This is GMcKCypress's Typepad Profile.
Join Typepad and start following GMcKCypress's activity
Join Now!
Already a member? Sign In
GMcKCypress
Recent Activity
It's a good list of counterarguments. I think it leaves out two more, thouugh. The "paperclip maximizer" argument has already been refuted in real life. We already have an immortal, amoral, super-humanly intelligent "organism" whose goal is to make as many paperclips as possible - it's called "ACCO Brands". It's failed to take over the world because it's not the only one of its kind. It has other competitors in the same intelligence class who are dedicated to making sure that it fails in its goal, because they have the same goal, and they are determined to win, not ACCO. More abstractly, entities exist in ecosystems, and are dependent on services provided by those ecosystems for their own survival. When they damage or destroy their ecosystem, they end up destroying themselves. The other argument is more technical, and comes from the realm of complexity theory. First of all, there are no physically realistic levels of computational capability beyond Turing equivalence, as Kaleburg comments above. Once you've learned to read and write, you can think any thinkable thought, given enough time and paper and pencils. There's no qualitiative difference available to be worried about. But maybe a superintelligence could just be so much blindingly faster and bigger than human intelligence that it amounts to a qualitative difference. Well, there's a question known as "P = NP?" that despite being the acknowledged as the most important problem in computer science, remains unresolved. All signs point to P being different from NP. This means that there are many properties of interconnected systems such as ecosystems, economies, societies of minds, or even individual minds, which as they grow become harder to determine faster than the capacity of any system to identify or control them grows. Even a superintelligence can only "rule the world" by fooling itself - an honest superintelligence will still just be muddling through, just like the rest of us.
1 reply
If Boeing hadn't discontinued its other single-aisle airplane, the 757, they woudn't have needed to make those terrible decisions to overdrive the 737's design with those excessively large engines. The tall landing gear on the 757 that made it look like a stork had plenty of clearance for engines as big as anyone could have wanted. The 757 was said to be a hot rod and a pilot's airplane. Even Donald Trump was flying around in one before he upgraded to Air Force One. But Boeing wanted a plane that could fly into low-tech airports that couldn't even provide their own stairs - the 737 is low enough that it can carry its own stairs. And here we are...
1 reply
If Boeing hadn't discontinued its other single-aisle model, the 757, they woudn't have needed to make those terrible decisions to overdrive the 737's design with those excessively large engines. The tall landing gear on the 757 that made it look like a stork had plenty of clearance for engines as big as anyone could have wanted. The 757 was said to be a hot rod and a pilot's airplane. Even Donald Trump was flying around in one before he upgraded to Air Force One. But Boeing wanted a plane that could fly into low-tech airports that couldn't even provide their own stairs - the 737 is low enough that it can carry its own stairs. And here we are...
1 reply
What's most remarkable about this announcement is the dead silence from the punditocracy in reaction to it. It appears that nobody has any ideas how to fix US healthcare beyond the politically locked-in ideas of giving people the freedom to suffer and die due to inability to pay, or Obamacare, or universal coverage via single-payer or single-provider systems. Since Buffett derives much of his wealth from insurance businesses, it seems unlikely that this group would go for anything that would destroy that industry. Maybe I have a bit more imagination than those pundits: to my eyes there are three classes of problems with the healthcare system. (1) Pay for service pricing incentivizes (profitable!) overtreatment without accounting for interventions' effectiveness at maintaining patients' health. (2) Massive intermediation creates incredibly complex (and profitable!) supply chains and equally complex (and profitable!) payment chains. (3) Lack of transparent pricing and monopolistic abuse of intellectual property laws, and the fat of the distribution of costs of illnesses that makes shopping for cost-effective treatments intractable, prevent free markets from driving costs out of the system. Other than the governmental single-payer/provider solutions, I see one private approach that hasn't been really tried yet, namely massive vertical integration. We're seeing bits of this with events like the merger of CVS and Aetna, and that could be just the beginning. Bezos has access to business process design expertise to drive out massive logistical costs ("Alexa, my baby girl has a rash, a headache and a fever, what should I do?") and partnering with Buffett's insurance expertise could help drive out payment costs. Of course, this was tried 30 years ago with "health maintenance organizations", and that didn't work out so well when for-profit HMO's discovered that is was cheaper to act as a Care Denial Organization than to deal with the complexity of millions of individual patient profiles. But in addition to its logistical expertise, Amazon's other core competence is its customer profiling capability. Keeping health records private to the standards set by HIPAA regulations while making them accessible to people and software with need to know is complex; building this kind of trust is hard; both Microsoft and Google have put their toes into health records management without obvious success. There are a handful of HMO's that seem to have been successful in achieving the original vision of delivering a system that puts patient health ahead of other factors, but they've all been limited to regional scale. I'd be thrilled to see this group produce a system that works at national scale. But as the saying goes, "I should live so long..."
1 reply
If the rest of the book is anything like the downloadable intro, I'll give it a pass. It's looking not like a coherent theory, which would be awesome in its complexity, but a pastiche of quasi-insights from half-understood theories in other fields. In particular, despite having a section title "it's the environment, stupid", the rest of that very section fails to comprehend that evolutionary adaptations are perfectly rational responses, in a stochastic way, to organisms' discovery that different individuals exist in different environments at any given time, and thereby identify different paths towards maximizing their respective fitnesses. There's a lot to be learned by economists from evolutionary biology, but Lo doesn't appear to have learned very much. I suppose that if you're still fighting the EMH wars, that's forgivable, but it's still disappointing.
Toggle Commented May 20, 2017 on Darwin Visits Wall Street at Economist's View
1 reply
No economist should criticize politicians' economic rantings for being out of touch with reality, when prominent economists see nothing wrong with their own theories being at best remotely related to reality. The beam in one's own eye, etc...
1 reply
I can never get over seeing economic authorities say things like "The purpose of these models is to make a deep theoretical point, likely of relevance to nearly any macro model, but not pretending to capture reality closely." People working on this kind of model need to move to math and philosophy departments where dissociation from reality is not a concern. But I suppose there's still hope for Blanchard, two of the 5 types of model are getting sort of close to the scientific method. And the reductionist/constructivist paradigm where emergent phenomena are derived from the properties and behavior of more fundamental elements, i.e.agent based models, even gets a footnote.
1 reply
Kremer's model was OK for its day, but I'd hope that it's been substantially superseded in the 25 years since. We have a lot more data about growth that an adequate model needs to account for than we did then. Charles Jones' 2015 article in the Handbook of Macroeconomics lists a good portion of it. Daron Acemoglu and his colleagues are collecting more even now, with some counterintuitive results. For example one of that group, David Autor, recently mentioned in an interview that globalization increases competition, reducing profit margins, leaving less funds available for R&D, slowing the rate of growth. Kremer's model leaves out resources as a factor affecting regional growth, which anyone who has read Jared Diamond's Guns, Germs and Steel would find immediately falsifying. (GG&S came out 4 years later than Kremer's paper.) In particular, Kremer's model has no way of explaining how colonial takeovers, with their advanced technologies, often reduce GDP per capita, as during the British Raj in India (see also "the resource curse"), or how India was the world's leading producer of scientific papers for many years before its recent growth takeoff, with little effect on GDP growth. If I were to build a model in Kremer's style today, I'd model "technology" as an autocatalytic set of individual technologies, operating over a diffusive field of regional initial population and resource substrates. The idea of autocatalysis had been around for 15 years when Kremer was writing, but the compute power needed to study it wasn't available. Today, your phone probably has more compute power than the biggest supercomputer in 1993, so raw GFLOPS are not the problem. The programming technology to model, as evolutionary groups of autocatalytic subsets, tool-building towers like 3D printing for rocket engines to launch communications satellites to carry TV shows, is still bleeding-edge.
1 reply
If you're looking for a sound-bite definition of "singularity", I think the defintion of a weak singularity as "a structural change that makes the future profoundly unpredictable given the past" is quite reasonable. The Kurzweil/Vinge strong singularity is more religious than scientific. Apologies if I've said this here before: In 2014 Cosma Shalizi listed some evidence that the Industrial Revolution qualifies as a technological singularity. (http://bactra.org/notebooks/699.html) Shalizi has a very good list of criteria that should be satisfied in order for a singularity to be recognized. As someone who did research in cell biology and in social sciences at different times long ago, I think there are a lot more in the history of life on the planet. Here's a list of past (and 1.666 future) candidates -- most of Shalizi's criteria are met by each one. Note that when I say "discovery" I mean discovery by "selfish genes" or "memes" that spontaneously replicate and are naturally selected for, not by individuals. 1. The sequestration of molecular replication in membrane-bounded cells 2. The switch from storage of genetic information in RNA sequences to its storage in DNA sequences. 3. The discovery of photosynthesis by the ancestors of cyanobacteria. 4. The discovery of the rules for cellular differentiation, adhesion, and migration that led to multicellular organisms 5. The discovery of backwards development by deuterostomes that led to internal skeletons rather than exoskeletons Not saying that arthropods are bad, just that endoskeletons are better at growing big. Deuterostome development certainly leads to other severe problems. 6. [not saying anything about all the mass extinctions that led to mammalian domination of land animal life] 7. The discovery of learning by imitation rather than individual trial and error 8. The invention of controlled fire 9. The discovery of information storage and retrieval from conspecifics by means of reverence for tribal elders, e.g. "old wives' tales" 10. The discovery and institutionalization of marketplaces 11. The invention of writing 12. The first industrial revolution 13. The second industrial revolution of information technology, robotics & biotechnology 14. The third technological revolution of controlled ecological engineering Humanity is involved in the most recent 8 singularities, Gordon seems to be interested only in industrial/technological revolutions. Robotics & longevity don't do happy things to median income when most people have minimal net worth to invest in compound-interest-yielding capital.
1 reply
I think we're seeing a Kuhn-type scientific revolution playing out in the 25 years since Bell wrote this. Not a word about "decoherence" in all this thinking. Yet Sidney Coleman in his 1994 "quantum mechanics in your face" lecture was already talking about how QFT works just fine without any measurement. And Zurek and Saleh in what should be a famous 1990 article in Physics Today showed the reshaping of the wavefunction of a "particle" under the influence of a wavelike "measurement apparatus" -- no sudden, magical "collapse" was needed. Nowdays, the frontiers of quantum mechanics are all about superposition and entanglement over extended times and distances, teleportation, backwards causation, non-demolition measurement beyond the Heisenberg limit, and other incredible stuff where the issues that Bell and Everett were worrying about are as archaic as concerns about the origin of "vitality" in living organisms are to a molecular biologist.
1 reply
Meanwhile, the tech industry itself is laying off about 369,000 people this year. http://spectrum.ieee.org/view-from-the-valley/at-work/tech-careers/analyst-ugly-year-for-tech-layoffs-and-its-going-to-get-worse
1 reply
Thompson/Stratechery seem to be ambivalent about the strategies - this article is paywalled, but other articles of his on the same site are open. According to your quote, this means that Stratechery is doomed unless they decide one or the other. There are two long tail populations that need to be coordinated - the long tail of niche readers and the long tail of niche authors. Most people are members of many niche audiences -- the "mass market reader" is as much of an oversimplification as the "representative agent" clones are in macroeconomic models. Niche readers need to be able to scan large quantities of sources in order to find the optimal match for their interests, and total free access is the only way for a niche publisher to provide enough content to that reader to allow for an informed decision. Some hybrid strategies may not be fatal. The academic open access model is one: niche market funded by authors who are themselves philanthropically funded (or industrially funded in the case of many tech and biomedical researchers). Sugar mommies and sugar daddies don't have to be big grantors - Patreon and others like it are based on the notion that aggregated settlement systems can enable micro-philanthropy. Ad substitution is a niche in this ecosystem that is not fully explored. Adblock Plus is soliciting paying advertisers to place well-behaved ads in place of blocked ones -- this is technically just as parasitic as the cuckoo laying its eggs in the nests of other birds, and the Ad Council is justifiably outraged. But I would seriously consider paying Google directly to forego its ad revenue for that space on the page if they would just shut up. There's a huge moral difference between informative marketing and emotionally seductive and extortionate advertising, which advertising has obliterated.
1 reply
Olivier Blanchard summarizes a lot of the problems with DSGE models, but then makes a remarkable statement: "A typical DSGE paper ... starts with an algebra-heavy derivation of the model, then goes through estimation, and ends with various dynamic simulations showing the effects of the distortion on the general equilibrium properties of the model. These would indeed seem to be the characteristics of a mature science: Building on a well understood, agreed upon body of science and exploring modifications and extensions." There's no mention of testing the modifications against reality and throwing out changes to the theory that don't improve its match to reality. Blanchard's "science" doesn't have the iterative improvement aspect that is characteristic of science in other fields. Apparently the mature science of economics is pure mathematics, not a natural science. Even when he suggests that this not may be a good thing, he states it in an extraordinarily diffident way: "It may well be that, for these purposes, reduced form models will continue to beat structural models for some time; theoretical purity may be for the moment more of a hindrance than a strength." And he doesn't seem to be aware at all of the irony of policymakers informing analyses of the effects of policy changes by using a class of model that because of its so-called microfoundations is said to be immune to any Lucas-type policy effects.
1 reply
GMcKCypress is now following The Typepad Team
Aug 14, 2016