When two protons collide, they release pyrotechnic jets of particles, the details of which can tell scientists something about the nature of physics and the fundamental forces that govern the universe.
Enormous particle accelerators such as the Large Hadron Collider can generate billions of such collisions per minute by smashing together beams of protons at close to the speed of light. Scientists then search through measurements of these collisions in hopes of unearthing weird, unpredictable behavior beyond the established playbook of physics known as the Standard Model.
Now MIT physicists have found a way to automate the search for strange and potentially new physics, with a technique that determines the degree of similarity between pairs of collision events. In this way, they can estimate the relationships among hundreds of thousands of collisions in a proton beam smashup, and create a geometric map of events according to their degree of similarity.
The researchers say their new technique is the first to relate multitudes of particle collisions to each other, similar to a social network.
“Maps of social networks are based on the degree of connectivity between people, and for example, how many neighbors you need before you get from one friend to another,” says Jesse Thaler, associate professor of physics at MIT. “It’s the same idea here.”
Thaler says this social networking of particle collisions can give researchers a sense of the more connected, and therefore more typical, events that occur when protons collide. They can also quickly spot the dissimilar events, on the outskirts of a collision network, which they can further investigate for potentially new physics. He and his collaborators, graduate students Patrick Komiske and Eric Metodiev, carried out the research at the MIT Center for Theoretical Physics and the MIT Laboratory for Nuclear Science. They detail their new technique this week in the journal Physical Review Letters.
Seeing the data agnostically
Thaler’s group focuses, in part, on developing techniques to analyze open data from the LHC and other particle collider facilities in hopes of digging up interesting physics that others might have initially missed.
“Having access to this public data has been wonderful,” Thaler says. “But it’s daunting to sift through this mountain of data to figure out what’s going on.”
Physicists normally look through collider data for specific patterns or energies of collisions that they believe to be of interest based on theoretical predictions. Such was the case for the discovery of the Higgs boson, the elusive elementary particle that was predicted by the Standard Model. The particle’s properties were theoretically outlined in detail but had not been observed until 2012, when physicists, knowing approximately what to look for, found signatures of the Higgs boson hidden amid trillions of proton collisions.
But what if particles exhibit behavior beyond what the Standard Model predicts, that physicists have no theory to anticipate?
Thaler, Komiske, and Metodiev have landed on a novel way to sift through collider data without knowing ahead of time what to look for. Rather than consider a single collision event at a time, they looked for ways to compare multiple events with each other, with the idea that perhaps by determining which events are more typical and which are less so, they might pick out outliers with potentially interesting, unexpected behavior.
“What we’re trying to do is to be agnostic about what we think is new physics or not,” says Metodiev. “We want to let the data speak for itself.”
Particle collider data are jam-packed with billions of proton collisions, each of which comprises individual sprays of particles. The team realized these sprays are essentially point clouds — collections of dots, similar to the point clouds that represent scenes and objects in computer vision. Researchers in that field have developed an arsenal of techniques to compare point clouds, for example to enable robots to accurately identify objects and obstacles in their environment.
Metodiev and Komiske utilized similar techniques to compare point clouds between pairs of collisions in particle collider data. In particular, they adapted an existing algorithm that is designed to calculate the optimal amount of energy, or “work” that is needed to transform one point cloud into another. The crux of the algorithm is based on an abstract idea known as the “earth’s mover’s distance.”
“You can imagine deposits of energy as being dirt, and you’re the earth mover who has to move that dirt from one place to another,” Thaler explains. “The amount of sweat that you expend getting from one configuration to another is the notion of distance that we’re calculating.”
In other words, the more energy it takes to rearrange one point cloud to resemble another, the farther apart they are in terms of their similarity. Applying this idea to particle collider data, the team was able to calculate the optimal energy it would take to transform a given point cloud into another, one pair at a time. For each pair, they assigned a number, based on the “distance,” or degree of similarity they calculated between the two. They then considered each point cloud as a single point and arranged these points in a social network of sorts.
Three particle collision events, in the form of jets, obtained from the CMS Open Data, form a triangle to represent an abstract "space of events." The animation depicts how one jet can be optimally rearranged into another.
The team has been able to construct a social network of 100,000 pairs of collision events, from open data provided by the LHC, using their technique. The researchers hope that by looking at collision datasets as networks, scientists may be able to quickly flag potentially interesting events at the edges of a given network.
“We’d like to have an Instagram page for all the craziest events, or point clouds, recorded by the LHC on a given day,” says Komiske. “This technique is an ideal way to determine that image. Because you just find the thing that’s farthest away from everything else.”
Typical collider datasets that are made publicly available normally include several million events, which have been preselected from an original chaos of billions of collisions that occurred at any given moment in a particle accelerator. Thaler says the team is working on ways to scale up their technique to construct larger networks, to potentially visualize the “shape,” or general relationships within an entire dataset of particle collisions.
In the near future, he envisions testing the technique on historical data that physicists now know contain milestone discoveries, such as the first detection in 1995 of the top quark, the most massive of all known elementary particles.
“The top quark is an object that gives rise to these funny, three-pronged sprays of radiation, which are very dissimilar from typical sprays of one or two prongs,” Thaler says. “If we could rediscover the top quark in this archival data, with this technique that doesn’t need to know what new physics it is looking for, it would be very exciting and could give us confidence in applying this to current datasets, to find more exotic objects.”
This research was funded, in part, by the U.S. Department of Energy, the Simons Foundation, and the MIT Quest for Intelligence.
The Media Lab has always been a place for students and researchers to work on building the future they want to live in. As such, many innovative companies have spun out of its multidisciplinary working spaces. Since 2013, however, the lab has taken a more deliberate approach to helping entrepreneurs create successful companies, with the E14 Fund.
The fund is designed to supercharge the Media Lab’s impact in the world by supporting the imaginative people that have walked its hallways. That support can come in the form of early-stage investments and fellowships that give students, alumni, and researchers a “runway” to turn their ideas into businesses. But E14 — named after the location of the Media Lab building — is much more than an investment fund.
“Even though the E14 Fund has the word ‘fund’ at the end of it, first and foremost we’re a community-builder,” Managing Partner Habib Haddad says. “When you put that first, you actually become a stronger venture fund by being an authentic community supporter.”
Haddad, fellow Managing Partner Calvin Chin, and their team of venture partners and advisors believe their efforts help turn the innovative ideas of the Media Lab’s community into companies that benefit society.
“I helped launch [E14] because I felt it addressed an important need in offering a runway to graduating students who wanted to deploy their research through a new company, and it's been successful at doing that,” Media Lab Director Joi Ito says. “But it's also become a really good community and a networking hub for Media Lab students, researchers, and alumni — not just ones working on building a company.”
A bold founding mission
The Media Lab is known for fostering a culture that blends future-focused research with current problems to inspire creative solutions. The lab’s community has also developed a reputation for pursuing bold, new ideas at rapid — some might say entrepreneurial — speed.
Throughout 2013, Ito was looking for a way to help students continue following the Media Lab’s “deploy” mantra after graduation. Ito believed a venture fund, if structured correctly, could increase the Media Lab’s impact on the world while also furthering its academic mission, by helping students focus on their coursework without worrying about losing support when they graduate.
In keeping with the lab’s experimental ethos, E14 began as what Haddad calls a “prototype fund” of $2 million in 2013. The money was used to provide both equity and nonequity funding and mentoring to recent graduates in the early stages of starting companies. Some of the carried interest from the fund would be donated back to MIT to support more research at the Media Lab (a practice that carries on today).
A handful of the earliest investments went on to raise much larger funding rounds and find commercial success, including the manufacturing app platform Tulip, robotic furniture company Ori, and urban technology company Soofa. With an eye toward building a community, E14 also provided slightly later-stage funding to Media Lab spinouts Formlabs and Affectiva.
Even as the early investments had success, it became clear the Media Lab would benefit from an even more involved investment team. E14’s next fund, launched at the end of 2017, raised $37 million.
“As that prototype fund was evolving, it turned out some of these companies needed much more capital and support,” Haddad recalls. “We said, ‘These are all great companies, but for each one, there’s a few others that could have been great spinoffs that didn’t have the right funding or support.’”
Since then, the E14 Fund has been advising everyone from Media Lab alumni starting their fourth company to current students nearing graduation. It’s also helped the lab’s member companies connect with researchers and created a more active community overall.
A pillar in the community
Today the E14 team regularly hosts events and gatherings for members of the Media Lab. The lab’s semiannual member week now includes a startup showcase, and alumni are becoming a more common sight in the hallways.
E14 also works with the other entrepreneurial resources on campus, giving students advice on what programs are worth exploring depending on what stage their idea is in.
The E14’s average funding is between $500,000 and $1 million, while its fellowship program offers a lower-risk way of pursuing the commercialization of an idea. Some of E14’s latest investments include ThruWave, a company using millimeter wave sensors to see through packaging; Figur8, which has developed a system that captures three-dimensional skeletal movement and muscle output; Wise Systems, which uses fleet dispatch and routing software to optimize deliveries; Canopy, which uses machine learning to preserve privacy on the internet; and Kiwi, which offers a robotic cropdusting system for distributing chemicals and seeds on precise plots of land.
By investing exclusively in Media Lab founders with a new technology, the E14 team spends far more time with founders — sometimes years, as students work toward graduation or alumni consider starting companies — compared to traditional investment funds, and helps researchers and scientists develop a more entrepreneurial mindset over time.
“We like to say we don’t have all the answers, but we ask a lot of questions,” Chin says. “It’s really just brainstorming and thinking about the kind of company they’re going to build, the way they’re going to build it, and the market they’re going to address. A lot of times in that process we’ll be making introductions. We’re trying to add value even if we’re not invested.”
In fact, beyond making investments, the E14 Fund bears little resemblance to other venture capital funds, in which partners typically guard their time like gold and focus on narrow areas of industries. The members of E14, conversely, say their areas of investment are entirely dependent on the interests of members of the Media Lab.
“Our model is serving the community, and then finding companies as a subset of that community,” Chin says. “We just always want to be listening to the community. If the community’s research interests or business interests go a certain way, we want to be there with the resources the community needs. That’s what we’re excited about going forward.”
As embryos develop, they follow predetermined patterns of tissue folding, so that individuals of the same species end up with nearly identically shaped organs and very similar body shapes.
MIT scientists have now discovered a key feature of embryonic tissue that helps explain how this process is carried out so faithfully each time. In a study of fruit flies, they found that the reproducibility of tissue folding is generated by a network of proteins that connect like a fishing net, creating many alternative pathways that tissues can use to fold the right way.
“What we found is that there’s a lot of redundancy in the network,” says Adam Martin, an MIT associate professor of biology and the senior author of the study. “The cells are interacting and connecting with each other mechanically, but you don’t see individual cells taking on an all-important role. This means that if one cell gets damaged, other cells can still connect to disparate parts of the tissue.”
To uncover these network features, Martin worked with Jörn Dunkel, an MIT associate professor of physical applied mathematics and an author of the paper, to apply an algorithm normally used by astronomers to study the structure of galaxies.
Hannah Yevick, an MIT postdoc, is the lead author of the study, which appears today in Developmental Cell. Graduate student Pearson Miller is also an author of the paper.
A safety net
During embryonic development, tissues change their shape through a process known as morphogenesis. One important way tissues change shape is to fold, which allows flat sheets of embryonic cells to become tubes and other important shapes for organs and other body parts. Previous studies in fruit flies have shown that even when some of these embryonic cells are damaged, sheets can still fold into their correct shapes.
“This is a process that’s fairly reproducible, and so we wanted to know what makes it so robust,” Martin says.
In this study, the researchers focused on the process of gastrulation, during which the embryo is reorganized from a single-layered sphere to a more complex structure with multiple layers. This process, and other morphogenetic processes similar to fruit fly tissue folding, also occur in human embryos. The embryonic cells involved in gastrulation contain in their cytoplasm proteins called myosin and actin, which form cables and connect at junctions between cells to form a network across the tissue. Martin and Yevick had hypothesized that the network of cell connectivity might play a role in the robustness of the tissue folding, but until now, there was no good way to trace the connections of the network.
To achieve that, Martin’s lab joined forces with Dunkel, who studies the physics of soft surfaces and flowing matter — for example, wrinkle formation and patterns of bacterial streaming. For this study, Dunkel had the idea to apply a mathematical procedure that can identify topological features of a three-dimensional structure, analogous to ridges and valleys in a landscape. Astronomers use this algorithm to identify galaxies, and in this case, the researchers used it to trace the actomyosin networks across and between the cells in a sheet of tissue.
“Once you have the network, you can apply standard methods from network analysis — the same kind of analysis that you would apply to streets or other transport networks, or the blood circulation network, or any other form of network,” Dunkel says.
Among other things, this kind of analysis can reveal the structure of the network and how efficiently information flows along it. One important question is how well a network adapts if part of it gets damaged or blocked. The MIT team found that the actomyosin network contains a great deal of redundancy — that is, most of the “nodes” of the network are connected to many other nodes.
This built-in redundancy is analogous to a good public transit system, where if one bus or train line goes down, you can still get to your destination. Because cells can generate mechanical tension along many different pathways, they can fold the right way even if many of the cells in the network are damaged.
“If you and I are holding a single rope, and then we cut it in the middle, it would come apart. But if you have a net, and cut it in some places, it still stays globally connected and can transmit forces, as long as you don’t cut all of it,” Dunkel says.
The researchers also found that the connections between cells preferentially organize themselves to run in the same direction as the furrow that forms in the early stages of folding.
“We think this is setting up a frame around which the tissue will adopt its shape,” Martin says. “If you prevent the directionality of the connections, then what happens is you can still get folding but it will fold along the wrong axis.”
Although this study was done in fruit flies, similar folding occurs in vertebrates (including humans) during the formation of the neural tube, which is the precursor to the brain and spinal cord. Martin now plans to apply the techniques he used in fruit flies to see if the actomyosin network is organized the same way in the neural tube of mice. Defects in the closure of the neural tube can lead to birth defects such as spina bifida.
“We would like to understand how it goes wrong,” Martin says. “It’s still not clear whether it’s the sealing up of the tube that’s problematic or whether there are defects in the folding process.”
The research was funded by the National Institute of General Medical Sciences and the James S. McDonnell Foundation.
Years before he set foot on the MIT campus, Kieran P. Dolan participated in studies conducted at MIT's Nuclear Reactor Laboratory (NRL). As an undergraduate student majoring in nuclear engineering at the University of Wisconsin at Madison, Dolan worked on components and sensors for MIT Reactor (MITR)-based experiments integral to designing fluoride-salt-cooled high-temperature nuclear reactors, known as FHRs.
Today, as a second-year doctoral student in MIT's Department of Nuclear Science and Engineering, Dolan is a hands-on investigator at the NRL, deepening his research engagement with this type of next-generation reactor.
"I've been interested in advanced reactors for a long time, so it's been really nice to stay with this project and learn from people working here on-site," says Dolan.
This series of studies on FHRs is part of a multiyear collaboration among MIT, the University of Wisconsin at Madison, and the University of California at Berkeley, funded by an Integrated Research Project (IRP) Grant from the U.S. Department of Energy (DOE). The nuclear energy community sees great promise in the FHR concept because molten salt transfers heat very efficiently, enabling such advanced reactors to run at higher temperatures and with several unique safety features compared to the current fleet of water-cooled commercial reactors.
"Molten salt reactors offer an approach to nuclear energy that is both economically viable and safe," says Dolan.
For the purposes of the FHR project, the MITR reactor simulates the likely operating environment of a working advanced reactor, complete with high temperatures in the experimental capsules. The FHR concept Dolan has been testing envisions billiard-ball-sized composites of fuel particles suspended within a circulating flow of molten salt — a special blend of lithium fluoride and beryllium fluoride called flibe. This salt river constantly absorbs and distributes the heat produced by the fuel's fission reactions.
But there is a formidable technical challenge to the salt coolants used in FHRs. "The salt reacts with the neutrons released during fission, and produces tritium," explains Dolan. "Tritium is one of hydrogen’s isotopes, which are notorious for permeating metal." Tritium is a potential hazard if it gets into water or air. "The worry is that tritium might escape as a gas through an FHR's heat exchanger or other metal components."
There is a potential workaround to this problem: graphite, which can trap fission products and suck up tritium before it escapes the confines of a reactor. "While people have determined that graphite can absorb a significant quantity of hydrogen, no one knows with certainty where the tritium is going to end up in the reactor,” says Dolan. So, he is focusing his doctoral research on MITR experiments to determine how effectively graphite performs as a sponge for tritium — a critical element required to model tritium transport in the complete reactor system.
"We want to predict where the tritium goes and find the best solution for containing it and extracting it safely, so we can achieve optimal performance in flibe-based reactors," he says.
While it's early, Dolan has been analyzing the results of three MITR experiments subjecting various types of specialized graphite samples to neutron irradiation in the presence of molten salt. "Our measurements so far indicate a significant amount of tritium retention by graphite," he says. "We're in the right ballpark."
Dolan never expected to be immersed in the electrochemistry of salts, but it quickly became central to his research portfolio. Enthused by math and physics during high school in Brookfield, Wisconsin, he swiftly oriented toward nuclear engineering in college. "I liked the idea of making useful devices, and I was especially interested in nuclear physics with practical applications, such as power plants and energy," he says.
At UW Madison, he earned a spot in an engineering physics material research group engaged in the FHR project, and he assisted in purifying flibe coolants, designing and constructing probes for measuring salt's corrosive effect on reactor parts, and experimenting on the electrochemical properties of molten fluoride salts. Working with Exelon Generation as a reactor engineer after college convinced him he was more suited for research in next-generation projects than in the day-to-day maintenance and operation of a commercial nuclear plant.
"I was interested in innovation and improving things," he says. "I liked being part of the FHR IRP, and while I didn't have a passion for electrochemistry, I knew it would be fun working on a solution that could advance a new type of reactor."
Familiar with the goals of the FHR project, MIT facilities, and personnel, Dolan was able to jump rapidly into studies analyzing MITR's irradiated graphite samples. Under the supervision of Lin-wen Hu, his advisor and NRL research director, as well as MITR engineers David Carpenter and Gordon Kohse, Dolan came up to speed in reactor protocol. He's found on-site participation in experiments thrilling.
"Standing at the top of the reactor as it starts and the salt heats up, anticipating when the tritium comes out, manipulating the system to look at different areas, and then watching the measurements come in — being involved with that is really interesting in a hands-on way," he says.
For the immediate future, "the main focus is getting data," says Dolan. But eventually "the data will predict what happens to tritium in different conditions, which should be the main driving force determining what to do in actual commercial FHR reactor designs."
For Dolan, contributing to this next phase of advanced reactor development would prove the ideal next step following his doctoral work. This past summer, Dolan interned at Kairos Power, a nuclear startup company formed by the UC Berkeley collaborators on two DOE-funded FHR IRPs. Kairos Power continues to develop FHR technology by leveraging major strategic investments that the DOE has made at universities and national laboratories, and has recently started collaborating with MIT.
"I've built up a lot of experience in FHRs so far, and there's a lot of interest at MIT and beyond in reactors using molten salt concepts," he says. "I will be happy to apply what I've learned to help accelerate a new generation of safe and efficient reactors."
For patients with kidney failure who need dialysis, removing fluid at the correct rate and stopping at the right time is critical. This typically requires guessing how much water to remove and carefully monitoring the patient for sudden drops in blood pressure.
Currently there is no reliable, easy way to measure hydration levels in these patients, who number around half a million in the United States. However, researchers from MIT and Massachusetts General Hospital have now developed a portable sensor that can accurately measure patients’ hydration levels using a technique known as nuclear magnetic resonance (NMR) relaxometry.
Such a device could be useful for not only dialysis patients but also people with congestive heart failure, as well as athletes and elderly people who may be in danger of becoming dehydrated, says Michael Cima, the David H. Koch Professor of Engineering in MIT’s Department of Materials Science and Engineering.
“There’s a tremendous need across many different patient populations to know whether they have too much water or too little water,” says Cima, who is the senior author of the study and a member of MIT’s Koch Institute for Integrative Cancer Research. “This is a way we could measure directly, in every patient, how close they are to a normal hydration state.”
The portable device is based on the same technology as magnetic resonance imaging (MRI) scanners but can obtain measurements at a fraction of the cost of MRI, and in much less time, because there is no imaging involved.
Lina Colucci, a former graduate student in health sciences and technology, is the lead author of the paper, which appears in the July 24 issue of Science Translational Medicine. Other authors of the paper include MIT graduate student Matthew Li; MGH nephrologists Kristin Corapi, Andrew Allegretti, and Herbert Lin; MGH research fellow Xavier Vela Parada; MGH Chief of Medicine Dennis Ausiello; and Harvard Medical School assistant professor in radiology Matthew Rosen.
Cima began working on this project about 10 years ago, after realizing that there was a critical need for an accurate, noninvasive way to measure hydration. Currently, the available methods are either invasive, subjective, or unreliable. Doctors most frequently assess overload (hypervolemia) by a few physical signs such as examining the size of the jugular vein, pressing on the skin, or examining the ankles where water might pool.
The MIT team decided to try a different approach, based on NMR. Cima had previously launched a company called T2 Biosystems that uses small NMR devices to diagnose bacterial infections by analyzing patient blood samples. One day, he had the idea to use the devices to try to measure water content in tissue, and a few years ago, the researchers got a grant from the MIT-MGH Strategic Partnership to do a small clinical trial for monitoring hydration. They studied both healthy controls and patients with end-stage renal disease who regularly underwent dialysis.
One of the main goals of dialysis is to remove fluid in order bring patients to their “dry weight,” which is the weight at which their fluid levels are optimized. Determining a patient’s dry weight is extremely challenging, however. Doctors currently estimate dry weight based on physical signs as well as through trial-and-error over multiple dialysis sessions.
The MIT/MGH team showed that quantitative NMR, which works by measuring a property of hydrogen atoms called T2 relaxation time, can provide much more accurate measurements. The T2 signal measures both the environment and quantity of hydrogen atoms (or water molecules) present.
“The beauty of magnetic resonance compared to other modalities for assessing hydration is that the magnetic resonance signal comes exclusively from hydrogen atoms. And most of the hydrogen atoms in the human body are found in water molecules,” Colucci says.
The researchers used their device to measure fluid volume in patients before and after they underwent dialysis. The results showed that this technique could distinguish healthy patients from those needing dialysis with just the first measurement. In addition, the measurement correctly showed dialysis patients moving closer to a normal hydration state over the course of their treatment.
Furthermore, the NMR measurements were able to detect the presence of excess fluid in the body before traditional clinical signs — such as visible fluid accumulation below the skin — were present. The sensor could be used by physicians to determine when a patient has reached their true dry weight, and this determination could be personalized at each dialysis treatment.
The researchers are now planning additional clinical trials with dialysis patients. They expect that dialysis, which currently costs the United States more than $40 billion per year, would be one of the biggest applications for this technology. This kind of monitoring could also be useful for patients with congestive heart failure, which affects about 5 million people in the United States.
“The water retention issues of congestive heart failure patients are very significant,” Cima says. “Our sensor may offer the possibility of a direct measure of how close they are to a normal fluid state. This is important because identifying fluid accumulation early has been shown to reduce hospitalization, but right now there are no ways to quantify low-level fluid accumulation in the body. Our technology could potentially be used at home as a way for the care team to get that early warning.”
Sahir Kalim, a nephrologist and assistant professor of medicine at Massachusetts General Hospital, described the MIT approach as “highly novel.”
“The development of a bedside device that can accurately inform providers about how much fluid a patient should ideally have removed during their dialysis treatment would likely be one of the most significant developments in dialysis care in many years,” says Kalim, who was not involved in the study. “Colucci and colleagues have made a promising innovation that may one day yield this impact.”
In their study of the healthy control subjects, the researchers also incidentally discovered that they could detect dehydration. This could make the device useful for monitoring elderly people, who often become dehydrated because their sense of thirst lessens with age, or athletes taking part in marathons or other endurance events. The researchers are planning future clinical trials to test the potential of their technology to detect dehydration.
The research was funded by the MGH-MIT Strategic Partnership Grand Challenge, the Air Force Medical Services/Institute of Soldier Nanotechnologies, the National Science Foundation Graduate Research Fellowships Program, the National Institute of Biomedical Imaging and Bioengineering, the Koch Institute Support (core) Grant from the National Cancer Institute, and Harvard University.
Following their graduation in 2016, two dual-degree students from the MIT Center for Real Estate (CRE) and the Department of Architecture — Kun Qian MSRED '16, MArch ’16 and Marwan Aboudib MSRED '16, MArch ’16 — asked Professor Dennis Frenchman if he would join with their firm, Tekuma, to create an international design practice.
“They came to me and said, ‘Look, we have this project opportunity in Jinan, [China],’” says Frenchman, the Class of ’22 Professor of Urban Design and Planning and now CRE director. “Would you like to join us?”
Frenchman said yes. The resulting urban design and innovation studio, Tekuma Frenchman, practices worldwide, applying Frenchman’s research at MIT to many scales of intervention — from planning cities for millions in China to art and cultural installations in the Middle East and Boston, Massachusetts. In addition to Qian, Aboudib, and Frenchman, the partnership includes urban designer Naomi Hebert and a staff of 10 working in Cambridge, Massachusetts; Dubai; and Beijing.
The firm’s projects include design of Seoul Digital Media City in South Korea; the Digital Mile in Zaragoza, Spain; Ciudad Creativa Digital, Guadalajara, Mexico; Media City: UK in England; Twofour54 in Abu Dhabi; Jinan North New District and Chanqing University City in China; and, more recently, projects in cities across the Middle East.
In 2018, Tekuma Frenchman won the Shenzhen New Marine City International Design Competition in China. Their design, titled “Ocean Edge,” will be home to 50,000 people on 5.5 square kilometers of reclaimed land.
The competition was part of China’s 13th Five-Year Plan for the Development of the National Marine Economy, which aims to advance manufacturing industry along China’s southern coast.
Shenzhen is a city of more than 12 million in the Guangdong province of southern China, where Hong Kong links to China’s mainland. Shenzhen was the first special economic zone in China that encouraged outside investment and is the home to high-tech industries in computer science, robotics, artificial intelligence, and data storage.
The urban design competition for Shenzhen New Marine City received over 140 design submissions from international firms. A jury of nine design professionals and senior academics selected Tekuma Frenchman’s proposal. The organizers wanted a scheme that would be both innovative and operable, and become a world-class demonstration site for the future of marine economy and sustainability.
Frenchman’s winning scheme integrates marine ecology, marine industry, marine culture, and coastal landscape, providing the design framework for a visionary development. Land reclamation for Ocean Edge has already begun, and key aspects of the development will be in place by 2022. It is expected that the project will take about 20 years to complete.
Piers and mangroves as a solution
The waterfront site poses many challenges. To minimize the use of fill and disruption of water flow, Tekuma Frenchman decided to put parts of the city on piers, islands, and autonomous floating structures.
A 1-kilometer central entertainment pier will connect Shenzhen’s convention center with the ocean and anchor recreation areas along the waterfront. This pier and boardwalk area includes a ferry terminal, port offices, deep-sea aquarium, theater, cinemas, clubs, water sports, seafood restaurants, and specialty retail shops.
Tekuma Frenchman’s design will regenerate an indigenous mangrove forest to protect the shoreline from waves, retain soil, support biodiversity, and help clean the water. The design also provides a habitat for fish. The growth of the forest will be monitored by sensors controlling the mix of freshwater runoff with salt water, ensuring an ideal habitat for optimum growth of marine fauna and flora. In this way, Ocean Edge both senses and responds to the natural environment.
Sea-level rise and storm surge from the South China Sea is a concern in Shenzhen. Ocean Edge helps prevent flooding by using the regenerated mangrove forest as a natural protection from storm surge. “There’s a sustainable matrix into which this contemporary city is built,” Frenchman says.
Promoting ocean industry
The heartbeat of the city is a new industry cluster dedicated to deep-sea exploration and resource extraction, using autonomous undersea vehicles.
Robotic vehicles will be researched, developed, and deployed from Ocean Edge to, for example, scavenge manganese nodules from the ocean floor or to manage fish and agricultural production. A spine of development, accessible directly to the water, will house private labs, academic research institutions, and public agencies devoted to understanding and exploiting the resources of the South China Sea.
Production and manufacturing for this industry cluster are woven in with housing and entertainment.
“Most of the new cities out there are built as places to consume — shopping, eating, culture,” Frenchman says. “What’s interesting about the Ocean Edge design in Shenzhen is its focus on making a productive city with emerging 21st century industries and lifestyles at its heart. Ocean Edge will become a key link in the chain of manufacturing cities which make up the Guangzhou-Shenzhen Innovation Corridor.”
“Over the years we have been researching and implementing a new methodology to design cities that celebrate production, making places that are human-centric and productive,” says Kun Qian. “We believe that the future is moving toward a productive urbanism, where companies from all economic sectors also participate in the shaping of our public realm and creating unique experiences for people. The Ocean Edge proposal is a great testimony to this approach.”
“What differentiates our firm is that our work goes beyond design,” says partner Marwan Aboudib. “The key is our integration of design with real estate economics and technology. Our understanding and ability to bring those domains together enables us to create more vibrant cities in which people can excel. We make it possible for cities to thrive, which creates stronger returns for businesses and residents.”
For a video of Tekuma Frenchman’s winning design, see Vimeo.
MIT professors David Sontag and Peter Szolovits don’t assign a textbook for their class, 6.S897HST.956 (Machine Learning for Healthcare), because there isn’t one. Instead, students read scientific papers, solve problem sets based on current topics like opioid addiction and infant mortality, and meet the doctors and engineers paving the way for a more data-driven approach to health care. Jointly offered by MIT’s Department of Electrical Engineering and Computer Science (EECS) and the Harvard-MIT program in Health Sciences Technology, the class is one of just a handful offered across the country.
“Because it’s a new field, what we teach will help shape how AI is used to diagnose and treat patients,” says Irene Chen, an EECS graduate student who helped design and teach the course. “We tried to give students the freedom to be creative and explore the many ways machine learning is being applied to health care.”
Two-thirds of the syllabus this spring was new. Students were introduced to the latest machine-learning algorithms for analyzing doctors’ clinical notes, patient medical scans, and electronic health records, among other data. Students also explored the risks of using automated methods to explore large, often messy observational datasets, from confusing correlation with causation to understanding how AI models can make bad decisions based on biased data or faulty assumptions.
With all of the hype around AI, the course had more takers than seats. After 100 students showed up on the first day, students were assigned a quiz to test their knowledge of statistics and other prerequisites. That helped whittle the class down to 70. Michiel Bakker, a graduate student at the MIT Media Lab, made the cut and says the course gave him medical concepts that most engineering courses don’t provide.
“In machine learning, the data are either images or text,” he says. “Here we learned the importance of combining genetic data with medical images with electronic health records. To use machine learning in health care you have to understand the problems, how to combine techniques and anticipate where things could go wrong.”
Most lectures and homework problems focused on real world scenarios, drawing from MIT’s MIMIC critical care database and a subset of the IBM MarketScan Research Databases focused on insurance claims. The course also featured regular guest lectures by Boston-area clinicians. In a reversal of roles, students held office hours for doctors interested in integrating AI into their practice.
“There are so many people in academia working on machine learning, and so many doctors at hospitals in Boston,” says Willie Boag, an EECS graduate student who helped design and teach the course. “There’s so much opportunity in fostering conversation between these groups.”
In health care, as in other fields where AI has made inroads, regulators are discussing what rules should be put in place to protect the public. The U.S. Federal Drug Administration recently released a draft framework for regulating AI products, which students got to review and comment on, in class and in feedback published online in the Federal Register.
Andy Coravos, a former entrepreneur in residence at the FDA, now CEO of ElektraLabs in Boston, helped lead the discussion and was impressed by the quality of the comments. “Many students identified test cases relevant to the current white paper, and used those examples to draft public comments for what to keep, add, and change in future iterations,” she says.
The course culminated in a final project in which teams of students used the MIMIC and IBM datasets to explore a timely question in the field. One team analyzed insurance claims to explore regional variation in screening patients for early-stage kidney disease. Many patients with hypertension and diabetes are never tested for chronic kidney disease, even though both conditions put them at high risk. The students found that they could predict fairly well who would be screened, and that screening rates diverged most between the southern and northeastern United States.
“If this work were to continue, the next step would be to share the results with a doctor and get their perspective,” says team member Matt Groh, a PhD student at the MIT Media Lab. “You need that cross-disciplinary feedback.”
The MIT-IBM Watson AI Lab took the trouble of making the anonymized data available and providing cloud computing support to students on the IBM cloud out of an interest in helping to educate the next generation of scientists and engineers, says Kush Varshney, principal research staff member and manager at IBM Research. “Health care is messy and complex, which is why there are no substitutes for working with real-world data,” he says.
Szolovits agrees. Using synthetic data would have been easier but far less meaningful. “It’s important for students to grapple with the complexities of real data,” he says. “Any researcher developing automated techniques and tools to improve patient care needs to be sensitive to its many nuances.”
In a recent recap on Twitter, Chen gave shout-outs to the students, guest lecturers, professors, and her fellow teaching assistant. She also reflected on the joys of teaching. “Research is rewarding and often fun, but helping someone see your field with fresh eyes is insanely cool.”