of information and access it via the Web. We also have a huge amount of feedback from users. And we have a large amount of software components. We can combine things and piece things together. “We’re solving a collection of problems which are acceptably probabilistic.”
Rod Brooks, an AI and robotics expert at MIT and entrepreneur: We have enough computer power to solve bigger and more complex problems. “Using machine learning and statistics we’ve managed to come up with algorithms which learn things acceptably well.”
David Ferrucci of IBM: “What’s exciting is the ability to generation hypothesis using induction and then track them back and evaluate evidence during an inductive process.”
11 a.m. Panel: What Can Technology Do Today and in 2020?
Andrew McAfee of MIT asks the question: Why can’t computers do things that a two year old child can do?
Rod Brooks, AI and robotics entrepreneur “There is progress but it’s in narrow subfields. But it can do great things. Google cars are an example: They don’t do a lot of things but they do a few things very well.”
Afred Spector of Google: Google Translate is another example. We’re up to 69 languages. We’re working on quality. “We want to get to the languages that are less spoken so all those populations will have access to the Web. We want to make the knowledge available to everybody.”
Another project: Making it possible so the machine automatically understands things so well that we can translate an image into a text description, or visa versa.
Brooks: He talks about the problem with manufacturing in the US. We keep going to high tech manufacturing, but that makes us too narrow. Not enough jobs created. He says we need to develop manufacturing that can employ a lot of people, by automating the low value pieces more but produce a wide variety of products. “The answer is in the masses–creating robots that people can interact with and use.”
David Ferrucci of IBM: “I see a future where computers can act as intelligent mediators that enable informed collaboration, for instance, between you and your healthcare team, so you can make better decisions about your treatment.”
Here’s McAfee talking about the coming capabilities of machines and their impact on jobs and job creation:
Panel: What Can Technology Do Today and in 2020?
Andrew McAfee of MIT asks the question: What would accelerate your work the most: data, computing power, or smart researchers?
Rod Brooks, MIT professor, and AI and robotics entrepreneur : We have enough computing power and data. “I have a bunch of smart PhDs, but you have to direct them in the right direction. You need the right reward structure for research.”
“We’d be better off if universities were smaller, had fewer people working form them, and focused on deep fundamental research. Let organizations like IBM do the applied research.”
Afred Spector of Google: “When we go to universities we’re surprised and disappointed that the faculties aren’t doing more high-risk research.”
Google sees a need for vastly more computing power. We need to do “deep learning.” It’s a new level of machine learning. Google has 5,000 PhD’s in computer science. “We need even more talent.”
“We need all three to get better.”
David Ferrucci of IBM: “Researchers, data and machines, in that order.”
12:20 p.m. Panel: How Will Technology Affect Productivity and Employment?
Erik Brynjolfson, MIT Sloan School of Management, asks: What does technology mean for technology and jobs?
David Autor, economics professor at MIT: There’s a long running debate. Does technology eliminate jobs? The stock answer is to call people who ask it Luddites.
We’ve seen incredibly rapid technology change over the past century and it eliminated a lot of farm jobs, but it created jobs elsewhere. “We’ve seen rising employment rates; and it raises productivity and incomes.”
However, there’s another side to this. Technology increases our efficiency but it can compete with workers and their skills. “Technology changes much faster than people can adapt.”
Middle-education and middle-skilled jobs are the vulnerable ones. That’s manufacturing jobs and administrative jobs.
“This creates real challenges. We should be worried. The set of opportunities are far more bifurcated than then used to be.”
“It’s leading to even more unequal distribution of wealth.”
Irving Wladawsky-Berger, former IBM executive and MIT lecturer: He says he has been focusing on technology based innovation in the service economy. So many of the new jobs are in services. About 80% of the service jobs are information-based jobs. Technology will be used more and more in this area. So these jobs will increasingly come under pressure, too.
This is another period of creative destruction.
New industries will be created that will create the mid-skilled and mid-education jobs. “I don’t know the answer”
“The top-down approaches to job creation aren’t working. We have to rely more on bottoms-up approaches—entrepreneurialism.”
Frank Levy, a labor economist at MIT: Keeps things in perspective. Everything we see today is colored by the recession. It doesn’t have a lot to do with technology—but with the collapse of the housing bubble.
At the same time, the middle-skill job problem is very real.
“Because of the recession, it’s going to be hard to get kids to get advanced education at the same time that the jobs that will come will require advanced educations.”
12:45 p.m. Panel: How Will Technology Affect Productivity and Employment?
Erik Brynjolfson, MIT Sloan School of Management, asks: How do we create new jobs for mid-skilled people?
Irving Wladawsky-Berger, former IBM executive and MIT lecturer: Cloud computing and other technologies can help entrepreneurs get started and build companies and hire people. So a lot of small companies will spring up—not the high tech companies but companies that take advantage of technology.
David Autor, economics professor at MIT: That’s good, but it won’t produce a lot of jobs. “Most people want to be employed. They want to work for somebody. If they have a choice, that’s what they do.”
Frank Levy, MIT: He calls for apprenticeships and case-based education to bring up the skills.
Wladawky-Berger: Germany has done a better job at creating the mid-skilled jobs.
Autor: The Germans have adapted more quickly than other developed economies. They brought up the skills and reduced wages for mid-skilled people, which made the people and the country more competitive.
There are decent middle-skilled job in health care, repairs, the trades. But all require post-high schools investments in skills. “But you can’t go to the Harvard of plumbing.”
Wladawsky-Berger: Can’t community colleges help fill the void.
Autor: “They’d like to, but their funding is being cut by states and communities.”
1 p.m. Panel: How Will Technology Affect Productivity and Employment?
Question from the audience: “I’m worried about how we communicate about the new capabilities of machines and their impact on jobs. Will people react against it?”
Irving Wladawsky-Berger, former IBM executive and MIT lecturer: There’s a consensus that just as we transitioned from the agricultural age to the industrial age, and literacy and education went up, in today’s world you need the next level of education. You need information-based literacy, and teamwork literacy. People who learn to use these tools can make a good living. If we can communicate that we’ll be okay.
Frank Levy, MIT: We should be clear about what machines can do and what they can’t do, and not talk a lot about the “singularity”—the point in the future when machines will be able to truly think.
“Just tell it straight in terms of what we know now. Don’t try to scare people.”
Here’s Wladawsky-Berger talking about the future of job creation:
1:25 p.m. Remarks from Martin Fleming, chief economist at IBM:
There have been five waves of technological change over the past three centuries. With each wave, the new technologies fundamentally altered the way business—and work—was done. “The business changes because the technology makes it possible to do so.”
With IBM’s Watson, for example, you enable the democratization of clinical decision making. The practice of healthcare can be fundamentally changed. Evidence-based medicine is made possible.
Each wave was also accompanied by an economic crash, typically at the time when the new technology is impacting the old ways of doing things but has not yet produced all of the productivity gains that are coming on a mass scale.
We’re now entering into a period where the economy is beginning to open up opportunities for the deployment of significant new innovations. “Radical new technologies will be deployed. New industries will be created.”
“This is about the transformation of the economy.”
At Harvard Business School now…
3:30 p.m. David Ferrucci, head of IBM’s Watson project, talks about how Watson came to be and where the technology is going.
Watson was a grand challenge aimed at driving important scientific advances. It gets people to think about the implications of technology–where is it today and where might it go.
Watson beat former grand champions at TV’s Jeopardy! quiz show.
He points out how much more difficult it is for a computer to have a conversation with a person than it is to play chess–a previous grand challenge that IBM took on in the 1990s when one of its machines beat the best chess player in the world. That’s because, in conversation, “context defines meaning.”
It’s a very difficult problem. Computers can’t relate words to experience.
“Jeopardy! helped us push the kind of technology that interprets natural language to determine meaning.”
4:00 p.m. David Ferrucci, the father of Watson, talks about how the technology came about and where it’s going.
We did 8000 experiments to develop Watson’s capability. There were lots of growing pains. Example: New York Times Headlines: An exclamation point was warranted for the “end of” this” in 1918. Watson’s answer: “a sentence.”
The project took 4 years.
Originally, running on a single PC, it took two hours for Watson to answer a single question. “The producers insisted that that would make for a boring game.” So they scaled the machine up to a 2880-core computing system.
The important thing about Watson is that it collects evidence and builds confidence in an ansers. when we think about applying it to medicine or law, we don’t think it’s providing the answer but providing useful suggestions–based on an evidence profile.
We’ll give a human decision maker the top answers to a question and the evidence and analysis that led Watson to those answers. This is about empowering the decision maker.
Here’s Ferrucci talking about how the software program can help transform the healthcare industry.