Science

That is in the FT, gated most likely, do subscribe!  In any case, here was to me the most interesting bit:

He [Tetlock] is trying to replace the public debates he describes as “Krugman-Ferguson pie fights” — a reference to the clashes over austerity between the economist and Nobel laureate, Paul Krugman and the economic historian, Niall Ferguson — with adversarial collaboration. “You give each side the opportunity to pose, say, 10 questions it thinks are probative and resolvable, and that it thinks it has a comparative advantage in answering” and then have the two sides give testable answers . . . Here is a very clear psychological prediction: people will come out of that tournament more open-minded than they otherwise would have been. You can take that one to the bank.”

More importantly, Tetlock ordered “…an apple fizz cocktail to go with haddock with a sauce of butter and musses.”

Here was the best single sentence from Tetlock, itself worth the price of an FT subscription:

“There is a price to be paid for feeling good about your beliefs.”

Robert Armstrong did an excellent job with the interview and piece.

“These are black boxes,” said Dr. Steven Joffe, a pediatric oncologist and bioethicist of the University of Pennsylvania, who serves on the FDA’s Pediatric Ethics Committee. “IRBs as a rule are incredibly difficult to study. Their processes are opaque, they don’t publicize what they do. There is no public record of their decision or deliberations, they don’t, as a rule, invite scrutiny or allow themselves to be observed. They ought to be accountable for the work they do.”

That is part of a longer and very interesting article on whether IRBs should be for-profit, or if we even at this point have a choice:

“This shift to commercial IRBs is, in effect, over,” said Caplan, who heads the division of bioethics at New York University Langone Medical Center. “It’s automatic and it’s not going back.”

Institutional review boards — which review all research that involves human participants — have undergone a quiet revolution in recent years, with many drug companies strongly encouraging researchers to use commercial boards, considered by many more efficient than their nonprofit counterparts.

Commercial IRBs now oversee an estimated 70 percent of US clinical trials for drugs and medical devices. The industry has also consolidated, with larger IRBs buying smaller ones, and even private equity firms coming along and buying the companies. Arsenal Capital Partners, for example, now owns WIRB-Copernicus Group.

But even if the tide has already turned, the debate over commercial review boards — and whether they can serve as human subject safety nets, responsible for protecting the hundreds of thousands of people who enroll in clinical trials each year — continues to swirl.

I am not well-informed in this area, but if you refer back to the first paragraph, perhaps nobody is.  That’s worrying.

For the pointer I thank Michelle Dawson.

Very few people imagined that self-driving cars would advance so quickly or be deployed so rapidly. As a result, robot cars are largely unregulated. There is no government testing regime or pre-certification for robot cars, for example. Indeed, most states don’t even require a human driver because no one imagined that there was an alternative. Many people, however, are beginning to question laissez-faire in light of the first fatality involving a partially-autonomous car that occurred in May and became public last week. That would be a mistake. The normal system of laissez-faire is working well for robot cars.

Laissez-faire for new technologies is the norm. In the automotive world, for example, new technologies have been deployed on cars for over a hundred years without pre-certification including seatbelts, air bags, crumple zones, abs braking systems, adaptive cruise control and lane departure and collision warning systems. Some of these technologies are now regulated but regulation came after these technologies were developed and became common. Airbags began to be deployed in the 1970s, for example when they were not as safe as they are today but airbags improved over time and by the 1990s were fairly common. It was only in 1998, long after they were an option and the design had stabilized, that the Federal government required airbags in all new cars.

Lane departure and collision warning systems, among other technologies, remain largely unregulated by the Federal government today. All technologies, however, are regulated by the ordinary rules of tort (part of the laissez-faire system). The tort system is imperfect but it works tolerably well especially when it focuses on contract and disclosure. Market regulation also occurs through the insurance companies. Will insurance companies given a discount for self-driving cars? Will they charge more? Forbid the use of self-driving cars? Let the system evolve an answer.

Had burdensome regulations been imposed on airbags in the 1970s the technology would have been delayed and the net result could well have been more injury and death. We have ignored important tradeoffs in drug regulation to our detriment. Let’s avoid these errors in the regulation of other technologies.

The fatality in May was a tragedy but so were the approximately 35,000 other traffic fatalities that occurred last year without a robot at the wheel. At present, these technologies appear to be increasing safety but even more importantly what I have called the glide path of the technology looks very good. Investment is flowing into this field and we don’t want to forestall improvements by raising costs now or imposing technological “fixes” which could well be obsolete in a few years.

Laissez-faire is working well for robot cars. Let’s avoid over-regulation today so that in a dozen years we can argue about whether all cars should be required to be robot cars.

Optical Illusion of the Year

by on July 3, 2016 at 4:55 pm in Games, Science | Permalink

Claims about clutter

by on July 2, 2016 at 11:09 am in Books, Education, Science, The Arts | Permalink

Tidy by category, not by location

One of the most common mistakes people make is to tidy room by room.  This approach doesn’t work because people think they have tied up when in fact they have only shuffled their things around from one location to another or scattered items in the same category around the house, making it impossible to get an accurate grasp of the volume of things they actually own.

The correct approach is to tidy by category.  This means tidying up all the things in the same category in one go.  For example, when tidying the clothes category, the first step is to gather every item of clothing from the entire house in one spot.  This allows you to see objectively exactly how much you have.  Confronted with an enormous mound of clothes, you will also be forced to acknowledge how poorly you have been treating your possessions.  It’s very important to get an accurate grasp of the sheer volume for each category.

That is from Marie Kondo, Spark Joy: An Illustrated Guide to the Japanese Art of Tidying, a recommended book.  Also never tidy the kitchen first, do not keep make-up and skin care products together, and “…the first step in tidying is to get rid of things that don’t spark joy.”

I have a related tip.  If you want to do a truly significant clean-up, focus only on those problems which are not immediately visible.  This will help you build efficient systems, and prepare the way for more systematic solutions to your clutter problems.  You’ll then be prompted to take care of the visible problems in any case.  If you focus on the visible problems instead, you will solve them for a day or two but they will rapidly reemerge because the overall quality of your systems has not improved.

Popular Science: A pilot A.I. developed by a doctoral graduate from the University of Cincinnati has shown that it can not only beat other A.I.s, but also a professional fighter pilot with decades of experience. In a series of flight combat simulations, the A.I. successfully evaded retired U.S. Air Force Colonel Gene “Geno” Lee, and shot him down every time. In a statement, Lee called it “the most aggressive, responsive, dynamic and credible A.I. I’ve seen to date.”

What’s the most important part of this paragraph? The fact that an AI downed a professional fighter pilot? Or the fact that the AI was developed by a graduate student?

In the research paper the article is based on the authors note:

…given an average human visual reaction time of 0.15 to 0.30 seconds, and an even longer time to think of optimal plans and coordinate them with friendly forces, there is a huge window of improvement that an Artificial Intelligence (AI) can capitalize upon.

The AI was running on a $35 Raspberry Pi.

AI pilots can plan and react far quicker than human pilots but that is only half the story. Once we have AI pilots, the entire plane can be redesigned. We can build planes today that are much faster and more powerful than anything that exists now but the pilots can’t take the G-forces even with g-suits, AIs can. Moreover, AI driven planes don’t need ejector seats, life-support, canopies or as much space as humans.

The military won’t hesitate to deploy these systems for battlefield dominance so now seems like a good time to recommend Concrete Problems in AI Safety, a very important paper written by some of the world’s leading researchers in artificial intelligence. The paper examines practical ways to design AI systems so they don’t run off the rails. In the Terminator movie, for example, Skynet goes wrong because it concludes that the best way to fulfill its function to safeguard the world is to eliminate all humans–this is an extreme example of one type of problem, reward hacking.

Imagine that an agent discovers a buffer overflow in its reward function: it may then use this to get extremely high reward in an unintended way. From the agent’s point of view, this is not a bug, but simply how the environment works, and is thus a valid strategy like any other for achieving reward. For example, if our cleaning robot is set up to earn reward for not seeing any messes, it might simply close its eyes rather than ever cleaning anything up. Or if the robot is rewarded for cleaning messes, it may intentionally create work so it can earn more reward. More broadly, formal rewards or objective functions are an attempt to capture the designer’s informal intent, and sometimes these objective functions, or their implementation, can be “gamed” by solutions that are valid in some literal sense but don’t meet the designer’s intent. Pursuit of these “reward hacks” can lead to coherent but unanticipated behavior, and has the potential for harmful impacts in real-world systems. For example, it has been shown that genetic algorithms can often output unexpected but formally correct solutions to problems [155, 22], such as a circuit tasked to keep time which instead developed into a radio that picked up the regular RF emissions of a nearby PC.

Concrete Problems in AI Safety asks what kind of general solutions might exist to prevent or ameliorate reward hacking when we can never know all the variables that might be hacked? (The paper looks at many other issues as well.)

Competitive pressures on the battlefield and in the market mean that AI adoption will be rapid and AIs will be placed in greater and greater positions of responsibility. Firms and governments, however, have an incentive to write piecemeal solutions to AI control for each new domain but that is unlikely to be optimal. We need general solutions so that every AI benefits from the best thinking across a wide range of domains. Incentive design is hard enough when applied to humans. It will take a significant research effort combining ideas from computer science, mathematics and economics to design the right kind of incentive and learning structures for super-human AIs.

That is the new book by Ben Wilson, and no it has nothing (directly) to do with Brexit.  Rather it is a survey of the technological breakthroughs of the 1850s and how they reshaped Great Britain and the globe more generally.  Here is one short bit:

Japan may have secluded itself from the rest of the world, but it had not closed itself off.  That was a distinction that people in the West were slow to grasp.  The shogun’s court subscribed to the Illustrated London News, for example, and the bakufu had acquired books and papers detailing global politics and scientific discoveries through their Dutch and Chinese trading partners.  This knowledge was strictly regulated, but the seeds of scientific enlightenment were diffused in small numbers across the archipelago.  Perry did not know it — and nor did many Japanese — but his telegraph was not the first on Japanese soil.

Other parts of this book which I enjoyed were on the Great Geomagnetic Storm of 1859, how the British saw a connection between the U.S. Civil War, and the origins of Reuters.

If you want a new Brexit-relevant title of interest, try Brendan Simms, Britain’s Europe: A Thousand Years of Conflict and Cooperation.

It can be incredibly frustrating when a virtual assistant repeatedly misunderstands what you’re saying. Soon, though, some of them might at least be able to hear the irritation in your voice, and offer an apology.

Amazon is working on significant updates to Alexa, the virtual helper that lives inside the company’s voice-controlled home appliance, called Amazon Echo. These will include better language skills and perhaps the ability to recognize the emotional tenor of your voice.

Researchers have long predicted that emotional cues could make machine interfaces much smarter, but so far such technology has not been incorporated into any consumer technology.

Rosalind Picard, a professor at MIT’s Media Lab, says adding emotion sensing to personal electronics could improve them: “Yes, definitely, this is spot on.” In a 1997 book, Affective Computing, Picard first mentioned the idea of changing the voice of a virtual helper in response to a user’s emotional state. She notes that research has shown how matching a computer’s voice to that of a person can make communication more efficient and effective. “There are lots of ways it could help,” she says.

The software needed to detect the emotional state in a person’s voice exists already. For some time, telephone support companies have used such technology to detect when a customer is becoming irritated while dealing with an automated system. In recent years, new machine-learning techniques have improved the state of the art, making it possible to detect more emotional states with greater accuracy, although the approach is far from perfect.

Here is the full story.  Here is my recent New Yorker piece on how talking bots will affect us.

The robot administers a small pin prick at random to certain people of its choosing.

The tiny injury pierces the flesh and draws blood.

Mr Reben has nicknamed it ‘The First Law’ after a set of rules devised by sci-fi author Isaac Asimov.

He created it to generate discussion around our fear of man made machines. He says his latest device shows we need to prepare for the worst

‘Obviously, a needle is a minimum amount of injury, however – now that this class of robot exists, it will have to be confronted,’ Mr Reben said on his website.

Here is more, with pictures of (slightly) injured humans, via the excellent Mark Thorson.

William W. Olney has a newly published paper on that topic, and it seems it helps a good deal:

This article investigates whether the global spread of the English language provides an inherent advantage to native English speakers. This question is studied within the context of the economics profession, where the impact of being a native English speaker on future publishing success is examined. English speakers may have an advantage because they are writing in their native language, the quality of writing is a crucial determinant of publishing success, and all the top economics journals are published in English. Using a ranking of the world’s top 2.5% of economists, this article confirms that native English speakers are ranked 100 spots higher (better) than similar non-native English speakers. A variety of extensions examine and dispel many other potential explanations.

“Similar” is a tricky word!  How similar can a Frenchman be?  I am not sure, but it does seem that growing up in the Anglo-American world may — language aside — be more conducive to patterns of thought which predict success in…the Anglo-American world.  Nonetheless this is an interesting investigation, even if I am not entirely convinced.  Note also that economics blogging is predominantly an Anglo-American enterprise, but I view that too as more about “mentality” than language per se.

For the pointer I thank the excellent Kevin Lewis.

Hermance

I am here to give a talk on randomized control trials, a public choice perspective.  Angus Deaton and Josh Angrist are among the other speakers, along with many people in the medical field.  The first question, not quite resulting from a controlled experiment, is whether this setting, on Lake Geneva, improves or worsens my mood…

A reader has been asking me this question, and my answer is…no!

Don’t get me wrong, I still think it is a stimulating and wonderful book.  And if you don’t believe me, here is The Wall Street Journal:

Mr. Hanson’s book is comprehensive and not put-downable.

But it is best not read as a predictive text, much as Robin might disagree with that assessment.  Why not?  I have three main reasons, all of which are a sort of punting, nonetheless on topics outside one’s areas of expertise deference is very often the correct response.  Here goes:

1. I know a few people who have expertise in neuroscience, and they have never mentioned to me that things might turn out this way (brain scans uploaded into computers to create actual beings and furthermore as the dominant form of civilization).  Maybe they’re just holding back, but I don’t think so.  The neuroscience profession as a whole seems to be unconvinced and for the most part not even pondering this scenario.

2. The people who predict “the age of Em” claim expertise in a variety of fields surrounding neuroscience, including computer science and physics, and thus they might believe they are broader and thus superior experts.  But in general claiming expertise in “more” fields is not correlated with finding the truth, unless you can convince people in the connected specialized fields you are writing about.  I don’t see this happening, nor do I believe that neuroscience is somehow hopelessly corrupt or politicized.  What I do see the “Em partisans” sharing is an early love of science fiction, a very valuable enterprise I might add.

3. Robin seems to think the age of Em could come about reasonably soon (sorry, I am in Geneva and don’t have the book with me for an exact quotation).  Yet I don’t see any sign of such a radical transformation in market prices.  Even with positive discounting, I would expect backwards induction to mean that an eventual “Em scenario” would affect lots of prices now.  There are for instance a variety of 100-year bonds, but Em scenarios do not seem to be a factor in their pricing.

Robin himself believes that market prices are the best arbiter of truth.  But which market prices today show a realistic probability for an “Age of Em”?  Are there pending price bubbles in Em-producing firms, or energy companies, just as internet grocery delivery was the object of lots of speculation in 1999-2000?  I don’t see it.

The one market price that has changed is the “shadow value of Robin Hanson,” because he has finished and published a very good and very successful book.  And that pleases me greatly, no matter which version of Robin is hanging around fifty years hence.

Addendum: Robin Hanson responds.  I enjoyed this line: “Tyler has spent too much time around media pundits if he thinks he should be hearing a buzz about anything big that might happen in the next few centuries!”

From Geoff Kaufman and Mary Flanagan:

The present research investigated whether digital and non-digital platforms activate differing default levels of cognitive construal. Two initial randomized experiments revealed that individuals who completed the same information processing task on a digital mobile device (a tablet or laptop computer) versus a non-digital platform (a physical print-out) exhibited a lower level of construal, one prioritizing immediate, concrete details over abstract, decontextualized interpretations. This pattern emerged both in digital platform participants’ greater preference for concrete versus abstract descriptions of behaviors as well as superior performance on detail-focused items (and inferior performance on inference-focused items) on a reading comprehension assessment. A pair of final studies found that the likelihood of correctly solving a problem-solving task requiring higher-level “gist” processing was: (1) higher for participants who processed the information for task on a non-digital versus digital platform and (2) heightened for digital platform participants who had first completed an activity activating an abstract mindset, compared to (equivalent) performance levels exhibited by participants who had either completed no prior activity or completed an activity activating a concrete mindset.

Here is also the press release, and for the pointer I thank Charles Klingman.

I have a short New Yorker piece on that question, here is one bit from it:

If Siri is sometimes sarcastic, could heavy users of the Siri of the future become a little more sarcastic, too?

For companies, there are risks associated with such widespread personification. For a time, consumers may be lulled by conversational products into increased intimacy and loyalty. But, later, they may feel especially betrayed by products they’ve come to think of as friends. Like politicians, who build up trust by acting like members of the family only to incur wrath when they are revealed to be careerist and self-interested, companies may find themselves on an emotional roller coaster. They’ll also have to deal with complicated subjects like politics. Recently, Tay, a chat bot from Microsoft, had to be disabled because it began issuing tweets with Nazi-like rhetoric. According to Elizabeth Dwoskin, in the Post, Cortana, another talking Microsoft bot, was carefully programmed not to express favoritism for either Hillary Clinton or Donald Trump. A product’s apparent intelligence makes it likable, but also offers more of an opportunity to offend.

Here is another:

And there are ways in which just knowing that bots exist will change us. If the bots are good enough, we won’t be able to distinguish them from actual people over e-mail or text; when you get an e-mail, you won’t necessarily be certain it’s from a human being. When your best friend writes that she’s also “looking forward to seeing you at the baseball game tonight,” you’ll smile—then wonder if she’s busy and has asked her e-mail bot to send appropriate replies. Once everyone realizes that there might not be a person on the other end, peremptory behavior online may become more common. We’ll likely learn to treat bots more like people. But, in the process, we may end up treating people more like bots.

Do read the whole thing.

Maybe so, here is the latest:

Where witchcraft beliefs are widespread, American University Economics Professor Boris Gershman found high levels of mistrust exist among people. Gershman also found a negative relationship between witchcraft beliefs and other metrics of relied upon for a functioning society, including religious participation and charitable giving.

It’s long been argued that witchcraft beliefs impede economic progress and disrupt social relations, and Gershman’s statistical analysis supports that theory. From a policy perspective, Gershman’s results emphasize the importance of accounting for local culture when undertaking development projects, especially those that require communal effort and cooperation. Gershman and other social scientists believe that education can help foster improved trust and decrease the prevalence of witchcraft beliefs.

Furthermore:

Parents in witchcraft-believing societies inculcate antisocial traits in children.

Second-generation immigrants from witchcraft-believing nations are less trusting.

Here is the summary statement, here is the full article.  Here are related papers by Gershman.  For the pointer I thank the excellent Samir Varma.