Ask HN: What are some problems solved in unrelated domains?
Elaborating, what are some problems that were being experienced by one industry or domain but were solved elsewhere and were adopted by the industry plagued by the problem? Even better if the answer was really obvious for the people who solved it or solving it involved cross-domain knowledge.
For context, I recently heard having a breadth of knowledge (along with depth in chosen areas) is really helpful in problem solving and I'm looking for concrete examples for the same. Kanban comes from the automotive industry, and works really well in the tech industry. The Toyota Production System is a brilliant set of principles and practices that fit well in most industries. Toyota sent engineers to optimize street kitchens some years ago. ISO 9001 was implemented in manufacturing, and is quite popular in many circles of hell. When I researched this I found that automotive Kanban bore absolutely no relationship to the computing industry's Kanban. Shigeo Shingo's book (https://www.amazon.co.uk/dp/0915299178/) is particularly emphatic that Kanban was a was for repeatable industrial processes and not, for example, for knowledge work. Automotive Kanban is not a "Todo"/"doing"/"done" board, by his account anyway. The tickets sat on individual work items and moved around the factory. It was about limiting the number of unfinished work items (which take up space and can be a sign that something is out of control). IMHO Kanban in the computing industry is an analogue of typical task scheduling algorithm on the CPU. Instead of planning the work ahead, you have a work queue and you work each item from it until either the work itself becomes blocked, or (if you're timeboxing/preempting) a fixed period of time expires. But more generally to the poster's question - like the example above, I think there is a lot in the project management space that was already solved by computer scientists. :-) > It was about limiting the number of unfinished work items (which take up space and can be a sign that something is out of control). As it happens, limiting unfinished work items (aka WIP) boosts software development productivity as well. The space being taken up is mental, but is also a bad sign of things spinning out of control. Unless you're in a startup, where most ideas are usually distractions, or trying to play catch up with a competitor. Well, sure. But if 'succumbing to distractions' is the issue, I'll note that this also tends to result in piles of semi-complete WIP. Patterns taken from manufacturing do not always translate well to software. Kanban is a notable exception. The "pull" model of kanban fits well with the Agile approach to late resource allocation to tasks. Kanban, while it was born in the repeating process world of manufacturing, translates well to the non-repeating, yet iterative nature of software projects. The black-scholes model for pricing options had a big impact on the financial industry, and the big breakthrough for solving the equation was to realize it could be turned into the heat equation that already had a solution. If I remember well, the insight came to Robert Merton, an economist who had also studied engineering mathematics. "Big impact" is understating it. It's caused at least one financial crisis when Scholes and Merton's hedge fund (https://en.wikipedia.org/wiki/Long-Term_Capital_Management) went boom in the late nineties. Since then the assumptions present in Black-Scholes and similar pricing models have caused considerable chaos elsewhere in finance. One of those things that perhaps it would have been better that they had not invented. Professionally it was great for the inventors though: they won the Nobel Prize at the time. The Ito calculus being applicable for solving Black-Scholes might be a better example for the original question :) I am not sure it's better if that was not invented - the question of option pricing doesn't go away, it's a fundamental need for real businesses producing commodities that need to hedge real risk around the weather, or selling into cyclical industries etc. LTCM (& it's failure) has as much to do with getting so big as to present a systemic risk to the entire financial sector, and over-leverage. No matter what formula they use to price their options, those two factors make for a dangerous combination. There is a perverse incentive to become "too big to fail", because if you truly reach that, the gains stay private whereas the losses are inevitably socialized to prevent taking down the whole system. It's also led to lower costs in just about every industry due to the ability to price obscure contracts and lower risk. An example is airlines need to deal with fuel price volatility when selling tickets now, and, if they take on more risk, they need to price that into tickets. Getting someone to hedge against that lowers their risk. The pricing of such hedges and contracts has led to less volatility for pretty much every industry. For about a decade, as I flew a lot, I met a lot of mid level executives at all sorts of companies. I routinely asked them to describe if their company used any derivative products, and which, and why. I was astounded to learn pretty much everyone did, from pricing plastics, to dealing with international trade variations, to getting financing for projects. There is a reason the derivatives market has a notional value of ~1 quadrillion, and it's not because the people using them are all idiots. Quite the contrary, this value is there because it adds value across every sector of the world economy, and business the world over voluntarily use these products because they deem it in their best interests, which almost universally is because it lowers costs and/or volatility. This [1] claimed 94% of the world's biggest companies used derivatives, and this was in 2009. I'd expect more and more of them to use derivatives over time as more products get priced. All this followed from the ideas in Black-Scholes, for pricing certain contracts, ideas which were later developed to handle all sorts of products. Before this, people did a lot of the pricing mostly by guesswork, making it much harder to design and use all these volatility reducing products. [1] https://derivsource.com/2009/04/23/over-94-of-the-worlds-lar... LTCM's blowup had nothing to do with any flaw in Black-Scholes. They did lose money on equity volatility strategies but it was because they were betting that volatility in the future would be lower than the volatility implied by Black-Scholes. Volatility ended up going up and they lost money. They lost money because their forecast of volatility was wrong, not because Black-Scholes was wrong about the market's view of volatility. Financial crises are what financial model assumptions become when they grow up. > One of those things that perhaps it would have been better that they had not invented. I think you're off base on this. The ensuing financial mishaps just made the model better, but the creation of Black-Scholes brought us from 0% to 90% of the way to where we are today, where modern derivatives insure product price swings for almost everything across the world. Since then, a patch-work of model changes have brought us closer to perfect pricing through the realization that many markets have higher kurtosis. Breath of knowledge is certainly useful. I'm sure there's a lot of redundant work across domains in universities, but here's a somewhat humorous example of someone recreating the trapezoidal method of approximating the area under a curve. They needed it in a medical context, and anyone who has done Calc 2 would tell you "that's an integral". It shows 402 citations on Google Scholar. https://scholar.google.com/scholar?cluster=18129095207210817... https://academia.stackexchange.com/questions/9602/rediscover... It's worth noting that some academics cite that paper as a joke, I'm not sure how many genuine citations there are. Contrary to popular belief problems solved in unrelated domains are very common in invention and innovation but people just hate to admit them, hopefully in the near future someone can honestly write a book on it. One of the most prolific scientists Albert Einstein had his single most productive year while working as a patent officer (Annus Mirabilis). I strongly suspect that he learned quite a lot from many patent applications, journal papers and books that he was required to read during his daily job, thus getting excellent and novel ideas from seemingly unrelated patents and discoveries. The classic modern example (sorry for the oxymoron terminology) is how a patented signal processing technique of radio astronomy research by CSIRO solved the wireless multi-path propagation problem that enabled wireless revolution from WiFi to 4G/5G. CSIRO's patent leads to the wireless OFDM invention that allows for much higher communication bandwidth especially on wireless environment where unmitigated multi-path interference is a deal breaker. But of course wireless people just hate to admit it [1]. [1]https://arstechnica.com/tech-policy/2012/04/how-the-aussie-g... A Hospital Races To Learn Lessons Of Ferrari Pit Stop https://www.wsj.com/articles/SB116346916169622261 Knowledge from pit stop procedures improves speed and decreases accidents in hospital settings. I'm not sure if it is possible to pinpoint the specific point of crossover, but the analysis of crowd dynamics owes a lot to the study of fluid dynamics, and the counterintuitive insight that placing an obstacle like a column in front of an exit can smooth out and speed the outflow of people by adding some turbulence (preventing the jams that can crush people) is one specific example that I recall. Ada Lovelace developed general purpose computer programming from her understanding of the encoding method implemented by the Jacquard loom. https://www.irenebrination.com/irenebrination_notes_on_a/201... If you look in the field of network analysis there’s several camps of researchers, namely: physicists, sociologists, mathematicians and computer scientists all solving the same problems but from different domains with different terms and drawing from different prior work. They all interestingly converge in the sub field of community detection or graph clustering. for me it was learning differential equations. From mechanical systems, chemical reactions, electrical circuits, ecological systems, control systems, epidemiology... Natural selection led to evolutionary optimization algorithms
Statistical mechanics to boltzmann machines, ising model to perceptron I was thinking of two more in this vein, but I'm not actually sure about the history of their derivation, so I'll pose them as questions: Was simulated annealing inspired by actual annealing (heating and slowly cooling metal/glass to toughen it)? Or did they just name it annealing after discovering it? What is the relationship between thermodynamic/statistical entropy and Shannon's information theory entropy (E = -p lg(p))? there are a lot of applications of physical theories in other sciences. annealing is just a descriptive name though. Shannons information entropy is a generalization of thermodynamic entropy. Brilliant! I've always wondered the relationship between these two. A real popular one would be QR codes. They were initially created to track manufacturing processes in the automotive industry. And you know how it is these days... QR codes are amazing and yet untapped, in my opinion. They're like barcodes but: - amount of data can vary from small to large
- error correction They don't have to just hold URLs and can contain raw information that custom programs can read. I love their use in it for virtual currencies as public/private addresses b/c it simplifies them for users. Venmo also picked it up not too long ago to easily pay/request to people you may not have friended which is useful for street vendors. Someone posted an example here about contact tracing and how it just leads to a web form to fill details in. Ideally, if we had our shit together, there can be some standard contact tracing format for QR and use programs to scan the data in to auto fill and submit your details to. My wife said to me yesterday "How do I use a QR Code?". Keep in mind, she's a civil engineer and one of the smartest people I know. I said "In what context?" because really, I didn't know either. "I have to scan this QR code if I go to my office. It's apparently posted at the door, and will let them know I've been there, for contact tracing." After a lot of playing around, we finally found out her phone's camera could automatically read QR codes. When we did, the QR code in question was a URL to a website with a form, requiring her to enter all her information. Not a fan of QR codes. It sounds like you aren't a fan of the way this QR code was used. I went to a restaurant the other day and it had one of those Coke Freestyle soda machines. There was a QR code on the screen which I scanned with my phone. This opened a website with drink choices. I put my cup under the dispenser and pushed the Barq's Root Beer button on the website on my phone which dispensed the drink from the machine. I was impressed. Wouldn’t pressing a button on a machine be easier? Normally I would completely agree, but perhaps in covid times it can prevent people from leaving germs on the buttons?? You’re right. It seems the perfect COVID-19 option. Completely touchless and fast. I was playing with a DMV kiosk and it took about half a second to recognize a key press. That’s a fair point! To me, this seems like a much more fragile and convoluted process than just having the buttons be on the machine itself. For one thing, the machine becomes unusable if there's any disruption to the network (wifi outage, dns issues, server maintenance, etc). I wonder what the security implications of this would be. Did the QR code connect your phone to the machine locally, or is there some menu out there for every freestyle coke machine on the open internet? If it's the latter I'd wager there's probably a way to remotely create a huuuge mess at some random five guys location. When I tested one of these machines out 2 months ago, a new QR code was generated each time the machine timed out to the home page. At this point, the old QR code and even the loaded webpage became useless. This means you can’t save the URL, leave the restaurant, then use it to control the machine later. In Denver, I've seen many restaurants that have eliminated their physical menu and switched to a QR Code at the table that brings you to their menu online. Not as nice, but definitely useful in these times. This one has been relevant for the last 10 years: https://picturesofpeoplescanningqrcodes.tumblr.com/ I feel like QR codes failed as a general UI tool. I get annoyed whenever I have to deal with one. They wee HUGE about 10 years ago. Now, not so much. I don’t know about other places, but they’re getting a lot of use in NYC right now as a method for viewing restaurant menus for outdoor dining. Here (Austin) it’s used in movie theaters that had seat food ordering and I gotta say thee
execution is flawless now that the camera has a built in QR. I think if QR could start over again and have support from the camera without the user needing to figure an app to use/get etc. things might have been different. I went to a restaurant the other day and it had one of those Coke Freestyle soda machines. There was a QR code on the screen which I scanned with my phone. This opened a website with drink choices. I put my cup under the dispenser and pushed the Barq's Root Beer button on the website on my phone which dispensed the drink from the machine. They also inspired similar usage for looking up a profile/resource in Snapchat, Spotify, Venmo, etc. Same in Germany. Also for payment with WeChat in China. It is big for payments in India too. I agreed with this take until 2020 and subsequent shutdowns of varying levels occurred in my U.S. state (Oregon) due to Covid-19. There are several eateries in my city utilizing them to great effect for menus and ordering as they have pivoted to outdoor and no-touch dining. And now that I have finally embraced them on this level, I notice and use them more: flyers around town, links to video tutorials on products I have purchased, local printed media, etc. That depends on where you are. Here every billboard and its cousin have a QR code. So does the beer at the local gas station. In Paris, most restaurants I visited recently have now done away with paper menus due to COVID and instead have QR codes on each table for the menus. Also Japan, being one of the earliest adopters, is still big on QR codes. And of course anything involving mobile and cryptocurrency transactions. And mobile payments in most of South and South East Asia. How so? They are effectively a `<link/>` IRL. I would say the UX is pretty good. I don't know how much easier it can get than pulling out your camera then taking a picture using the camera app. Their use has exploded since Covid in dense cities: I regularly have to use them to fill out pre-screening forms for facilities ("do you have symptoms X or have been in contact..."), see restaurant menus, pay bills (receipt at a restaurant has a QR code printed on the bottom, link to an online checkout on your phone, so you don't have to handle the POS machine), etc. I love QR codes, mostly because they are in-app now, and not restricted to one app or one company. The convenience is too good. I find them pretty easy to deal with now, they used to be annoying when you needed a specific app to scan them though. Apple is coming out with it's own version soon that should be popular (at least with iOS people): https://www.macrumors.com/2020/05/18/ios-14-leak-apple-qr-co... Really? I think they are quite a success, especially for making payments easy, for example I really like Apple's implementation of an animated secure QR code [1] for migrating data from one device to another. It looks like undulating point cloud, and it has no formal name (but is described in 2 Apple patents as an "invisible optical label") [1] https://apple.stackexchange.com/questions/301563/what-is-the... Have you seen the 1978 series "Connections" from James Burke? History of science and technology is full of these. A bit late to the party, but: Statistics and Robotics, Statistics and Digital Communications. Many filters in Robotics (Kalman Filter, Particle Filter) and many concepts in digital communication (Modulation Schemes, Filters, ...) are firmly rooted in statistics and require a decent applied understanding of the area. This may not be specific enough, but Elon Musk seems to have brought a software-engineering-like perspective to both SpaceX and Tesla, although the latter hasn't been an unqualified success in all respects, applying rapid iteration to a product and process while the process is running at scale has caused occasional blips: https://jalopnik.com/tesla-model-y-owners-have-found-home-de... We sometimes refer to software systems as held together with spit and baling wire, but it's never literally true. Others have noted the general influence in the opposite direction of the Toyota Production System on software engineering methodologies (most notably in that family of processes labeled "Lean"), but it is pretty clear Tesla could have stood to take a bit more direct influence from TPS on their assembly line. We haven't seen anything like that at SpaceX (despite much armchair prognostication), but then, even as they ramp up production and launches, each rocket is still somewhat a bespoke product, and no one is going to get in trouble for holding up the production schedule when a problem is noted. Each rocket is still very much a pet, rather than cattle. It remains to be seen if the pace of iteration will slack off (or start conforming to a more regular punctuated cadence) as SpaceX continues to ramp up their capacity. The approach to safety in the aviation industry is all about learning from mistakes. Whereas in Healthcare it is often more about winning an inevitable legal battle. There have been a few efforts recently to try and apply aviation safety lessons to healthcare. Interesting story on it here https://www.bbc.co.uk/programmes/p02x3vwh It's a good question. It can be quite infuriating to watch people build inferior monstrosities to solve problems that have well known and efficient solutions. I see examples everywhere of engineers trying to design custom algorithms for problems that can be trivially solved using linear programming or constraint programming. Not that custom algorithms can't beat generalized solvers...just that you aren't likely going by to be the person that designs it when you're a mid level engineer on a time budget and haven't researched any related work. Beyond that, I think cross breeding some insights from other fields could bring some efficiency improvements to problems that are squarely in the domain of computer science. OS scheduling, for example, has plenty of related people/materials scheduling parallels in the Operations Research field, and possibly some in Auction Theory as well. Same for load balancing (e.g. The Min Cost Multi-Commodity Flow problem). Frances Arnold, 2018 Nobel laureate for Chemistry, pioneered the use of directed evolution, which she conceived as an iterative lab process, to create enzymes with improved and/or novel functions, e.g. bacteria that produce the biofuel isobutanol. FRANCES H. ARNOLD
Nobel Prize in Chemistry 2018 https://www.nobelprize.org/womenwhochangedscience/stories/fr... Exploring protein fitness landscapes by directed evolution That's not really an unrelated field - thats just bioengineering and synthetic biology. Thats one of the more chemistry related chemistry novels in recent times actually, since the lack of a biology nobel has led to so many biology breakthroughs getting shoved into chemistry instead that its become a bit of a running joke. This year is no exception with CRISPR being looped under chemistry even though it woupd make the most sense under a separate new bio nobel. Yes, it is bioengineering. What Arnold brought to synthetic biology is an engineer's approach to iterative process. Rather than designing a new enzyme at the molecular level, Arnold constructed a process that would yield the design for her. Yes - but that's not interdisciplinary in the way you think it is. Designing a new enzyme at the molecular level is basically still not a thing that humans have figured out how to do. Her approach was revolutionary in the same way that the invention of a telescope was revolutionary to physics. You could try and say it was some grand synthesis of optics and astronomy, but that's a hell of a stretch, they're both physics. Evolution is just biology, and directed evolution is bioengineering. If you wanted to claim that Arnold invented modern bioengineering and laid the framework for synthetic biology, I could buy that. Just like the telescope laid the foundation for modern astronomy. But saying it's a perspective from an unrelated field doesn't make sense when it's one of the foundational defining cornerstones of this field. I see what you mean about this not being interdisciplinary. What Arnold did was to apply a biochemical process to a problem in biochemistry. > If you wanted to claim that Arnold invented modern bioengineering and laid the framework for synthetic biology, I could buy that. Well, I was just responding to the request for "some problems that were being experienced by one industry or domain but were solved elsewhere and were adopted by the industry plagued by the problem? Even better if the answer was really obvious for the people who solved it or solving it involved cross-domain knowledge". I thought of Arnold because I recalled that at the time she was awarded the Nobel prize she said she didn't know a lot of enzyme biochemistry, and, having come to the discipline late, could never have hoped to catch up on the detailed knowledge that her peers had spent a lifetime cultivating. [From Wikipedia]. What she did know about was engineering, having graduated in 1979 with a B.S. degree in mechanical and aerospace engineering from Princeton University, where she focused on solar energy research. In addition to the courses required for her major, she took classes in economics, Russian, and Italian, and envisioned herself as becoming a diplomat or CEO, even considering getting an advanced degree in international affairs. She took a year off from Princeton after her second year to travel to Italy and work in a factory that made nuclear reactor parts, then returned to complete her studies. Medical MRIs came out of the technology for radio astronomy: https://armaghplanet.com/astronomy-magnetic-resonance-imagin... Not completely unrelated fields but the Chen-Simons theory is a quantum field theory named after mathematicians Shiing-Shen Chern and James Simons(founder of Renaissance Technologies) who introduced the Chern-Simons form. Maybe control version? From software to a lot of other information industries. Also, not an industry, but I think that some mathematicians use category theory to translate insights from one domain to other. (Trying the recipe itself here to find a meta-recipe) I think an insight itself from category theory to a more general recipe could be: don't try to move "laterally" to other industry, but go first "up" from a specific solution to a more general interpretation, and see what more you can solve "down" from there. I don't think this is valid. Which other field is there where you can revert changes to a previous version? Certainly not hardware manufacturing. Versioning schematics and manuals is actually a great place for version control in manufacturing. One of my first projects out of school was a web app where our engineers could document the assembly steps for a given sub-assenbly, then string those sub-assemblies together into a manual for a specific product. The sub-assembly instructions would be revision controlled, and products would be revision controlled. The model was conceptually similar to git submodules. I don't recall exactly, but there was some legal authority (I think it was a U.S. state?) That uses Git for versioning laws, using pull requests as a way of usefully visualizing changes. Most domains where you produce knowledge as your main output would work. Version control is integrated in Office for example. Second order differential equations, they seem to pop up everywhere. Particularly relevant to classical mechanics. Also gives rise to the metal model of second order consequences of actions. The concept of bootstrapping. How to start something from nothing that eventually becomes self sustaining. Compound interest and exponential growth. Particularly relevant to investing as well as the personal ‘growth’ mentality. The natural exponent that was discovered from compound interest is everywhere in mathematics and physics. Duck tape: Originally came from the thought that seals on ammo boxes would cost soldiers precious time on the battlefield. Today it's used for seal and repair in everyday life. Space & rocket science research needed some speciality materials research, and led to heart valves replacement materials. Paul Bach-y-Rita used a grid of electrodes to stimulate the tongue, enabling a form of vision in blind people. Seeing With The Tongue – Paul Bach-Y-Rita – http://antonyhall.net/blog/seeing-with-the-tongue-paul-bach-... Zachary's Karate Club [1] was one of the early social networking approaches for identifying an eventual split in organization. Hedy Lamarr developed frequency-hopping, the forerunner of spread-spectrum communications used in WiFi, GPS, and Bluetooth, to control torpedoes in a way that avoided jamming by the Nazis, from her understanding of George Antheil's symphonic use of pianola rolls to synchronise multiple player pianos. https://www.wired.com/2011/08/0811hedy-lamar-george-antheil-... Hospitality and Healthcare Some of the "top" hospitals in the US are downright miserable to go to as a patient. Well, one classic is "Tai's formula is the Trapezoidal rule" https://pubmed.ncbi.nlm.nih.gov/7677819/ Where do you draw the line? Econophysics takes the natural science physics and tries to solve economic problems with it. There's also some application in the finance sector. Graph theory - organic chemistry I wonder about that stuff too and sometimes I think some problems are really like: I suspect a big one is "regular" engineering vs software engineering. Regular engineering has schedules and (apart from the apple campus) can put up a building in a predictable amount of time. Also testing and maintenance and more. Software creation has the characteristic that if you are doing the same thing over and over you are doing it wrong. Construction and manufacturing are much about doing the same thing the same way every time. This is why a lot of techniques do not translate well to/from software and other disciplines. I would argue that 90% of all software share the same premise. And considering nearly every construction process is behind time and over budget, the premise itself is flawed. Bee waggle dance found to be analogous to a 2d projection from a 6D flag manifold, complete with explanation of how the dance changes between a figure 8 shape to a round shape at a certain indicated distance- not much of a problem solved though, perhaps bringing more questions than answers... https://www.discovermagazine.com/planet-earth/quantum-honeyb... I'm not sure your question is correct given your premise. Is the breadth of knowledge related to cross domain discoveries? I would say bringing narrow skills like management or IT or marketing to a company is what you want to aim for. I'm not convinced cross domain issues exist at rates higher than internal domain issues. I think they are orders of magnitude lower.