Recently I gave a keynote speech in Mahindra University, Hyderabad as part of a 2-day workshop on “Data Science for the Industry”. Great opportunity to share my thoughts on Data Sciences/AIML technologies and industry use cases.
I talked about various problems to be solved by these rapidly advancing technologies. One of them was “Are You Human?” question. The problem is created by AIML technology and also solutions need to come from AIML technology. Basically, how does any IT system distinguish between humans and machines during transactional interfaces?
Is this problem important enough to worry about? Yes. I will give you both technical and commercial reasons for it.
First Commercial reason.
I am sure all of you heard of the Twitter take-over bid by Elon Musk, CEO Tesla for US$44 Billion. The deal was cancelled by Elon Musk due to inability to determine % of non-human or BOT users on Twitter. Elon Musk accused Twitter of using incorrect algorithms to determine BOT users and under estimating the real BOT numbers. The issue is now in legal dispute.
Many commercial decisions are based on number of customers. For e.g., number of people visiting the website determines cost of advertisements and royalty payments to the website content authors.
Second Technology reason.
The age of Digital has transformed IT landscape across the enterprises and use of web, mobile phones, chatbots and IOT devices are the norm and not exceptions. All of these channels are communicating with the enterprise IT systems and getting business executed i.e. placing orders for products, registering service issues etc. At the same time automation is also become a norm and Robotic Process automation tools are widely used in enterprises. Many cases they use various technologies to simplify data entry by using a single screen input and on the background simulating multiple screens data entry to various enterprise systems. These interactions are legitimate and should be flagged as non-human but legitimate approved interfaces.
I am sure now you are convinced about the importance of the problem.
Now let us come to main topic of our blog i.e., CAPTCHA.
All of us have used on-line or mobile Banking to do banking transactions. Most of us would have encountered some thing called CAPTCHA. The system throws a set of characters twisted in a wavy curvy fashion and system expects the interacting person to see the image and do the right interpretation and enter it back to the system for confirmation. Some examples are given below.
The system generates random sequence of case sensitive alpha-numeric characters such as 263S2V. This is twisted in to an image as you see above and shown back to the interacting agent. It is assumed that automated systems will fail to interpret this correctly and only human can interpret and type back the same set of characters 263S2V.
What is full form of Captcha? “Completely Automated Public Turing Test to tell Computers and Humans Apart”.
When was this invented? Between 1997 to 2003. The most common type of CAPTCHA (displayed as Version 1.0) was first invented in 1997 by two groups working in parallel. In 2000, CMU professors Luis Von Ahn, Manuel Blum and John Langford wrote a paper titled “Telling Humans and Computers Apart (Automatically) or How Lazy Cryptographers do AI”. The term CATCHA was coined in 2003 by Luis von Ahn, Manuel Blum, Nicholas J. Hopper and John Langford.
This form of CAPTCHA requires someone to correctly evaluate and enter a sequence of letters or numbers perceptible in a distorted image displayed on their screen. Because the test is administered by a computer, in contrast to the standard Turing test that is administered by a human, a CAPTCHA is sometimes described as a revised Turing test.
I am sure you are wondering how come a 20-year-old technology is still in use for Hitech Banking industry as a digital security solution?
This technology is very cumbersome and frustrating for all humans. Younger people with sharp 20-20 eye vision may be able to get it right first time but it still adds 10-15 seconds to the transaction. Senior citizens, people with glasses, people with vison disability, people with low quality display units, poor lighting in capital or not with all the wavy and squiggly images. In addition, the systems are badly designed and it forces me to re-enter all the data fields till I get my Captcha right.
In the last 20 years, the AIML technology has improved exponentially. Hand writing recognition and image recognition technologies are very good and they can easily recognize the Captcha transformed images. I can go one step further and say that if a senior citizen customer gets the CAPTCHA first time right, then the bank should assume it a fraud!. Unfortunately, the banks assume the exact opposite, which was the original basis of the CAPTCHA technology.
Various other ideas such as speed of data filling were also considered as part of CAPTCHA. Humans do take time to type the data while automatic BOTS can do it at super-fast speeds. However, RPA based automation systems will always be fast and they are genuine systems interactions. Also, it is so easy for a BOT to slow things down by waiting few seconds before submitting the data and fool the timing algorithm.
We have seen big discussions on evolution race between the prey and the hunter in the biological world. The deer evolves to get stronger legs to out run the tiger. Tiger evolved stronger lungs to sustain long chases. Same way as the AIML technologies evolving so fast to mimic human interactions, we need to get better technologies to solve the “Are You Human?” problem.
More later. Do share your views.
Regards,
L Ravichandran
AiThoughts.Org
No Comments yet!