Seth Teller's Random Ideas

This is a collection of random ideas. All of them require more thought and engineering to be developed. If you love or hate any of them, please email me at
On the other hand, if you get rich from any of them, please send me a check or...

make a donation!

These are arranged in reverse-chronological order, i.e., newest ideas first.

2014, Heat Wick for Global Warming What is the basic problem of global warming: there's too much heat building up in our atmosphere, making it turbulent. If we could somehow chill the earth -- without changing our present practices -- we could keep our status quo growth going somewhat longer. Where would the heat go? Along a human-made wick, of course, oriented perpendicularly to the earth's atmosphere, so that we can direct the heat away as efficiently as possible from that exceedingly thin layer. But where would the heat go? There are only two possibilities -- to outer space, beyond the edge of our atmosphere, or to the depths of earth itself, where we would be relatively insensitive to its accumulation there. Why not make it a top priority to find a way to construct such "wicks," and place one at each Pole. What sort of technology could possibly constitute a wick?

2014, Browser Plug-In for Product Recalls: The Consumer Product Safety Commission does not have a way to publicize its recall notices effectively. Suppose instead they funded development of a browser plug-in that would detect text or images associated with a recalled product, and visually highlight those items whenever the user happened to view them. The trick would be making the false positive rate acceptably low. But even if it were a bit high, users could ignore spurious warnings. And of course the CPSC would have to talk browser developers into including the plug-in by default, or otherwise publicizing it.

2014, Hybrid Representational Voting: Here is a way for all voters to share their vote fairly with that of their representative. Fairness here is designed to allow each citizen to exercise 1/Nth of a vote, or decline to vote; the fraction of the total vote declined, since it has not been expressed directly by any individual voter, goes to the citizens' elected representative to exercise by "default." For example suppose a district of 100 people puts an issue to a vote, during which 25 citizens vote YES, 30 citizens vote NO; and 45 citizens do not vote. The representative's vote would then have weight 0.45, compared to 0.55 for the total votes of the voters. Suppose M <= N citizens choose to vote, with the "ith" citizen casting vote V_i in {0,1}; the remaining (1-M) citizens choose not to vote; and the representative casts the vote R in {0,1}. The hybrid vote is then just the appropriate weighted average: Vote = ( sum[i=1..M] { V_i } + R * (1-M) ) / N Above, the balance of power between the representative and her or his citizens is N:1 -- the representative is much more powerful than the people. But that power can be dialed down or up simply by adjusting the relative weights of the votes. Moreover, we can make fraud difficult by generating a secure email, memorializing the vote, to the unique address owned by the voter. The address fits the bill perfectly, as it can be easily associated with the holder of that social security number. (It could be forwarded to the holder's routine inbox, so that one wouldn't have to go check another email server to see if votes had been cast in one's name.)

2014, Objective Measure of Leadership Monitor the evolution of the real-time voting display. Measure how often each representative is the first, or early, in expressing a position which ends up winning the vote. For party-line voting, no one would be expected to be first more often than anyone else, but for issues requiring leadership, the representatives leading others would tend to appear early in voting.

2012, Electrostatic Vacuuming: Suppose we could put an electrical charge on all the dust and dirt in the carpet or on the floor, and an opposite charge on the vacuum cleaner head; wouldn't this make the head that much more effective at pulling the material off of the floor?

2011, Cued speech recognition for people with disabilities: Suppose someone with a disability is attempting to use a speech recognition system for computer control or environmental control in a very noisy environment. The system could "cue" the user visually, by displaying words or phrases with utterances that are predicted to have relatively high SNR for that environment! The cued terms could change rapidly over time as the user's surroundings change.

2011, Foveating Camera: Modern cameras provide either high-resolution and a small field of view, or low resolution and a wide field of view. Trying to do both high-res and high-FOV yields an infeasibly high bandwidth requirement; there are too many pixels to extract from the camera in any reasonable shuttering time. A "foveating camera" could give the best of both worlds, without any moving parts. The idea would be to fabricate a high-resolution sensor array, but arrange the internal I/O circuitry in a tree structure to admit either high-res extraction of any small region, or low-res extraction of any large region, both subject to some upper limit on I/O bandwidth. Such a camera could foveate on elements of its surroundings in much the way a human eye does.

2011, Interstate Water Pipelines for Flood Reduction: Will we ever get to the point where we can control weather? Suppose we can't (or at least, don't do so quickly) and want to address the problem that some areas of the country have drought, and some have flood. Can we build a water pipeline with the capacity to ease floods in flood zones? How much volume are we talking about? How would we pump it? Would sediment be an issue? I'm thinking something that would run along the interstate highway system, with enough capacity to de-inundate a flooded city in hours (yes, it would have to be a very large pipe).

2011, Living Wage Verification: My wife and I were talking about how so many workers here and in the third world etc. live in terrible conditions, with low wages, etc. We agreed that we would be willing to pay extra for goods that came with some guarantee that all the workers involved in producing them had been paid a living wage. It's hard to see how to implement this kind of guarantee with traditional means -- each supplier/vendor would somehow have to prove that they had paid their employees etc. back through the chain. Even if there were paid human monitors throughout the system, how could corruption be prevented? Suppose we assume that every worker has a mobile phone or some other device that's cryptographically "identified" with that worker. Is there some way for a purchaser, at any level of the supply chain, to secure a guarantee that everyone below that level has been paid fairly? The base case would be the worker at the "leaf" who does some labor and is paid some amount, in cash. Somehow that worker would give the payer a "proof" that fair payment had been made, which the payer could then aggregate and pass upward when s/he was paid in turn. Of course the method would have to be designed so as not to admit coercion, corruption etc.

2010, Health-Care Provider's Hand-Washing Medallion: Develop a small device worn around the neck like a medallion. It observes the patients with whom a doctor or other health-care provider interacts, and detects hand-washing activity. It buzzes a warning when the doctor encounters a new patient and has not washed his/her hands in the interim.

2010, Rapidly-deployable oil-slick containment wall (prompted by the Gulf of Mexico underground oil leak): When an underwater source is spewing oil that floats to the surface and spreads, quickly surround it with a neutrally-buoyant wall that extends several meters vertically above and below the ocean surface. Close the wall on itself to make a circle around the oil. Then set the inside on fire, or use other devices to vacuum the oil, etc.

2010, Self-configuring Human-Computer Interfaces: When a computer or robotic system (like the self-driving wheelchair under development in our lab) is first introduced to its user, have the system probe the user's sensorimotor capabilities so that it can appropriately tailor its human-computer interface. For example, the configuration phase could: speak some text at a variety of loudnesses and ask questions to determine whether it has been understood; show some text in a variety of languages and font sizes to determine the user's first language, eyesight and reading ability; ask the user to perform motor tasks with known outcomes and gauge the results; etc. As mentioned above, we developed this idea for robotic wheelchairs, but it applies to essentially any person using any device.

2010, Nano-Nukes: Build a nuclear reactor at nanoscale with tiny fluidics to handle the nuclear and control materials, and a small turbine to convert steam into heat. Package the whole thing into the size of a standard battery, designed to produce power for (say) a hundred years.

2009, Collapsible Automotive Nose Cone: Design a nose cone that extrudes forward of a car at highway speeds to greatly reduce air resistance. Make it collapsible on impact so that it does not increase the injury risk to other parties during collisions.

2009, Protecting Athletes from Heat Death: At the start of each sporting event (e.g. marathon, triathlon, football game), have each participant swallow a small capsule with temperature and other sensing, and a radio to transmit its data. Station data receivers throughout the race course or along the sidelines of the playing field such that the players will frequently pass by some receiver. Have each capsule opportunistically offload its data trace when it can. Implement a filter-and-relay method to alert the competition organizers when the core temperature of any participant is about to exceed safe limits.

2009, Fine-grained Computational Economics: People have probably already thought of this, but in case they haven't: Instead of trying to come up with low-parameter models that predict economic activity, why not characterize individual people by an appropriate number of parameters -- salary, saving behavior, social context, etc. -- and run a simulation with many such individuals, chosen from a distribution matching that of the real world, to produce an emergent outcome? Think how revolutionary this could be if such a process were shown to "predict" historical economic information with high accuracy! [Update: A Mr. Frank Hirsch tells me that this has been tried, under the name "Agent-Based Computational Economics". Perhaps the difficult part is modeling each agent's repertoire of behavior and reactions to circumstances at any given moment.]

2009, Financial Transparency: Require all companies to maintain an RSS feed of every financial transaction they make. Use a naming scheme to assign a unique, persistent identifier to each financial instrument, along with metadata describing the parties, the date, time, valuation, etc. Third parties could then monitor, aggregate, archive, and analyze the feeds to look for discrepancies or bad actors.

2009, Anti-Bird Strike equipment pack for jet engines: Develop a microwave radar device as an after-market equipment package for jet engines. It would use radar to sense the presence (at speed) of birds in front of the engine. When birds were present, it would pulse them with radar, causing them enough discomfort that they would move out of the path of the engine.

2008, Anti-terrorist consumer devices: Secretly build GPS hardware into consumer-grade video cameras (I know, it's already being put into some cellphones and cameras). Secretly, say steganographically, encode the photographer's location and (true) recording time and date into any exported AVI, MPG file etc. Not as consumer-readable metadata, but as an NSA-readable bit overlay.

2007, Transportation via Ballistic Pod Launch from Electromagnetic Rails: Build infrastructure near each major city of EM rails firing ballistic pods with glider-like control systems, programmed to land on airstrips near each major city. After passengers disembark, pods could be sent back empty, or reused for other passengers. [Update: Elon Musk is doing something like this,]

2007, Truly no-stick cooking surface: Combine a griddle with an air-hockey table, so the food floats on air as it is being cooked.

2006, Garbage harvesting robot: There are several huge garbage pits on land, and garbage clusters floating in the ocean, that are full of valuable materials. Some of the land pits are picked over by human trash-pickers; the marine clusters are presumably investigated for food by animals. Idea: design robots that could burrow into these piles and identify and extract materials such as plastics, glass and metals. This probably isn't cost-effective today, but will be when resources get more scarce and energy (for production of new materials) gets even more expensive.

2006, Solar-powered condenser: Arrange for hot ambient air (e.g. in a drought region) to pass over a chilled surface, using solar power. Collect the resulting condensation drip in a cistern. A tiny unit could produce a few cups of clean water per day. The challenge of course is to do this very cheaply for developing regions. [Update: DARPA has issued a solicitation for research on such "water from air" devices.]

2006, Delayed development: Many children suffer brain damage from oxygen deprivation (drowning, freezing, commotio cordis etc.) at a young age, then age out of the period during which their brains have high plasticity. Perhaps this period of relative plasticity could be extended artificially in order to improve their chances of recovery.

2006, MEMS ship hulls: Paper a ship hull with a MEMS material (half-buried discs with rotation axis lying in the hull tangent plane?) to zero the effective viscosity between the hull and the surrounding water. Modulate the material's behavior to slow the ship rapidly.

2005, Group proximity sensors: For groups of children or tourists. Sensors form an ad hoc network, and measure distance to one another. Alarm sounds if any individual moves further than X meters from the group centroid or group leader.

2005, Voice-commandable wheelchair: Imbue a power wheelchair with sensors and computation to form a mental model of the world, and the ability to navigate through the world in order to reach the user's desired location. Useful for those with insufficient motor ability to move or joystick their own wheelchair. [Note: in collaboration with Prof. Nick Roy of the MIT Aero/Astro Department, and Dr. Bryan Reimer of the MIT AgeLab, I have begun to develop exactly this device.]

2005, Smart emergency exit doors: Put a sensor in the door frame to determine whether it is passable or not. Have a very loud speaker over the door directing people toward or away from it.

2005, Transcribed voice-based reminders: Enable your mobile phone to take voice dictation, the contents of which it would then transcribe and email to you. [Note: a company called "Jott" has since implemented this in 2008, with an 800 number and off-board speech transcription; the resulting text is emailed and SMS'ed to the user!]

2005, Automatic-reminder pill containers: Make pill containers/caps that remind you when your next pill should be taken. For example suppose you should take a pill three times a day. Then the cap could reset each time it is opened, and count down with an 8- or 6-hour timer, indicating when the timer has reached zero. At this time you would open it and take the pill, resetting the timer for another cycle.

2005, Passively powered bike helmet light: Power it from passive head motion; perhaps a weighted magnet bouncing around inside a coil?

2004, Queries as prediction: Data-mine the internet query stream in order to predict near-term future events. Query activity probably runs a few minutes ahead of stock trades, for example. Google and others are probably already doing this.

2004, Unsuccessful Internet searches: Consider the set of all failed Internet searches, i.e., searches that return no results. Suppose the search engine retained these for some months, then when similar queries were observed, offered to put the searchers in touch? Could be an interesting community-building mechanism. Another possibility would be to repeat the query at intervals, and notify the user when it produced results. [Note: Google has since implemented this idea as "Google Alerts".]

2004, Location-based reminders: Arrange for your phone/PDA to remind you of things based not (only) on the time, as with a calendar, but on your location. Examples: "next time I pass a hardware store, remind me to buy X," or "tell me when I pass a mailbox."

2004, MEMS shoe soles: Put ABS in your shoes. Have a sensor in your shoe that detects slipping, and within a few milliseconds, increases the friction coefficient of your shoe to stop the slip.

2004, Ice cube with constant surface area: Is it possible to design an ice cube shape that exhibits constant area as it melts? What shape should it be?

2003, Nano-cloth: Design clothing with dynamically adjustable porosity, or even vesicles, that trap or admit air to the desired degree. Thus the insulating properties of the clothing can be changed over short time scales as the user's environment changes (going from indoors to outdoors, or sun to shade, etc.) or as the user's body temperature changes (from exercise to rest).

2003, Smart File Cabinet: Develop a sheet-fed scanner that, as it ingests pages of a document, also listens to your speech about the document. Thus you can describe the document as you feed it in. The scanner also performs OCR on the document. Later you can ask the device about any document you've scanned, keyed by contents or verbal description. [Update: one of my students developed a prototype of this device as his Master's thesis.]

2003, Hair-dryer in your dashboard: Why wait for your car engine to warm up to have heat? Simply put a hair-dryer's electric heating element in the air ducting and power it from the car's electrical system. Once the engine heats up, switch off the heating element. Assuming you can limit the fire risk, this would make the car much more comfortable while adding only a few bucks to its cost. It might even reduce energy usage by cutting pre-drive idling time.


Page maintained by