Category: Uncategorized

I have an article at Fortune.com/CNN Money on the possible future impact of technologies like Watson on the job market:

Below are links to a series of videos on IBM’s website that describe Watson and how it works. I highly recommend watching these, while keeping in mind the possible implications for job automation.

One thing I have tried to emphasize here and in my book, The Lights in the Tunnel, is that a true thinking machine is really NOT required in order to automate a large proportion of knowledge-based jobs: all that’s required is a sufficiently powerful specialized (or narrow) artificial intelligence application. I think Watson is a pretty strong indication that things are trending in that direction.

Watson is not based on custom databases of information, but rather uses machine learning techniques to parse and analyze information in “as is” form — in other words, books, magazines, encyclopedias, etc. (It’s worth noting that the real purpose behind Google’s massive book scanning project is reportedly to someday make the information accessible to an AI application, so IBM is certainly not the only company working on this problem).

Watson uses a massively parallel architecture that launches multiple (perhaps thousands) of queries using different algorithms simultaneously. The best answer—and a confidence level—is then selected using experience acquired via machine learning (in other words, by practicing with sample questions).

Here’s a quote from the IBM website: “Watson’s first test will be on Jeopardy!, but the real test will be applying the underlying data management and analytics technology across different industries.”

Here are the videos:

What is Watson?

Expert Interviews

Building Watson (Includes explanations of how the technology works and practice matches showing improvement over time.)

AP reports that IBM is testing a new computer called “Watson” (after IBM’s founder) that will be a contestant on the “Jeopardy!” TV game show. The computer has already won a practice round:

The system, which is powered by 10 racks of IBM servers running the Linux operating system and has 15 terabytes of random-access memory, or RAM, has been in the works for four years. It has digested encyclopedias, dictionaries, books, news, movie scripts and more, IBM says. It has access to the equivalent of 200 million pages of content. It is not connected to the Internet, so it does not do Web searches.

The company says Watson rivals a human’s ability to answer questions posed in natural language — rather than computer code — with speed, accuracy and confidence. Unlike earlier computers, it can deal with “Jeopardy’s!” subtleties of language, including puns and riddles.

IBM scientist David Ferrucci, a leader of the Watson team, said last month that using “Jeopardy!” to develop the computer system “is going to drive the technology in the right directions.”

“It asks all kinds of things,” he said. “It has the confidence aspect — don’t answer if you don’t think you’re right. You also have to do it really quickly.”

Watson is reminiscent of IBM’s famous Deep Blue computer, which defeated chess champion Garry Kasparov in 1997. But while chess is well-defined and mathematical, “Jeopardy!” presents a more open-ended challenge.

My guess is that technologies like this will find their way into “the cloud” and that they’ll eventually be deployed to automate a great many tasks and jobs (call centers, customer service, tech support) throughout organizations (see my previous post on Google, Cloud Computing and Machine Learning).  According to the article, the episode with the computer will air on Feb 14-16.

Update

Singularity Hub has a video from Engadget that shows Watson in action (highly recommended).  Watson seems spookily reminiscient of HAL.

Here’s also an older video from YouTube:

 

An economist argues against the views I’ve  expressed here. Perhaps mainstream economic thought on this issue will begin to shift around the time that we start seeing articles entitled “Rise of the Robo Economists.”

Many lower wage jobs are currently protected from automation primarily because technology cannot yet replicate a person’s ability to recognize and manipulate objects. That is changing, and once affordable  robotic technologies begin to outperform people in tasks that require sophisticated hand-eye coordination combined with moderate-level decision making skills, millions of jobs will be at high risk.

Here are a couple of videos showing these technologies in action, along with articles at Singularity Hub:

Color Sorting Robot:

Robot to Pick Only Ripe Strawberries:

The thing to keep in mind is that the capabilities of technologies like these will certainly accelerate over the coming decades. The degree of progress we see over the next ten years will be dramatically more than what we saw over the last decade. Moore’s Law, for example, would imply a factor of 32 increase (5 doublings) in general capability.  There will also be dramatic cost reductions.

As soon as these technologies find their way into applications that offer a profitable economic trade-off vs. employing workers, they will be widely deployed. Competitive pressures will require this. The primary danger to the US economy will probably be when jobs in the service sector become heavily threatened.  In countries like China, these technologies are likely to accelerate manufacturing automation and ultimately make it difficult to maintain employment as workers continue to migrate from rural areas to manufacturing centers.

Note that both videos are Japanese. Japan is on the leading edge of robotic and automation technologies. So far the unemployment rate in Japan has remained relatively low (although there are many underemployed workers), but this is largely due to regulation and social/cultural attitudes.  The story in the United States— where businesses have complete freedom to slash their workforces—is likely to be quite different.

Yves Smith at Naked Capitalism has a good post on the declining economic value of college, and the looming danger of massive student loan defaults.  Shockingly, a full 50% of college graduates are winding up underemployed:

Take note: half the recently-minted college grads are in jobs that do not require a college degree.

Now if these graduates were going to college for the mere love of learning, and didn’t mind working at Home Depot because they could work on a novel in their garret, this picture might not be quite as terrible as it looks. But I sincerely doubt that anyone in the US goes to college to become a working class intellectual.

But the economic (as opposed to social and personal) value of higher education is exaggerated. The widely-touted College Board claim that lifetime earnings for college grad outpace those of mere high school grads by $800,000 does not stand up to scrutiny. The author of the 2007 study which the College Board relied upon disclaims that estimate and says $450,000 is a better figure. Mark Schneider, a vice president of the American Institutes for Research, who used actual earnings data of graduates ten years after college, and allowed for other factors such as taxes, pegged the difference at $280,000.

And these estimates are averages. Students who are drawn to fields such as architecture, which require advanced education but are not terribly well paid, will fare less well.

Read the whole post here.

Also, check out this September 2009 post on BusinessWeek‘s blog, which includes the “appalling” graph below:

Unfortunately, I think there is every reason to believe that the problem will get worse.  Technology will increasingly be leveraged to automate the knowledge worker jobs that are often taken by new college graduates, and this is likely to hit especially hard at the entry-level.

I also think the future impact of offshoring is underestimated. We cannot escape the reality that  intellectual capability within the population is subject to a normal distribution. This implies that, collectively, India and China have more smart people…than the United States has people.  In the future, technology will make it even easier for the millions of people on the right flank of Asia’s bell curve to compete directly with Americans for knowledge-based jobs.

Here is a section from The Lights in the Tunnel in which I discuss the future of college education:

Nearly everyone agrees that a college degree is generally a ticket to a brighter future. In the United States in 2006, the average worker with a bachelor’s degree earned $56,788, while the average high school graduate earned a little more than half this amount, or $31,071. Workers with graduate or professional degrees earned a still higher average salary of $82,320. While the primary motive for the majority of individuals to pursue advanced education is almost certainly economic, we would all agree that education also conveys many other benefits both to the individual and to society as a whole. A person with more education seems likely to enjoy a generally richer existence, to have an interest in a greater variety of issues and is perhaps also more likely to be focused on continuing personal and professional growth. A more educated society is generally a more civil society with a lower crime rate. An educated person is likely to hang out in the library—rather than on street corners.

The unfortunate reality, however, is that the college dream is likely at some point to collide with the trends in offshoring and automation that we have been discussing in this chapter. The fact is that college graduates very often become knowledge workers. As we have seen, these jobs—and in particular more routine or entry level jobs—are at very high risk. The danger is that as these trends accelerate, a college degree will be seen increasingly not as a ticket to a prosperous future, but as a ticket to a job that will very likely vaporize. At some point in the future, the high cost of a college education, together with diminishing prospects for college graduates, is likely to begin having a negative impact on college enrollment. This will be especially true of students coming from more modest backgrounds, but it will have impact at all levels of society.

This is, obviously, a very unconventional view. Most economists and others who study such trends would probably strongly argue exactly the opposite case: that in the future, a college degree will be increasingly valuable and there will be strong demand for well-educated workers.

This is essentially the “skill premium” argument—the idea that technology is creating jobs for highly skilled workers even as it destroys opportunities for the unskilled. I think the evidence clearly shows that this has indeed been the case over the past couple of decades, but I do not think it can continue indefinitely. The reason is simple: machines and computers are advancing in capability and will increasingly invade the realm of the highly educated. We’ll likely see evidence of this at some point in the form of diminished opportunity and unemployment among recent graduates and also among older college-educated workers who lose jobs and are unable to find comparable positions.

We may not see an actual closing of the gap in average pay for college v. non-college graduates because opportunities for workers of all skill levels are likely to be in decline. I am not suggesting that high school graduates who would have otherwise gone to college will chose to remain completely unskilled, but I do think there is likely to be a migration toward relatively skilled blue collar jobs if there is a perception that these occupations offer more security.

As new high school graduates begin to shy away from a course leading to knowledge worker jobs, they will increasingly turn to the trades. As we have seen, jobs for people like auto mechanics, truck drivers, plumbers and so forth are among the most difficult to automate. The result may well be intense competition for these relatively “safe” jobs. As high school graduates who might previously have been college-bound compete instead for trade jobs, they will, of course, end up displacing less academically inclined people who may have been a better fit for those jobs. That will leave even fewer options for a large number of workers.

We see evidence of this trend already in the daily news. Newspapers routinely report that people are specifically seeking jobs that can’t be offshored. Much is made of new “green collar jobs that cannot be outsourced.” While this is certainly a desirable development, we have to acknowledge that the bulk of these jobs are going to involve installing solar panels, wind turbines and so forth. They are trade jobs; not jobs for college graduates.

The cost to society of such a turn away from education would be enormous. It would damage the hopes, dreams and expectations of our children and potentially rob them of things that we ourselves have come to take for granted. Those workers whose prospects were diminished by a new influx of more “book smart” competitors would become even more dispirited and more likely to turn to crime or other undesirable alternatives. This hash new reality would fall most heavily on people in disadvantaged sectors of the population. Finally, and perhaps most chillingly, a trend away from college would rob us of talent we may well need in the future.

There’s a very good article in the New York Post on the polarization of the job market and the disappearing middle class:

I’ve been blogging here extensively about the likelihood that various forms of automation will eventually create significant technological unemployment. Advanced robotics will certainly play an important role in that once it becomes cost-effective to replace even low wage service workers with machines.

I find it interesting that very few other people seem to be particularly concerned about this issue. Here are two recent articles that seem quite enthusiastic about the robotic future, but give no thought at all to the possibility that robots might someday contribute significantly to unemployment:

Scientific American: “Robot be Good” assumes we’ll soon have autonomous robots interacting directly with humans in environments like nursing homes for the elderly. The article concerns itself with the problem of ethical behaviour among robots.

Technology Review: “Why Japanese Love Robots” looks at how Japanese culture tends to favor robots and see them as helpful and friendly, while Americans are more likely to see them as menacing.

A Google search will bring up plenty of other articles on the coming robotics revolution, and you’ll find quite a few similarities:

• Unless you’ve happened upon one of the relatively few articles that deal specifically with robots creating unemployment, you’ll find little or no mention of this issue. The negative side of robotics nearly always involves physical threats: robots that will hurt people, take over, get out of control, etc.
• Personal robots will eventually do all kinds of useful things around the house. Someday, we’ll all have robotic personal assistants. This is often touted as a huge new consumer market.
• Robots will do some jobs but they will invariably be dangerous jobs (police and military) or jobs that no one wants or where there are worker shortages due to low wages (caring for the elderly).

The thing is that for a robot to autonomously run around the house doing a variety of tasks requires a very sophisticated level of technology. If that technology is developed and becomes affordable then it will certainly make its way into a variety of commercial applications—in fact, it may well be deployed there first.

It seems to me that if we have affordable personal robots that are actually capable of doing anything useful, then that technology implies that millions of jobs will be at risk in areas like:

• stocking shelves in supermarkets and other retail stores
• moving materials in stores and warehouses
• providing security in a variety of settings

If the technology is there and if it is cost-effective, then businesses are not going to pass up the opportunity to deploy it.  The standard response from economists is that we don’t need to worry because new jobs will be created in other areas. I really wonder what kinds of jobs the economy would create for these workers.

SingularityHub recently reported that a Silicon Valley area hospital is announcing layoffs at the same time it begins to employ robots: 

El Camino Hospital in Silicon Valley is looking to cut expenses, so they’ve invested in 19 Aethon TUG robots. These smart carts can haul supplies around the hospital, making deliveries and pickups at a fraction of the costs of human workers. El Camino recently announced that it would further be cutting costs by firing up to 140 workers from its two facilities in Los Gatos and Mountain View.

It should be noted that most of the layoffs are probably not directly related to the decision to use robots. Nonetheless, I think this shows that even healthcare—the one field on which nearly everyone pins hopes for significant job growth—is not immune to automation. It also demonstrates that the economic tradeoff between robots and even relatively low wage/low skill jobs is beginning to tip in favor of the machines.

Economists often speak of “polarization” in the job market. The belief is that technology has primary impacted middle skill jobs, leaving plenty of high wage opportunities for the well-educated as well as lots of low skill service jobs with very low wages. As I’ve been arguing here, I think this is what has been happening so far—but it will not continue to be true indefinitely.  Automation will push up into the high wage areas via technologies like narrow artificial intelligence/expert systems, while it penetrates lower skill job sectors with more affordable robotic technologies. In general, I think economists have a serious problem with analyzing past data, determining a trend, and then assuming it will continue basically forever.  Technologies change rapidly.

In spite of this news, I think that healthcare will certainly remain one of the most promising areas for future employment. However, the same cannot be said for other lower skill jobs in the commercial arena.  While delivering medical supplies in a hospital may not be an especially high-skill job, it is certainly not an unimportant job. If robots can be trusted to autonomously navigate crowded hospital corridors to deliver medical supplies in a timely fashion, then they can and will be used in other commercial settings like warehouses and retail stores.  

In fact, CNET News reported that Wal-Mart was already looking into the use of inventory robots back in 2005. These robots would have prowled the aisles at night taking complete store inventories—a job that is, of course, currently done by workers. One has to wonder how long it will be until Wal-Mart and its competitors begin to look seriously at robots in a number of work areas.  Jobs involving shelf-stocking, inventory control, and materials moving are all likely to be susceptible at some point.

The scary thing is that for many workers these are really the jobs of last resort. This is where people who lose good jobs in manufacturing or other areas often end up. What options will these people have if even these jobs are someday much less plentiful?

Google recently announced a new machine learning engine that it will make available to software developers. Machine learning is a form of artificial intelligence (AI) in which an application can learn from processing real data and become more proficient over time. By making the tool available, Google will enable businesses and entrepreneurs to use AI in wide range of new applications.

In the coming years, artificial intelligence is going start showing up in more and more places. AI will be incorporated into productivity applications and into the enterprise software used by large companies. I’m not talking about science-fiction level general artificial intelligence (“Open the pod bay doors, HAL”), but rather specialized or narrow forms of AI. Narrow AI applications can already land jet aircraft and beat virtually any human being in a game of chess. In the near future, they will be able to do far more.

Google’s new AI tool is being offered as part of the company’s cloud computing strategy. Cloud computing is a new model in which computer hardware resources as well as application software are made available on an as-needed basis, in much the same way that utilities like electric power are provided.

The thing you should know about cloud computing is that it tends to concentrate information, power and income. The information technology resources of thousands of businesses and organizations will increasingly “migrate into the cloud.” One immediate result of this is increased concentration and automation of jobs. Information technology workers are already seeing significant job losses as a result of the move toward cloud computing.

Once artificial intelligence becomes integrated into the cloud, the effect will quickly be felt by far more than just IT professionals. Anyone with a knowledge-based job will be highly susceptible. Organizations will get flatter as more middle managers are eliminated. It’s also quite possible that AI tools will be used to amplify the capabilities of low wage off-shore workers—allowing them to move up the value chain and compete directly with professionals who have high skill and experience levels.

And AI-enabled cloud computing isn’t just about direct job automation: it will also allow larger organizations to leverage economies of scale, perhaps as never before. Companies like Wal-Mart and the big box retailers will gain, while smaller businesses continue to lose. Sophisticated applications will make it easier to run larger, more complex organizations with fewer people, and that will be an important enabler of corporate consolidations. Low interest rates are already driving a new wave of merger activity on Wall Street, and you can be sure that mass layoffs will follow.

The point here is that technologies like cloud computing and narrow AI are going to result in less opportunity for most workers—while concentrating income and power in the hands of the few (as if that is a new story). Corporations will need fewer managers and knowledge workers, while at the same time many of the small business opportunities that have traditionally led to middle class, or even upper middle class, success will continue to evaporate. The demise of the blue-collar middle class is already pretty much a done deal. College educated white-collar workers—even those with relatively high incomes—are next in line.

The broader trends that are driving income concentration and the destruction of the middle class—globalization, advancing technology, supply side economics—are of, course, not Google’s fault. However, within the IT field Google is becoming a poster child for the concentration of wealth and power: and it is making important contributions that will accelerate the process.

But here’s the rub: Google’s current business model is almost entirely dependent on a world in which income—and therefore purchasing power—is at least somewhat reasonably distributed. Google’s revenue comes primarily from its AdWords program, which allows businesses of all sizes to place highly targeted online advertisements.

AdWords is an enormously successful money machine, and it works because businesses know that among Google’s huge number of users there will be a significant slice of traffic with a high interest in a particular product or service. Here’s the thing though: AdWords advertisers aren’t interested in reaching web surfers. They want customers—customers with discretionary income.

In the long run, as income becomes more and more concentrated—as more average people in the population find themselves unemployed or forced to take lower wage jobs—the businesses that advertise on Google are inevitably going to see more surfers and fewer paying customers. As that happens, they will drop out of the program entirely, or they will be willing to pay less for the ads, and Google’s revenue will have to decline. If the economy continues on its seemingly relentless path toward increased concentration of income and consumption, then at some point, Google’s advertising model will no longer be an especially effective way to reach the few people who still have money to spend.

Of course, if the entire economy continues on that path, then the viability of Google’s business model may be the least or our worries. We already have BMW owners sleeping in their cars, and upper crust New Yorkers worrying about civil unrest or even revolution. Watch out.

Note: For more on AI, unemployment, the concentration of income, and the impact on Google’s business model, see the free PDF of  The Lights in the Tunnel: Automation, Accelerating Technology and the Economy of the Future (pages 67-73, 81-84, and 180-183).