Making Small Payments Work

Cashless payments are getting a lot of attention nowadays, but have you seen an auto not affiliated to Ola which accepts a cashless payment? I haven’t seen one! I believe that the notion of cashless payments is a bit forbidding for a significant fraction of the people, including many auto drivers. When I talk about auto drivers, it includes people manning small shops as well. I had filed a related post earlier. Visit

http://newstudentresearch.blogspot.in/2016/07/the-challenge-of-small-payments.html

A society that does not facilitate cashless transactions is a senseless society. The issue is not merely one of black money; it is also about avoiding waste of people’s time.

Why do so many of our people insist on continuing to use cash? I believe that that technical people involved should study the difficulties new users of technology face and work to eliminate these difficulties. Using technology should make transactions easier, not harder! The app on a mobile phone should require very little from the customer making a payment or the driver receiving it. You should be able to touch the app’s icon, point the phone’s camera at the auto’s meter or a similar display in a shop and touch the icon a second time. That should transfer money from your bank account to the recipient’s bank account or digital wallet. Who has a wallet, who uses a bank account, what is the name of the bank, etc., should be irrelevant at the moment the transaction occurs. Neither the customer nor the recipient of the payment should have to enter any numbers or addresses into a phone.

Character recognition is robust enough to let the app read off the amount to be paid as displayed; the app should similarly read the recipient’s Unified Payments Interface address from a sticker on the meter. (For information on UPI, visit Unified Payments Interface). 

While the front end described above has to be implemented, UPI already provides you the infrastructure to support it. I believe that student projects can easily achieve the goal of demonstrating such a user-friendly front-end for cashless transactions, integrated into the UPI network. A word for those who wish to carry out this project: print information to be read by the app using an Optical Character Recognition (OCR) font. This may not be necessary if you use a standardized font carefully chosen for the purpose, but it may simplify your task.  

Research related to financial inclusion

                              A socio-economic survey camp, Howrah, West Bengal

                           Report by Devinder Sharma, http://indiatogether.org/poverty-in-rural-india-poverty 
                                          Picture: Wikimedia Commons

There is a desire in India that the country should soon become a “cashless economy”. It is important that researchers in the fields of banking, commerce, management, ethnography, design, computer science should look into the problems that need to be addressed in taking banking to the poor and the illiterate.

The following questions are worth studying in this context:

What are the skills and knowledge required to operate a bank account, to use a debit card, and to use an ATM? Can any literate person manage these tasks? What level of literacy does one require? Can we do surveys to find out if people with low levels of education have (or do not have) useful access to banking? What are the practices for issuing cheque-books, debit cards etc.? Have any studies answering questions like these? Where have they been published? 

These are problems of major interest in India. “India currently has the largest population of illiterate adults in the world with 287 million”, said a 2014 report in “The Hindu” newspaperhttp://www.thehindu.com/news/national/indias-illiterate-population-largest-in-the-world-says-unesco-report/article5631797.ece

“A 2015 Unesco report said that in terms of absolute numbers, India - with 28.7 crore illiterates - was the country with the largest number of adults without basic literacy skills in 2010-11 compared to 2000-01 when it had 30.4 crore illiterates”. Visit http://timesofindia.indiatimes.com/india/Literacy-rate-up-but-so-is-illiteracy/articleshow/50749744.cms

A publication by the Indian Banks Association said “Normally no cheque book facility is provided to illiterate persons and blind persons. However, to meet periodic repayment of retail loans, utility bills etc. we will consider issuing of cheque book with safeguards to protect your interest”. Visit http://www.iba.org.in/bcsbi_code.asp and look for the paragraph with the heading “8.1.5 Special Accounts”.

Compare the change mentioned in the Unesco/Times-of-India reference cited above. Illiteracy declined by only 1.7% from 30.4% to 28.7% over the first decade of the 21st century. We hear the phrase “financial inclusion” fairly frequently. Unless the plight of the 28.7 Crore is well understood and documented, it would be difficult to extend the benefit of banks accounts, ATMs and cards to them. We should also find out if all those that are said to be literate do in fact benefit from banking. Are there several crores among them who find banking too intimidating to use in any significant manner?

Using ATMs to do currency tracking

                                   Photo: "ProjectManhattan", From Wikimedia, 

          Published under Creative Commons Attribution-Share Alike 4.0license.

This post is a sequel to the one at http://newstudentresearch.blogspot.in/2016/11/the-real-big-data-sources-and-laundries.html I will assume that the reader has read that post before reading this one.

Most ATMs do not keep track of who was issued which currency note. Here I will argue that this can be done with suitable software modifications. Let me make one question at the outset. Limits set for withdrawals and deposits using ATMs do not permit large transactions. So, why bother about ATM transactions? The answer is that not including ATMs in surveillance could create loopholes that would be quickly exploited. For instance, that a fair amount of “black money” is likely to be “white-washed” by using paid intermediaries to deposit smaller amounts (say, Rs 40,000 each) into their own bank accounts. These will be withdrawn later and given back to the holders of black money. So, ATM transactions should also be covered by currency tracking. How can this be done?

Currency notes are loaded into ATMs in multiple trays. New currency notes being delivered to banks can be in tray-sized stacks with contiguous serial numbers. Old notes to be loaded into ATM trays can be counted by modified currency counting machines (CCMs) so that modified ATM software can keep track of what high denomination notes were issued to which customer. Most computations and data storage can be done on centralized servers on a network that drives the ATMs. This will require that each ATM reports to a central server before a transaction how many high denomination notes are remaining in each tray loaded into it. They should also report how many high denomination notes are issued to which customer from which tray, every time a withdrawal takes place. The ATM need not handle the serial numbers at all, as they will be stored on the server. So, if an ATM has reported that its Tray No 3 has 630 notes left, and that it is now issuing 20 notes from that tray to the customer carrying out the transaction, the server can find out the 20 serial numbers of the notes issued, from its own files. Deposits through ATMs can be reported easily, as bank staff have to count notes in such deposits anyway; they can use a modified CCM.

Do we need to modify ATM hardware to do currency tracking? No. There seems to be no reason to do that.

So, we conclude that it is not very difficult track currency notes being handled by an ATM. Software changes necessary do not have to be done nationwide in a few weeks. The changes being made do not change the primary functions of an ATM; they only add a few additional functions.

The Real Big Data – sources and laundries of black money

The recent demonetization raises interesting questions that can trigger student research. What can we do with currency counting machines that read serial numbers of high denomination notes they handle and send them to a computer? Cash being deposited by customers can be counted and read by them keeping track of which notes are being deposited into which bank account. They can send a list of serial numbers to a computer, to link them to a given account. Payments of any significant amount of money over the counter can also be covered similarly. I will deal with serial number tracking of notes handled by ATMs in my next post; for the moment, assume that those notes can also be tracked.

Will this lead to draconian surveillance? What risks does such a technology pose?

Visualize a currency tracking system in which every bank branch reports serial numbers of notes it issues, or receives from, customers to a central authority identifying the customer involved. This could have many uses. For instance, suppose the police have found 10,000 Rs with a terrorist who was apprehended or killed. If an individual or organization had received those notes from a bank branch in the previous week, investigators would love to have that information.

A single note going from A to B may not mean much; A might have paid his milkman, who bought chappals from B with that money and B might have deposited the cash in his bank account. However, if ten notes had all gone from A to B within about a week there is some probability that A transferred them to B directly. What we are talking about is a system that will show currency flow from any given account A to another given account B over a given period. Flow through random middlemen would not explain a significant number of notes that were issued to A turning up in account B by sheer accident. In any case, a high flow of identified currency notes from A to B could flag it for the attention of Income Tax authorities, even if it is not dependable evidence of wrong doing. 

The research question is this: Can we identify a dozen economically significant applications that would justify the cost of this surveillance? We have thousands of students who are studying the use of big data techniques. They could be students of management, economics or engineering. Some of them could perhaps investigate this question.

A couple of concepts that may be relevant to research. We can define a bank account as a possible source of “black money” if a significant number of high denomination notes issued to it do not turn up in the banking system for a specified period. The presumption would be that A has paid it to someone who hoards cash. Similarly, we can define C as a “possible laundry” if high denomination notes not in circulation for a long time turn up in C’s bank account frequently!

To give an example, it would be worth finding out if employees in specific types of economic activities such as trading in gold and jewellery, or construction, are potential laundries. The presumption is that they could be paid in cash, using (unaccounted) money received in cash from customers. Similarly, if A is a possible source of black money it would be worth asking if he invests a lot in real estate or in jewellery.

Now, come to the feasibility of reading the serial numbers. Automated reading of numeric characters in a few known fonts has always been easy. This becomes easier when those numbers occur in specific locations on precisely cut sheets of paper and are in a unique colour contrasting with the background. In some series, the Reserve Bank of India has printed serial numbers using varying size font instead of uniform height numerals. That does not pose a major problem for automated reading of numbers.

There are some commercial products that handle serial numbers on currency notes: http://www.made-in-china.com/showroom/fionafocusbanker/product-detailIeaJGNTChXfy/China-Fb-810-Money-Counter-with-Serial-Number-Reading-Function.html

What about the serial numbers being unique or otherwise? It does not matter. We are looking at the probability of a currency note traveling from A to B. Rare coincidences of a large set of notes with the same numbers (as set mentioned above) traveling from some C to some D at about the same time is statistically improbable.

Another research question. Traditionally, currency notes are said to be fungible. The presence of machine readable serial numbers threatens this fungibility. Payments in cash are generally believed to ensure privacy. So, would currency tracking threaten this privacy? If I was a college student buying a bottle of whiskey, would I be worried that the authorities would inform my mom and dad about it? Surely, they have others things to do! I need not worry. Who should? Those who use the privacy of cash flow to evade taxes? Should we protect their privacy?

One last question: What would be the utility of a partial system in which deposits and payments over the counter are the only ones tracked, and that too only if the amounts exceed a certain value? Does the reporting of serial numbers of notes involved in such transactions add value? Or is it enough to identify the bank account holder and the amount involved?

My aim here is to share what I think are interesting questions. Not to provide answers!    

Good old tiles with embedded solar panels

I wish to trigger several student projects around the idea of using roof tiles to tap solar energy.  I think that students of architecture, design, or civil/electrical engineering would find this most interesting.

At the higher end, Tesla Motors in the US has showcased advanced batteries for home use to store solar energy as well as efficient and attractive solar tiles. Visit

https://techcrunch.com/2016/10/31/why-teslas-new-solar-roof-tiles-and-home-battery-are-such-a-big-deal/

However, affordability in countries like India makes it necessary for them to develop their own solutions. I urge students in such countries to try their own designs. The greatest value would be in areas where power failures/cuts are frequent, and of course in rural areas which are not electrified. As the cost of electrical power keeps going up at the consumer end, many rural homes may choose solar solutions to cut expenditure. 

Look at what is called a Mangalore Tile in India: 

http://www.btiles.com/roofing-tiles/mangalore-roof-tiles.html

It is usually made of clay and is 10 to 16 inches in size. I think that there is some logic in making solar tiles of the same size. Such solar tiles can be used along with the traditional Mangalore Tiles to build a roof and use only a fraction of the roof area to produce electricity. An alternative would be to make a big tile 20 x 32 inches in size, to replace four Mangalore Tiles. A large tile would be more susceptible to breaking during transportation and installation. So, some reinforcement should be considered to make them less fragile. Ideally, the tile should provide a connector on the lower side, that is inside the dwelling, to minimize wiring required. It would be great if the tile provides for supporting a lithium ion battery below it. This would enable keeping the battery inside the house. Care would have to be taken to insulate the battery from the heat that would emanate from the tile during daylight hours. An air gap between the tile and battery would help. It would be nice if the tile also provides mechanical support for suspending a small ceiling lamp with a reflector and LED. Then the wiring required would be minimal. 

How would you turn the lamp on or off? A hanging power cord with a switch at the lower end? Or a remote-control mechanism that responds to the TV set’s remote control when it is pointed at the tile? 

If you manage to get all this done, you can then consider if you can buy a DC driven small fan and incorporate that as well into the system. By then you might consider using two or more big tiles connected in parallel. The tile design should provide for serial or parallel connection of individual tiles.  

The National Renewable Energy Laboratory (NREL) of the US provides information about solar energy incident at different places in North West India.  http://mnre.gov.in/National%20Renewable%20Energy%20Laboratory/contents/solar_products.html

In general, most places in India seem to receive about 4 to 7 kWh/m² per day. That means that a big tile (as defined above) would generate only a fraction of this energy. Such a tile would have energy generating area somewhat less than 20x32 sq. inches, which is about 0.4 sq. meter.  A tiled roof would receive direct sunlight only for a few hours per day, and the energy it receives would depend upon the place it is in, the slope of the roof and the direction it is facing. Then there is the efficiency of energy conversion, perhaps about 10%. My guess is that with good engineering one could expect to get about 0.05 kWh per big tile, enough to power one or two 6W LED lamps for a few hours a day.

There are solar lighting solutions in the market that could provide parts for experiments with tile based solar energy system. Visit, for instance, 

_{http://www.amazon.in/Lighting-Belifal-Electric-Charger-charging/dp/B0132IAGDW?tag=googinhydr18418-21&tag=googinkenshoo-21&ascsubtag=77e5dfcc-31a2-4f56-bfb6-221423773093 or} 

_{http://ledsolar.in/product/solar-home-lighting-system-dc-12v-by-belifal-with-2-led-bulbs-electric-charger-battery-usb-port-for-mobile-charging-and-20w-solar-panel/}

_end

Personal Health Records on cell phones

I bought a new cell phone recently and was happy to find a facility it provided for creating a simple Electronic Health Record (EHR) for myself. EHRs can be quite complex as you can find by looking through the document http://www.nhp.gov.in/NHPfiles/Comments_invited_on_the_draft_of_Revised_EHR_Standards_for_India.pdf

What will work in the near future are simple forms of records that can be created on one’s cell phone by oneself or by a hospital. Typically, the EHR should provide for capturing information such as:

•      Name, date of birth, blood type, and emergency contacts
•      Name and phone number of doctor
•      Major illnesses and surgeries (giving dates)
•      A list of medicines being taken with dosages
•      Any allergies
•      Any diseases one has had over many years
•      Major illnesses suffered by one's parents, brothers or sisters
•      Habits such as smoking and alcohol consumption 

I urge students to experiment with simple apps to enable the creation and use of EHRs. Why an app? Why not a simple file created with a text editor?

I think an app can provide many advantages such as the following:

• The app will enable content to be created with prompting, and to be displayed when necessary. Users with sufficient knowledge may handle the app themselves. In other cases, someone else could handle the app on the user’s behalf. The app may protect the contents by going through a dialog that prevents careless handling by users who cannot cope with the complexity of handling medical information using a text editor.
• The app may allow a hospital staff member to access the data using Bluetooth connection and update it. Design issues that arise include safeguards to protect privacy.
• A well-designed app may allow the storage of doctor-generated information on the cell phone, replacing handwritten notes. This may include a section on medicines to be issued by a pharmacy giving dosages and directions for use.
• The app may allow a display of directions for taking the medicines and may also provide an alarm at the right time, once a day or more often as required.
• The app may give information in the patient’s own language whenever this is possible.

Different apps may provide different levels of capabilities, but should ideally be able to handle migration of essential information between apps at different levels as the user changes his cell phone or app. An interesting question that arises is that of securely storing doctor-generated information. Perhaps this can be digitally signed in apps running on smart phones with adequate resources.

The important point to note is that projects like this work best in environments in which the developer works closely with a doctor or hospital. Pure technology is not enough.  

end

A book and an article about remarkable innovators in the space race

You must read the blog post by Prof Vivek Wadhwa of CMU, at

http://wadhwa.com/2016/09/19/the-renegade-who-dreamed-of-going-into-space-and-started-a-race/  

It tells us about the Penguin book How to Make A Spaceship by Julian Guthrie.

It also talks about the great aircraft designer (visit About Burt Rutan and  https://en.wikipedia.org/wiki/Burt_Rutan ).  I had passed through the Seattle Airport a few days ago and had spent a few minutes admiring the mock up of the Voyager aircraft Rutan had built for flying around the world non-stop, without refueling, thirty years ago. (Visit https://en.wikipedia.org/wiki/Rutan_Voyager ) So, I was doubly thrilled to read Vivek Wadhwa’s article today.

There is a great quote from Burt Rutan I would like to bring to your attention as I conclude this short note:

“breakthroughs help define our species – without them, we get boredom and mediocrity and low expectations for the future” (Visit http://www.ethanzuckerman.com/blog/2006/10/14/burt-rutan-on-breakthrough-innovation/ ).

Auditory cues for directions and distance

Smart phones provide facilities that can be used in new ways to help the visually challenged. They can also be used to help vehicle operators like pilots and car drivers. They could be important in special environments like space and underwater. 

Cartoon: https://pixabay.com/en/boy-dance-music-headset-headphone-1271898/

Imagine a visually challenged student on the school playgrounds. A beep-beep sound at a set frequency can give him information about the direction and distance of the entrance of the main building from his current position. Assuming that he listens to this information through stereo earphones, the directional cue could be given by a suitable, short delay between the signals sent to the two ears. Further, the signal sent to the ear facing the school main gate could be louder by a computed amount. Important locations like the main building entrance could be marked by a cell phone acting as a GPS beacons for this purpose, broadcasting an ID number and a code indicating the nature of auditory cue that the listener should receive. The student’s smartphone would have a good compass that gives the software a reading of the direction the student is facing at a given instance. In future there could be a gyroscopic sensor to provide short term accuracy in direction sensing. 

It is possible to represent a direction, say north, by a steady whistle at a low volume. This could be in the form of a common complex waveform sent to the two ears, but one of them being delayed in a suitable manner, enabling the user to sense where north is in relation to the direction he is facing.

Multiple locations can be represented simultaneously by different types of beeps. Two dimensions are available to uniquely identify a signal – one being the number of beeps/second, and the other, the frequency of the underlying sub-carrier that carries the beep. Either one of these cues could give distance information. 

Some interesting research questions are:

1.   How many different locations can be signaled effectively?

2.   Can we offer training to increase the number of locations that a user can be aware of at a given time?

3.   What are the best ways of presenting distance/direction information to a user?

4.   Do we always need earphones? Or, can we use stereo-speakers in a car to give the driver some information about cars near him that he cannot see?

5.   Can the car speakers also give the driver information about other vehicles approaching or receding from his own? Can this, for instance, be done by change of the sub-carrier frequency carrying the beeps representing one of the nearby cars?

6.   What is a good auditory code to represent a section of a street? For instance, could a street be represented by a few virtual objects moving up and down the street, sending appropriate beeps?

end 

A Beacon in a Balloon to Broadcast Alerts after an Aircraft Mishap

An AN 32 aircraft of the Indian Air Force was reported lost over the Bay of Bengal on July 22, 2016. Visit https://en.wikipedia.org/wiki/2016_Indian_Air_Force_An-32_disappearance 

Incidentally, Wikipedia offers a very interesting article on the aircraft AN 32; it mentions that the “estimated price for a modernized AN-32 version is 15 million US dollars”.

This made me revisit a posting in this blog dated March 22, 2014. Visit http://newstudentresearch.blogspot.in/2014/03/the-need-to-invent-secondary-data.html

I have another possibility to explore in this posting. Is it possible to design and develop a device that, like an ejection seat, would be released after an aircraft mishap to carry information and broadcast it in the short wave band? The device, possibly mounted in an external pod, would monitor data such as the flight path, altitude, attitude and acceleration and store them in a secondary data recorder (SDS). Ideally, it would have its own sensors and operate independently of the aircraft’s instruments. The flight path would be recorded as a series of points in terms of GPS coordinates. The device, after ejection, would inflate a helium balloon and release it. The SDS would constitute one part of the payload of the balloon, along with a short wave transmitter. The alerts would be broadcast using a non-directional shortwave antenna sending out essential information giving GPS coordinates of where the device was ejected as well as the current position of the balloon. If, after a couple of days, the balloon comes down due to leakage of helium, its payload would continue sending wireless signals to enable its recovery. It would float if it lands on water.

Such a solution would present many challenges to a designer. What would be the difficulties in launching a helium balloon from a device ejected from a flying aircraft? What would be the security risks in using such a device with military aircraft? What should be done to make the SDS easily recoverable? Can the balloon drop it with a small parachute on wireless command? What would the transmitter power be, and what antennas would it require?

The proposed solution has several attractive features. The alert could be received at multiple airports within minutes of a mishap, enabling the quick launch of missions to recover any survivors. The system will work over land and sea. Search aircraft and ships would not have to spend weeks searching over tens of thousands of square kilometers. 

* * * 

The challenge of small payments

The huge population of India and the tradition of making small payments in cash only has created a problem. It is difficult to get 5 Rupee coins leave alone smaller ones. It is very common for shop keepers and auto rickshaw drivers to tell you they do not have change. I suspect that the Reserve Bank of India (RBI) finds it costly to mint coins and to print small denomination notes.

Innovation is called for. Imagine you can “beam up” amounts such as Rupees 57.50! All you should be doing is to enter 57.50 into an app, point your cell phone at a device in the shop and click a button. The amount should be transferred and a receipt in the name of the merchant, giving a transaction reference number, should be saved in your app. The app implementer can choose the mode of signaling, using the LED flash on the cell phone, or using Bluetooth or WiFi. Ideally the app should automatically turn on Bluetooth or WiFi even if the customer has not switched them on. The app can switch them off after the payment.

What about the transaction cost at the Bank’s end? Don’t worry, the banks know various ways of passing this cost on to the merchant! In any case, this cannot be very high; further any such cost will be more than compensated by faster handling of small payments at the merchant end.  

- - - - - -

Research Science Initiative - IIT Madras

The following post by Dr Sasikumar of CDAC would be of value to every young student interested in scientific research.

https://www.linkedin.com/pulse/rsic-amazing-initiative-iit-madras-m-sasikumar?trk=mp-reader-card

                                         - - - - -

Unfriendly human-machine interfaces

A few days ago, my wife got a text message from an app-based cab company. It told her she had not used their cabs for some time and offered her an incentive. It gave her some code like MYCAB and indicated that if she used it at the time of making her “next” booking she could get Rs 50 off from the bill. This incentive was offered for the next two rides.

I wondered what was wrong. The designer of this scheme must have gone to some extradinary design school! Anyone else would have implemented a mechanism to give her Rs 50 off for the next two rides without a code.

There is a possible reason. Their software provides for special codes to be recognized for giving discounts, but does not provide for marking a set of customers as qualifying for a specified discount for a specified number of times. If this true, we have to blame the software designers and not the company that uses the software.

I should, however, say that the design of the software did some good to the company. My wife did not remember the text message and the code it mentioned, when she booked a cab. Further she discovered that when she did another booking later on, the system did not accept the code. Remember, the message had said your “next” two rides! I don’t know what had happened.  I suspect that the system had invalidated the code on the grounds that she had booked a cab without using the code! So, the later booking did not qualify as the “next” booking! Surely a very smart and endearing practice to enforce using your software!

- - - - -

Articles and videos introducing computer science topics

You have some time free and would like to spend it on learning something interesting? Here are some recommendations. This list was originally put together for use by a few students who are going to engineering school in a couple of months. However, it would be useful to a variety of people, young and old, who are interested in learning some computer science, irrespective of the subject they studied, or are studying, for their degree.

1.       https://www.edx.org/course/introduction-computer-science-mitx-6-00-1x-7
You can audit any EDX course and pay no fees. Video lectures are good. You do not have any commitment to complete the course. You can take online exams and get your mark sheets

2.       Introduction to cellular technology. A very useful introduction. http://www.radio-electronics.com/info/cellulartelecomms/cellular_concepts/mobile-basics-concepts.php  A set of URLs for more information on specific topics is given at the end. They are useful to get additional information.

3.       http://vfu.bg/en/e-Learning/Computer-Networks--Introduction_Computer_Networking.pdf    

4.       http://www.ucl.ac.uk/archaeology/cisp/database/manual/node1.html

5.       Linear Algebra http://www.math.northwestern.edu/~len/LinAlg/

The books on Linear Algebra and on Discrete Maths do not have to be completed at a stretch. You can read one or more chapters at a time.

6.       http://www.cse.iitd.ernet.in/~bagchi/courses/discrete-book/fullbook.pdf

For most topics the Wikipedia chapter is usually the best introduction. Books, articles and other forms of text are not the only sources of scholarly information. Selected videos from Ted talks and YouTube are very useful too. I find the videos at  www.edx.org providing serious coverage of a topic most useful for learning from. Ted talks and YouTube videos are shorter and provide introduction, inspiration and overviews. 

In general, I trust sites that are in the .edu domain more than the ones in .com domain. But there are many useful articles on .com sites as well; for example, https://www.batterystuff.com/kb/articles/solar-articles/solar-info.html

You will find the following items interesting.

7.       http://inventionconvention.com/americasinventor/dec97issue/section2.html

8.       http://ed.ted.com/lessons/the-turing-test-can-a-computer-pass-for-a-human-alex-gendler

9.       http://ed.ted.com/lessons/the-turing-test-can-a-computer-pass-for-a-human-alex-gendler#review

10.   https://www.ted.com/talks/steve_jobs_how_to_live_before_you_die

11.   https://www.youtube.com/playlist?list=PLF7032F8EB1A4F9E2

                                                                        ----

A Computer Vision Challenge

Some people type into some food log a list of items they ate and get an estimate daily calorie intake. Some, find it impractical. My daughter says that she would like her activity tracker look at her plate of food at each meal and update its log! I thought it was crazy to ask for that when I heard it two years ago. I am wiser now, and think this can be achieved by a number of steps of innovation. As a first step, I announce this challenge, named Preethi’s Challenge in honor of the original thinker!

1.   The challenge is to recognize the items and quantities of food from their photographs. I visualize that you would use a tablet to take the picture and run the app on the same device. However, you are free to do it any other way, like using a webcam or a smart phone for picture taking and a laptop to run the app.

2.   If you have bought ready-to-eat food and its container has a bar code to identify it, the app is welcome to scan it and use the info.

3.   The app would use pictures with training information entered into “learning software” that will use machine learning techniques to recognize the food item and estimate its quantity and calorie content.

4.   You will train the app, giving it photographs of named objects annotating them with weight and information like calories/hundred-grams. Don’t worry about the training effort. Training one app would serve millions of users who are going to use copies of that app!

5.   You can use a standard plate of known diameter.

6.   You could also have only one item of food on the plate at a given time (at least at the start of the project).

7.   A good solution is to have the photographs taken by looking down, say at 45 degrees below the horizontal. This would give information about the item of food in X, Y and Z dimensions. The plate’s outline provides information on the scale. If the app finds it difficult to estimate calories from that photograph, it could ask you for a second photograph taken from another angle.

8.   The app should calculate the calorie content to some reasonable accuracy, say 10%.

9.   You can allow the app to ask for additional information when it finds it necessary, for instance, when it sees a plate with vegetables. It could ask, say, if it is boiled potatoes.

10.If you want to cheat, you could use a small plate that sits on a kitchen weighing device showing the total weight in digital form. Your app should recognize the numbers.  It could of course use stored information on the weight of the plate and deduct that.

11.You can make a second version of the app, which should take in more complex photographs, say with multiple food items, to make the app easier to use.

One last word: Use every trick you can think of to make this an easy-to-use app. Technology alone may not provide the best solution!

Why can't we tell consumers the truth about alcohol?

Reuters reported on May 3, 2016 http://www.reuters.com/article/us-johnson-johnson-talc-verdict-idUSKCN0XT20L a legal case in which Johnson and Johnson has been ordered to pay $55 million in compensatory and punitive damages for not warning customers on the risks of using of talcum powder. The company had been told to pay $72 million earlier, in a similar case. In the context of this, I look for clarity on labeling the health risks caused by alcoholic drinks.

The intention behind writing this post is to stimulate research towards quantifying risks consumers are exposed to, and discussions about adequate warnings to consumers. Is society lax in demanding warnings about the risks of alcohol consumption? If so, why?  Are consumers being fooled by statements that seem to imply that “moderate” alcohol consumption poses no cancer risk?

It is worth considering categories of risk posed by different substances, as listed by the International Agency for Research on Cancer (IARC), which is an intergovernmental agency forming part of the World Health Organization of the United Nations. The top few categories are:

Group 1: carcinogenic to humans: There is enough evidence to conclude that it can cause cancer in humans.

Group 2A: probably carcinogenic to humans: There is strong evidence that it can cause cancer in humans, but at present it is not conclusive.

Group 2B: possibly carcinogenic to humans: There is some evidence that it can cause cancer in humans but at present it is far from conclusive. 

Alcohol consumption and tobacco smoking have been labelled carcinogenic to humans (Group 1) by IARC in Press Release No 196 of 2009. Visit https://www.law.cornell.edu/uscode/text/15/1333 to see the very tough law that requires every cigarette packet to say that cigarettes are addictive, cause fatal lung disease, cancer, strokes and heart disease and that smoking can kill you. I have not given a full list, but you can read them in the reference given. 

Compare this with the warning specified by the Alcoholic Beverages Act, USA:

(1) According to the Surgeon General, women should not drink alcoholic beverages during pregnancy because of the risk of birth defects.

(2) Consumption of alcoholic beverages impairs your ability to drive a car or operate machinery, and may cause health problems.

Also, compare this with the warning on my bottle of an Indian beer bottle: 

  ALCOHOL CONTENTS LESS THAN 5%V/V

  DRINKING IS INJURIOUS TO HEALTH

The World Health Organization says about cancers that "Around one third of cancer deaths are due to the 5 leading behavioural and dietary risks: high body mass index, low fruit and vegetable intake, lack of physical activity, tobacco use, alcohol use".

WHO reports that tobacco kills around 6 million people each year. More than 5 million of those deaths are the result of direct tobacco use while more than 600 000 are the result of non-smokers being exposed to second-hand smoke.

Now consider the connection between alcohol consumption and cancer risk. Visit Alcohol and Cancer Risk put out by the (US) National Cancer Institute. It lists many types of cancer that have been shown to be linked with alcohol consumption. 

WHO says about alcohol that “in 2012, about 3.3 million net deaths, or 5.9% of all global deaths, were attributable to alcohol consumption”.

Examine all this in the light of what is reported about risks from talcum powder usage. The Wikipedia article https://en.wikipedia.org/wiki/Talc says that “Suspicions have been raised that its use contributes to certain types of disease, mainly cancers of the ovaries and lungs. It is classified in the same 2B category in the IARC listing as mobile phones and coffee”. It also says that “The US Food and Drug Administration (FDA) considers talc (magnesium silicate) to be generally recognized as safe (GRAS) for use as an anti-caking agent in table salt in concentrations smaller than 2%.”

I am not saying that using talcum powder is harmless. My case is that considering the far higher risk posed by alcohol, the warnings provided to customers are inadequate. The data given above shows that cancer deaths due to tobacco smoking worldwide were approximately 1.6 million. I believe that adequate attention has been given to this and effective warnings have been made legally necessary. On the other hand, the data given above shows that “3.3 million net deaths” worldwide in 2012 were attributable to alcohol consumption. To merely warn me that “drinking is injurious to health” is gross injustice.

There is another form injustice in labeling alcoholic beverages. Visit Alcohol calorie calculator This NIH publication says that “Alcohol beverages supply calories but few nutrients and may contribute to unwanted weight gain. If you need to lose weight, looking at your drinking may be a good place to start”. If you buy lemon drop candy for a child, the nutrition information tells you that it provides 18 calories per piece, but my bottle containing a pint of beer that provides 180 calories is not required to carry any label about that.

end

City Location Codes (CLCs)

Mammals, particularly carnivores like tigers, have evolved rather simple ways of marking trees to show that they form part of the boundary of their territory. Forget any imagery you associate with that statement! We will go on to discuss more civilized forms of marking up places in a city or town for vehicles ordered through apps to meet you!

The problem I discuss here may be relatively more significant to people like me living in a fast-growing, crowded city using multiple languages.  GPS, I find is not enough to guide a cab driver trying to find me in a crowded area. I am one of 300 people he finds in a place where his GPS says I am. If I make the mistake standing at an intersection of two roads, he does not know which one of the four corners I am in. On a straight road, it is difficult to convey on which side of the road I am waiting. I don’t know from which direction he is coming towards me. He calls me on the phone sometimes, but we do not always find a common language. Even if that is no problem, I am often unable to describe where I stand. Street names are kept a state secret here! In any case, there are so many streets, and so many streets with popular names different from official names, nobody seems to use street names anymore. Landmarks are fine, but they are usually not where I stand!

I would like to suggest a solution, not entirely original. Cities like Singapore have numbers painted at prominent places where cabs pick up passengers. I will suggest that we use a code of the form BL 959,574. If you are a resident of Bangalore and you are using an app, “BL” could be pre-filled. If you insist, we can add a country code and say INBL instead of BL, to make it clear we refer to Bangalore, India. Suppose you are at a cab pickup place displaying a prominent dark green colored sign board, with BL 959,574 painted on it in white. All you need to do is to enter BL 959,574 into the app to say where you are. The cab driver can locate you very easily. Incidentally, in this code BL 959,574 stands for Tipu Sultan’s Palace in Bangalore.

The proposal is that a two-letter code should identify a city; I would suspect about 300 cities/towns could find easily a two-letter code each that identifies them uniquely. For smaller towns, I would suggest three letter codes.

How do you assign a numeric code? Take the latitude and longitude of a given location, each written as a decimal number rounded off to three significant places after the decimal point. Leave out the integer part. For instance, Tipu Sultan’s palace is at a place close to the point (12.959 N, 77.574 E). For our purpose, we simplify this as (959, 574) and write it as BL 959,574. Small errors do not matter as long the sign saying BL 959,574 is close enough to the place indicated and is very visible to people looking for it. If your city fits in a 100 KM x 100 KM square, this system is enough to give you unique codes for locations in the city. Any shopkeeper could mention such a code in advertisements to say where the shop is. You can give a code to your visitor to say where your home is. However, I hope that the pickup points with display signs would be regulated so that it is at a place suitable for at least 30 people to wait, protected from rain/sun, with space on the street for five vehicles to stand at a time.

Search engines such as Google could easily incorporate these codes into their city maps. Users can specify “from” and “to” addresses succinctly and accurately to apps used for ordering cabs. It is worth stating that the proposal primarily aims to make GPS location information easy to use by humans. Google API offers facilities for finding Latitude/Longitude information of a given place. It also supports mapping a point identified by such geocoding information onto a map. Students should experiment with such facilities to implement user-friendly interfaces using CLCs.

There is an interesting idea relevant to those who design apps to use CLCs. Consider CH 202,271. It is easy to find that this is the CLC of Edayanchavadi, Chennai, if you know its GPS coordinates: 13.202 N, 77.271 E. But there are places south of Latitude 13.000 N in Chennai. So, how does an app determine that the given CLC corresponds to Longitude 13.202 N and not to 12.202N? This requires that the app have access to information in some form indicating CH CLCs are to be mapped to the Latitude range 12.845:13.262 and the Longitude range 80.080:80.329. (The latitude and longitude ranges given are only indicative. App implementers should double check these ranges).

A question that may arise in the reader’s mind is this: Do you expect the common man to use a location indicator like 959,574? Yes, after all the common man does use ten-digit mobile phone numbers! The simplest way to get such location indicators is through search engines, audio search engines, contact information stored on phones, etc. Don’t under-estimate the common man! He learns quite a bit, when necessary.

I have another suggestion to simplify urban life. App based transport companies such as Uber and Ola should sell caps of distinct colors for their users to wear. Then drivers would know who is likely to be the customer that called them.  Their vehicles could sport a flag with the chosen color for the users to identify their cab.

End