Thursday, June 20, 2019

Is Hydrogen the Storage and Carrying Technology for Carbon-Free Energy?

Fossil fuels store energy until they are burned. Solar and wind power generate electricity, but don't store it. As a result, they are intermittent sources of electricity, requiring back-up generation capacity that is typically still supplied by fossil fuel. Could hydrogen become a way of storing energy from renewable power sources? The International Energy Agency, in a report on The Future of Hydrogen, describes what would be needed to make this happen (June 14, 2019, accessing report requires free registration). 

At present, hydrogen is mostly produced from fossil fuels, and used only in fairly narrow applications. The report notes: 
Hydrogen is almost entirely supplied from natural gas and coal today. Hydrogen is
already with us at industrial scale all around the world, but its production is responsible
for annual CO2 emissions equivalent to those of Indonesia and the United Kingdom
combined. Harnessing this existing scale on the way to a clean energy future requires
both the capture of CO2 from hydrogen production from fossil fuels and greater supplies
of hydrogen from clean electricity. ... Today, hydrogen is used mostly in oil refining and
for the production of fertilisers. For it to make a significant contribution to clean energy
transitions, it also needs to be adopted in sectors where it is almost completely absent at the moment, such as transport, buildings and power generation.
So making the shift to a more use of hydrogen will require a lot of change. The overall vision would be that some of the sources of renewable energy like solar, wind, and hydro could become hydrogen farms, separating hydrogen from water. Instead of building electrical power lines from these facilities, there would need to be a system for storing and transporting the hydrogen they produce, similar to all the ways that natural gas (and liquified natural gas) are transported today. There would need to be early adopters of hydrogen fuel cell vehicles, perhaps focusing first on organizations with fleets of vehicles like trucks or buses. Building would need to be designed or retrofitted to use hydrogen as a source for heating, cooling, and electricity.

These changes are substantial!  As the report notes, there have been a few previous moments when hydrogen was widely discussed as a method of storing and carrying energy: the 1970s, the 1990s, and the early 2000s. When oil prices declined after about 2011, government R&D spending on hydrogen also declined.
However, if technological progress can continue to drive down the costs of hydrogen, the potential benefits are also substantial, because hydrogen technology offers a way in which renewable power that now generates electricity could instead be used to address sectors of the economy where electricity has proven to have practical drawbacks. The report notes:
The increased focus on reducing [greenhouse gas] emissions to near zero by mid-century has brought into sharp relief the challenge of tackling hard-to-abate emissions sources. These emissions are in sectors and applications for which electricity is not currently the form of energy at the point of end use, and for which direct electricity-based solutions come with high costs or technical drawbacks. Four-fifths of total final energy demand by end users today is for carbon-containing fuels, not electricity. In addition, much of the raw material for chemicals and other products contains carbon today and generate CO2 emissions during their processing. Hard-to-abate emissions sources include aviation, shipping, iron and steel production, chemicals manufacture, high-temperature industrial heat, long-distance and long-haul road transport and, especially in dense urban environments or off-grid, heat for buildings. Rapid technological transformations in these sectors have made limited progress in the face of the costs of low-carbon options, their infrastructure needs, the challenges they pose to established supply chains, and ingrained habits. ... As a low-carbon chemical energy carrier, hydrogen is a leading option for reducing these hard-to-abate emissions because it can be stored, combusted and combined in chemical reactions in ways that  are similar to natural gas, oil and coal.

Tuesday, June 18, 2019

The Economic Value of Household Production: 1965-2017

Gross domestic product is not the total amount of output produced; instead, it is a a measure of what is bought and sold in markets. Pretty much every intro class in economics will point out to students that when I clean my own house, cook my own meals, look after my children, or or mow my lawn, that "household production" doesn't show up in GDP. But if I hire someone to do household production tasks, then that output gets counted as part of GDP.

For a number of situations where the limitations of GDP are obvious, the US Bureau of Economic Analysis publishes "satellite" accounts, where it calculates what a different and broader measure of economic output would look like. In this spirit, Danit Kanal and Joseph Ted Kornegay have written "Accounting for Household Production in the National Accounts: An Update, 1965–2017," in the June 2019 Survey of Current Business.

The overall approach is to look at data on time use in household production, estimate the cost of hiring that time in the market, and then add this output to the standard conventional measure of GDP. They write:
The largest impact when including household production in GDP stems from the inclusion of nonmarket services. Nonmarket services measure the value of time spent on home production tasks. ... To compute household production, we first aggregated household production hours across seven categories: housework, cooking, odd jobs, gardening, shopping, child care, and domestic travel. The value of nonmarket services is the product of the wage rate of general-purpose domestic workers and the number of hours worked. This method assumes a market-cost approach to valuing nonmarket household services. ... BEA's current GDP measure treats consumer purchases of durable goods as consumption. In contrast, this satellite account treats such purchases as investment and adds the services of consumer durables to personal consumption expenditures. 
Some interesting fact patterns emerge from thinking about economic output in this way:

When this calculation is carried out for 1965, the revised GDP with household production included is 37% higher than the conventional measure. For 2017, the revised GDP with household production included is 23% higher.

Thus, if you are someone who sometimes uses per capita GDP as a quick-and-dirty measure for social well-being (a sin of which I've been guilty now and again), taking household production into account shows that the US standard of living is higher than the conventional measurement.

Why does adding a value for household production have a smaller effect now than a half-century ago? "Household production has declined in significance over time as more women engage in market work." In particular, the number of hours spent in household production by nonemployed women has declined substantially.

The growth rate for total economic output is slower. Measured in nominal dollars, the growth rate of traditional GDP is 6.5% per year from 1965 to 2017, while the annual nominal growth rate of output falls to 6.3% per year with household production in included. In effect, the declining hours spent on household production mean that the relative size for this part of the economy is shrinking.

As usual with economic statistics, any one number is going to have serious limitations, and so looking at a variety of interrelated measures will provide a more in-depth picture. Here, the authors are just presenting fact patterns, not hypothesizing about underlying causes. But presumably there are a variety of changes behind these patterns, like fewer children in the average household, the spread of household technologies like the dishwasher and the microwave, and household which choose to purchase some services (meals eaten out, house-cleaning, yard work) rather than producing it themselves. 

For my blog posts on a couple of previous reports from the Survey of Current Business on this topic, see: 


Monday, June 17, 2019

Interview with Rachel Glennerster: Development and Aid

Rachel Glennester has her finger on the pulse of both development economicsresearch and real-world development policy. She was the long-time Executive Director of the Abdul Latif Jameel Poverty Action Lab based at MIT, and now has taken a position as Chief Economist of the primary UK agency for developiment aid, the Department for International Development. She was interviewed by Robert Wiblin and Nathan Labenz at the 80,000 Hours website. You can listen to the 90-minute podcast or read a transcript at "A year’s worth of education for under a dollar and other‘ best buys’ in development, from the UK aid agency’s Chief Economist," by Robert Wiblin and Keiran Harris (December 20, 2018).

The importance of investigating basic descriptions of situations in development research
A lot of development programs just fail because they’re trying to solve a problem that doesn’t exist. ... The first really important thing you’ve got to do is really understand what the issue is in any given area. If we’re worried about girls not going to school because of menstruation, well, let’s start by finding out whether they actually don’t go to school more when they’re menstruating. That’s a really basic, obvious thing. But we actually need more work on that kind of understanding the context, understanding the problem, is really important first step. ...
Here is an example. I did a project looking at how to improve immunization rates in India ... Only 3% of kids in this part of India were getting fully immunized. Given that immunization is one of the most effective things that you could do, that rate is just appallingly low. There were a number of theories about why that could be, and a lot of people said ... they don’t trust the formal health system. There was also a question of, so the clinics are often closed, so is that the problem? ... Is it nurse absenteeism that’s the problem? Or is it just a behavioral economics thing that you’re happy to get your kid immunized, but you’ll do it tomorrow? ...
What we saw in the data is a lot of people got their kid immunized with one immunization, but they failed to persist to the end of the schedule. Which already, that’s just descriptive data and it starts to tell you, it’s not that they distrust the system or that they think that immunizations are evil, because they’re getting their kid one immunization. It’s more question of persistence. Now, fixing the supply problem increased the number of people getting the first shot, and the second shot, but again, it failed to fix this persistence problem. Where the incentive effect worked, was it helped people persist to the end. That tells you that one of the big problems was this persistence problem. It tells you a lot about why immunization isn’t happening.
How Randomized Control Trials Test Specific Interventions, But Reveal Bigger Lessons
I think actually RCTs should not be seen as looking at testing this specific program, they should be seen as testing big questions that can then influence policy. For example, you might test a specific project on education. A lot of the work on education has suggested that the most effective thing we can do in education is to focus on the learning within the classroom. It’s not about more money, it’s not about more textbooks, it’s not about … And that’s what governments spend their money on. They spend it on teachers and textbooks, mainly teachers. But more teachers doesn’t actually improve learning. More textbooks doesn’t improve learning. But that’s what the Indian government is spending their money on. ...
If you look at the data, just descriptive data, again, the power of descriptive data … within an average Indian classroom in 9th grade, none of the kids are even close to the 9th grade curriculum. They’re testing at somewhere between 2nd grade and 6th grade. No wonder they’re not learning very much, because the teacher, the only thing that a teacher has to do by law in India is complete the curriculum, even if the kids have no idea what they’re talking about. So yes, you have RCTs testing very specific interventions; all of the ones that worked were ones that got the teaching down from the 9th grade curriculum to a level that the kids could actually understand. Now the lesson from that, the big lesson for the Indian government if they were ever to agree to this, is change your curriculum. That’s the biggest thing that you could do. Reform the curriculum and make it more appropriate to what children are doing. So yes, you’re testing little things, but you’re coming out with big answers.
Helping Countries When They have Committed to Structural Change
At DFID, we have shifted a lot of emphasis relatively recently into trying to do more on economic transformation, under the recognition that the biggest reductions in poverty as you say, have come from big transformations in economic policy. So the big opening up of India and China towards more market-oriented economies … And I’m not saying markets solve everything; they absolutely don’t, but when you’ve got a system as screwed up as Communist China, making prices have some influence moves you an awful long way, and can really help transform the economy. And the same happened in India, and you saw massive reductions in poverty, by just a move towards a slightly more sensible economic policy.
When I was recently doing my ranking of what are most effective things that DFID could do, we were saying, “Well, if there are cases of countries that are as screwed up as China, helping them move to a more effective economic management, that’s gotta be the most effective thing that we could for poverty. You can’t do that as an outside donor, unless someone’s willing to do it. So where you see … I would say Ethiopia at the moment is going through a tremendous reform, and we really ought to be focusing attention, and helping Ethiopia in that transition. Tremendous potential, because they’re absolutely fundamentally changing policy there in ways that could be really beneficial to the poor. So jump on those opportunities, but you can’t really make them happen. It’s something that the developing country themselves has to decide to do, then help them as much as you can.
Then there’s the question of what do you do to promote economic development in countries that aren’t going through this fundamental reform process. You can nudge them a bit in the right way ... But we don’t always have all the tools that we need to make economic transformation happen.
What would be one piece of economic policy advice for India? 
Robert Wiblin: Yeah. If you could advise Narendra Modi, the PM of India, on one policy issue, hopefully get him to take you quite seriously, what do you think you would talk about, in that meeting? ...
Rachel Glennerster: I would try and persuade him to put in place markets for carbon.
Robert Wiblin: Oh, interesting, really? Okay. Explain that, sorry. I didn’t expect that.
Rachel Glennerster: Right. For a couple of reasons. So, one is just climate change is going to have huge impacts on the poor, and India is a big emitter of carbon. I firmly believe that, if you get prices right, there’s lots of things that people would do differently, if the price … That are reasonably cheap, but we’ve so screwed up prices that they don’t have the incentive to do it. ... And the final point is the health impacts in India of burning coal are just extraordinary, unbelievable health costs of all those coal fire power- ...
Robert Wiblin: Yeah, I was listening to a radio program the other day that was saying that it was taking about 10 years off the average life, for people in some of these cities like Delhi or Mumbai. I was like, it’s like the equivalent of smoking cigarettes, maybe even more so, which is absolutely crazy.
Rachel Glennerster: Yes, it’s the equivalent of heavy smoking every day. And you think about kids, breathing that in.

Robert Wiblin: Yeah, the air pollution thing makes sort of sense, but wouldn’t you then perhaps want to put in a program that just taxes air pollution? Or, do you think that taxing coal comes pretty close to doing that, or taxing carbon in general, is pretty close to an air pollution tax?

Rachel Glennerster: I would also love to do stuff on pollution, but a lot of this is coming from coal, and obviously then you also have climate impacts. I’d have to work through … I haven’t gone through all the detailed numbers of how much of those particulates are coming from coal, and how much are coming from other things. But the double whammy of ...

Robert Wiblin: Both climate change and saving just very large numbers of lives.

Rachel Glennerster: Yes.

Sunday, June 16, 2019

Mortality Rate of Children Over the Last Two Millennia

The global mortality rate of infants and youth up to the age of 15, based on an average of many studies, was almost one-half (46.2%) for the two millennia up to about 1900. By 1950, it was 27%. By 2017, it had fallen to 4.6%,.

The global infant mortality rate for children under the age of 1, again based on the average of many studies, was more than one-quarter (26.9%) for the two millennia up to about 1900. By 1950, it was 16%. By 2017, it had fallen to 2.9%

Here's a figure showing the patterns from Max Roser at the "Our World in Data" website (June 11, 2019). Of course, you will need to expand the version here, or go to the other website, to see the details.

I'll put off all the arguments over reasons why this happened and what it means for public policy for another day. It's Father's Day today, I just want to take a few minutes and marvel at this fundamental change in what it means to be a parent in the 21st century, especially in a high-income country. My children were much less likely to die.

Saturday, June 15, 2019

Marijuana Policy: Choosing Between Disastrous or Unpalatable

Slowly and with considerable uncertainty, the United States is altering its marijuana laws. Mark A. R. Kleiman offers an overview of the state of play and the likely tradeoffs in "The Public-Health Case for Legalizing Marijuana" (National Affairs, Spring 2019).  He writes:
John Kenneth Galbraith once said that politics consists in choosing between the disastrous and the unpalatable. The case of cannabis, an illicit market with sales of almost $50 billion per year, and half a million annual arrests, is fairly disastrous and unlikely to get better. The unpalatable solution is clear: Congress should proceed at once to legalize the sale of cannabis — at least in states that choose to make it legal under state law — for recreational as well as "medical" use. ...

First, as a practical matter, cannabis prohibition is no longer enforceable. The black market is too large to successfully repress. The choice we now face is not whether to make cannabis available, but whether its production and use should be legal and overt or illegal and at least somewhat covert. Second, because cannabis is compact and therefore easy to smuggle, a state-by-state solution is unworkable in the long run. States with tighter restrictions or higher taxes on marijuana will be flooded with products from states with looser restrictions and lower taxes. The serious question is not whether to legalize cannabis, but how.
Kleiman offers an overview of the legal status of marijuana, and also makes some key points about the evolution of the market.

Marijuana is a cheap high, even at the current illegal price, and legalization is likely to make it cheaper.
Cannabis, even as an illegal drug, is a remarkably cost-effective intoxicant, far cheaper than alcohol. For example, in New York City, where cannabis is still illegal, a gram of fairly high-potency material (say, 15% THC by weight) goes for about $10. A user can therefore obtain 150 milligrams of THC for $10, paying about 7 cents per milligram. Getting stoned generally requires around 10 milligrams of THC to reach the user's bloodstream, but the smoking process isn't very efficient; about half the THC in the plant gets burned up in the smoking process or is exhaled before it has been absorbed by the lungs. So a user would need about 20 milligrams of THC in plant material to get stoned, or a little less than $1.50 worth. For a user without an established tolerance, intoxication typically lasts about three hours. That works out to about 50 cents per stoned hour. ... So it costs a typical man drinking beer about $4 to get drunk — typically for a couple of hours — and staying drunk costs an additional $1 per hour. That's at least double the price per hour stoned offered by the illicit cannabis market.
For a number of users, marijuana use has adverse health effects.
Over the past quarter-century, the population of "current" (past-month) users has more than doubled (to 22 million) and the fraction of those users who report daily or near-daily use has more than tripled (to about 35%). Those daily or near-daily users account for about 80% of the total cannabis consumed. Between a third and a half of them report the symptoms of Cannabis Use Disorder: They're using more, or more frequently, than they intend to; they've tried to cut back or quit and failed; cannabis use is interfering with their other interests and responsibilities; and it's causing conflict with people they care about. ... Frequent users report using about 1.5 grams (equivalent to three or four joints) per day of use. With increasing prevalence, increasing frequency, and increasing potency, the total amount of THC consumed has likely increased about sixfold since the early 1970s.
A "state's rights" approach isn't likely to work well for marijuana.

Cannabis is simply too easy to smuggle across state lines. If cannabis is cheap anywhere, it will be available and fairly cheap everywhere. The same would be true if states were to adopt starkly different tax or regulatory policies, as these would likely generate large price differences in their respective legal cannabis markets.  ...
Even a very small difference would be more than enough to support a large illicit market, as the state and local taxation of tobacco has proven. New York State has fairly heavy tobacco taxes, and New York City adds a substantial local tax. Virginia, by contrast, taxes tobacco much more lightly. The result is that a pack of cigarettes that retails for under $5 in Virginia sells for $13 in New York City — a difference of $8 per pack. Due to this price gap in the legal tobacco market, more than half of all cigarettes sold in New York City are contraband: mostly genuine brand-name products purchased in bulk in Virginia and driven 250 miles to New York. There, they are resold for about $9 per pack by many of the same retailers who sell full-priced, legal cigarettes — mostly convenience stores in low-income neighborhoods. ... 
The same would be true for product regulation: If Massachusetts allows the sale of the solid concentrates used for the dangerous practice of "dabbing" (flash-vaporizing a hefty chunk of concentrate with a blowtorch in order to inhale a huge dose all at once), then for New York to try to forbid it would be a virtual invitation to smuggle. The states with the lowest taxes and the loosest regulations would wind up effectively dictating policy to the rest of the country.
What might be some general directions for federal-level marijuana legislation?
What would a public-health-friendly legalization program look like? The goals of such a policy would be the elimination or near-elimination of the illicit market and its replacement with a licit market delivering product of certified purity and known chemical composition, while minimizing the growth in heavy or hazardous use and use by minors. Its means would include taxation or minimum unit pricing (to prevent the otherwise inevitable collapse of cannabis prices); product regulation; and limits on marketing to prevent the cannabis industry from promoting the misuse of its product the way alcohol sellers encourage heavy drinking. ... Retail sales clerks — so-called "bud-tenders," now paid the minimum wage plus a sales commission, and thus given strong incentives to encourage overconsumption — could also be licensed, required to have extensive training in pharmacology and in preventing and recognizing Cannabis Use Disorder, and bound to a fiduciary duty to give advice in the interests of the consumer rather than with the goal of maximizing sales. ...Consumers could also be required, before being allowed to purchase cannabis, to pass a simple test showing they're aware of the risks and of basic precautions. More radically, they could be required to establish for themselves (and the stores could be required to enforce) a weekly or monthly purchase quota, as a nudge toward temperance. ... All of this will have to be done in the face of fierce opposition from the for-profit cannabis industry, if there is one.

For a previous post on the evolution of marijuana laws and markets, see "Canada Legalizes Marijuana: What's Up in Colorado and Oregon?" (October 22, 2018).

Friday, June 14, 2019

The "Right" and "Wrong" Kind of Artificial Intelligence for Labor Markets

Sometimes technology replaces existing jobs. Sometimes it create new jobs. Sometimes it does both at the same time. This raises an intriguing question: Do we need to view the effects of technology on jobs as a sort of tornado blowing through the labor market? Or could we come to understand why some technologies have bigger effects on creating jobs, or supplementing existing jobs, than on replacing job--and maybe even give greater encouragement to those kinds of technologies?

Ajay Agrawal, Joshua S. Gans, and Avi Goldfarb tackle the issue of how artificial intelligence technologies can have differing effects on jobs in "Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction" (Journal of Economic Perspectives, Spring 2019, 33 (2): 31-50). Perhaps someday "artificial intelligence" will be indistinguishable from human intelligence. But the authors argue that at present, most of the developments in AI are really about "machine learning," which involves using computing power to make more accurate predictions from data. They write (citations omitted):
The majority of recent achievements in artificial intelligence are the result of advances in machine learning, a branch of computational statistics. ... Machine learning does not represent an increase in artificial general intelligence of the kind that could substitute machines for all aspects of human cognition, but rather one particular aspect of intelligence: prediction. We define prediction in the statistical sense of using existing data to fill in missing information. As deep-learning pioneer Geoffrey Hinton said, “Take any old problem where you have to predict something and you have a lot of data, and deep learning is probably going to make it work better than the existing techniques.”
The authors are using "prediction" in a very broad sense: "As an input into decision-making under uncertainty, prediction is essential to many occupations, including service industries: teachers decide how to educate students, managers decide who to recruit and reward, and janitors decide how to deal with a given mess." Here are a few examples from their paper, some fairly well-known, others less so. 

AI and Brain Surgery
For example, ODS Medical developed a way of transforming brain surgery for cancer patients. Previously, a surgeon would remove a tumor and surrounding tissue based on previous imaging (say, an MRI scan). However, to be certain all cancerous tissue is removed, surgeons frequently end up removing more brain matter than necessary. The ODS Medical device, which resembles a connected pen-like camera, uses artificial intelligence to predict whether an area of brain tissue has cancer cells or not. Thus, while the operation is taking place, the surgeon can obtain an immediate recommendation as to whether a particular area should be removed. By predicting with more than 90 percent accuracy whether a cell is cancerous, the device enables the surgeon to reduce both type I errors (removing noncancerous tissue) and type II errors (leaving cancerous tissue). The effect is to augment the labor of brain surgeons. Put simply, given a prediction, human decision-makers can in some cases make more nuanced and improved choices. 
AI and Tax Law
Blue J Legal’s artificial intelligence scans tax law and decisions to provide firms with predictions of their tax liability. As one example, tax law is often ambiguous on how income should be classified. At one extreme, if someone trades securities multiple times per day and holds securities for a short time period, then the profits are likely to be classified as business income. In contrast, if trades are rare and assets are held for decades, then profits are likely to be classified by the courts as capital gains. Currently, a lawyer who takes on a case collects the specific facts, conducts research on past judicial decisions in similar cases, and makes predictions about the case at hand. Blue J Legal uses machine learning to predict the outcome of new fact scenarios in tax and employment law cases. In addition to a prediction, the software provides a “case finder” that identifies the most relevant cases that help generate the prediction.
AI and Office Cleaning
A&K Robotics takes existing, human-operated cleaning devices, retrofits them with sensors and a motor, and then trains a machine learning-based model using human operator data so the machine can eventually be operated autonomously. Artificial intelligence enables prediction of the correct path for the cleaning robot to take and also can adjust for unexpected surprises that appear in that path. Given these predictions, it is possible to prespecify what the cleaning robot should do in a wide range of predicted scenarios, and so the decisions and actions can be automated. If successful, the human operators will no longer be necessary. The company emphasizes how this will increase workplace productivity, reduce workplace injuries, and reduce costs.
AI and Bail Decisions
Judges make decisions about whether to grant bail and thus to allow the temporary release of an accused person awaiting trial, sometimes on the condition that a sum of money is lodged to guarantee their appearance in court. Kleinberg, Lakkaraju, Leskovec, Ludwig, and Mullainathan (2018) study the predictions that inform this decision ... Judges will continue to weigh the relative costs of errors, and in fact the US legal system requires human judges to decide. But artificial intelligence could enhance the productivity of judges. The main social gains here may not be in hours saved for judges as a group, but rather from the improvement in prediction accuracy. Police arrest more than 10 million people per year in the United States. Based on AIs trained on a large historical dataset to predict decisions and outcomes, the authors report simulations that show enhanced prediction quality could enable crime reductions up to 24.7 percent with no change in jailing rates or jailing rate reductions up to 41.9 percent with no increase in crime rates. In other words, if judicial output were measured in a quality-adjusted way, output and hence labor productivity could rise significantly. 
AI and Drug Discovery
A company called Atomwise uses artificial intelligence to enhance the drug discovery process. Traditionally, identifying molecules that could most efficiently bind with proteins for a given therapeutic target was largely based on educated guesses and, given the number of potential combinations, it was highly inefficient. Downstream experiments to test whether a molecule could be of use in a treatment often had to deal with a number of poor-quality candidate molecules. Atomwise automates the task of predicting which molecules have the most potential for exploration. Their software classifies foundational building blocks of organic chemistry and predicts the outcomes of real-world physical experiments. This makes the decision of which molecules to test more efficient. This increased efficiency, specifically enabling lower cost and higher accuracy decisions on which molecules to test, increases the returns to the downstream lab testing procedure that is conducted by humans. As a consequence, the demand for labor to conduct such testing is likely to increase. Furthermore, higher yield due to better prediction of which chemicals might work increases the number of humans needed in the downstream tasks of bringing these chemicals to market. In other words, automated prediction in drug discovery is leading to increased use of already-existing complementary tasks, performed by humans in downstream occupations.
Some of these examples fit the mental model that robots driven by AI are going to replace human workers. Other suggest that AI will make existing workers more productive. It has become common, when looking at effects of technology on labor markets, to focus on the idea that a given job  has a bunch of tasks. A new technology replace most or all of the tasks a certain job, that job may be eliminated. It the technology creates the need for a bunch of new tasks, brand-new job categories may be created. Or often, a new technology may just cause a job to evolve, by replacing some tasks and creating a need for other tasks to be carried out. 

These differing pathways suggest that it might be able to differentiate, at least to some extent, between uses of artificial intelligence that are especially likely to be efficiency-enhancing for existing workers and job-creating for others, and uses of artificial intelligence that are more likely to be job-replacing in a way that saves a little money for employers but doesn't have large efficiency gains. 

For example, an article in Axios described a discussion with James Manyika, director of the McKinsey Global Institute. Manyika notes that in doing AI research: "If your goal is human-level capability, you're increasing the probability that you're doing substitutive work ... If you were trying to solve this as an economic problem, you'd want to develop AI algorithms or machines that are as different from humans as possible." Manyika suggests a few examples of AI-based research that are less likely to replace human workers, because they don't mimic human capabilities: "augmented reality," "AI systems that can predict how proteins are folded, or how to route trucks better," and "robots that can see around corners, or register sounds outside our hearing range."

Daron Acemoglu and Pascual Restrepo tackle this question in a short nontechnical essay "The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand" (IZA Discussion Paper No. 12292, April 2019)
"Most AI researchers and economists studying its consequences view it as a way of automating yet more tasks. No doubt, AI has this capability, and most of its applications to date have been of this mold: e.g., image recognition, speech recognition, translation, accounting, recommendation systems, and customer support. But we do not need to accept that this as the primary way that AI can be and indeed ought to be used. ...
It is possible that the ecosystem around the most creative clusters in the United States, such as Silicon Valley, excessively rewards automation and pays insu¢ cient attention to other uses of frontier technologies. This may be partly because of the values and interests of leading researchers (consider for example the ethos of companies like Tesla that have ceaselessly tried to automate everything). It is also partly because the prevailing business model and vision of the large tech companies, which are the source of most of the resources going into AI, have focused on automation and removing the (fallible) human element from the production process. ...

All in all, even though we currently lack definitive evidence that research and corporate resources today are being directed towards the "wrong" kind of AI, the market for innovation gives no compelling reason to expect an efficient balance between different types of AI. If at this critical juncture insufficient attention is devoted to inventing and creating demand for, rather than just replacing, labor, that would be the "wrong" kind of AI from the social and economic point of view.
As one example, Acemoglu and Restrepo point out that individualized classroom teaching, enabled by AI, will not eliminate the need for teachers--and may even increase it. As they write: "Educational applications of AI would necessitate new, more flexible skills from teachers (beyond what is available and what is being invested in now), and they would need additional resources to hire more teachers to work with these new AI technologies (after all, that is the point of the new technology, to create new tasks and additional demand for teachers)." AI enabled-tools could go well beyond feeding students multiple-choice questions with continually adjusting levels of difficulty, and  provide a kind of feedback that is just different from what any classroom teacher can provide. 

Thursday, June 13, 2019

Some Snapshots of the Global Energy Situation

"Global primary energy grew by 2.9% in 2018 – the fastest growth seen since 2010. This occurred despite a backdrop of modest GDP growth and strengthening energy prices. At the same time, carbon emissions from energy use grew by 2.0%, again the fastest expansion for many years, with emissions increasing by around 0.6 gigatonnes. That’s roughly equivalent to the carbon emissions associated with increasing the number of passenger cars on the planet by a third." Spencer Dale offers these and other insights in his introduction to the the 2019 BP Statistical Review of World Energy. It's one of those books of charts and tables I try to check each year just to keep my personal perceptions of economic patterns connected to actual statistics.  Here are a few figures that jumped out at me. 

One main drive of the rise in world energy use is economic growth in emerging market countries. The horizontal axis of this figure shows average energy use per person. The vertical axis shows the cumulative share of total world population. The yellow line shows the pattern for 1978, while the green line shows four decades later in 2018. 


From the caption under the figure: "In 2018, 81% of the global population lived in countries where average energy demand per capita was less than 100 GJ/head, two percentage points more than 20 years ago. However, the share of the global population consuming less than 75 GJ/head declined from 76% in 1998 to 57% last year. Average energy demand per capita in China increased from 17 GJ/head in 1978 to 97 GJ/head in 2018." The figure constructed from national-level data on average energy consumption. Thus, the big jump the blue line at right about 100 GJ/head is the population of China. Overall, the shift from the yellow to the blue line shows how energy consumption is rising in emerging market economies. 

The sources of global energy consumption are also shifting. Oil consumption as shown by the green line is falling as a share of global energy consumption. (Just to be clear, total consumption of oil-produced energy rose in 2018, but it's rising more slowly than overall energy consumption, so the share of the total declined.) Coal remains pretty much the same share of global energy consumption as it has been for the last 30 years. Natural gas has risen. Hydro power is about the same. Nuclear energy is about the same as the last 20 years. Renewables like wind and solar are up, but still only about 5% of total energy consumption. 
High oil prices, reducing the quantity demanded, are part of the economic picture as to why the share of energy produced by oil has declined. The figure shows oil prices back to the Pennsylvania oil boom of the 1860s, with the light-green line showing prices adjusted for inflation. Oil prices have been volatile since the 1970s, but they seem at present to be rising to the middle of their range over the last 4-5 decades. 

What about renewable energy and carbon emissions? From Spencer Dale's overview: 
Renewable energy appears to be coming of age, but to repeat a point I made last year, despite the increasing penetration of renewable power, the fuel mix in the global power system remains depressingly flat, with the shares of both non-fossil fuels (36%) and coal (38%) in 2018 unchanged from their levels 20 years ago. This persistence in the fuel mix highlights a point that the International Energy Agency (IEA) and others have stressed recently; namely that a shift towards greater electrification helps as a pathway to a lower carbon energy system only if it goes hand-in-hand with a decarbonization of the power sector. Electrification without decarbonizing power is of little use. ... On the supply side, the growth in power generation was led by renewable energy, which grew by 14.5%, contributing around a third of the growth; followed by coal (3.0%) and natural gas (3.9%). China continued to lead the way in renewables growth, accounting for 45% of the global growth in renewable power generation, more than the entire OECD combined.
Here's is a figure showing how electricity is generated around the world. Coal still leads the way, by far. Natural gas is on the rise, while oil is dropping. "Renewables," which is led by wind, but also includes solar and smaller categories like geothermal and biomass, is on the rise, but still under 10%.

Finally, here's a table showing carbon emissions in 2018. This is a trimmed-down version of a bigger table in the text. It mainly shows carbon emissions by region. (CIS is "Commonwealth of Independent States," which refers to the remnants of what was once the Soviet Union.) Notice that the Asia-Pacific region accounts for half of all global carbon emissions, with China alone accounting for more than one-quarter of global carbon emissions. Also, the average annual growth rate of carbon emissions was negative for North America and for Europe from 2007-2017, but rising during that time frame in Asia Pacific, as well as Africa, the Middle East, and South/Central America. As I've written before, a meaningful approach to limiting or reducing global carbon emissions will need to include North American and Europe, but our participation won't be nearly enough.