The seeds of the information age were sown in the middle of the 15th century when Johannes Gutenberg laid the foundation for the printing press, which helped humans pen down their thoughts, copy and duplicate their work for others, thereby unleashing the intellectual tap of the world like never before. The written word retains its place today. Yet, for a long time now, ever since the advent of the mass television culture, humanity has been recording a lot of video. The internet was the platform, while any person in the world could now shoot anything and upload them for the world’s internet population to see, provided they looked for it.
More or less, the foundations of our modern information society lies in video. Videos are everywhere. They’ve allowed the world’s storytellers to reach audiences all over the world, and literally echo the term ‘global village’. It is likely to remain the dominant form of conversation and information for the foreseeable future, with minor changes here and there in its formats.
Yet, the question arises, if we are able to literally ‘put an eye on the world and record everything’, so to speak, can we mine it to improve the way humanity has functioned so far? Amit and Aaron are the two founders of the Collective Intel platform. Today, at Bangalore Insider, we sat down with Amit Dhand to understand how video analytics and intelligence will likely unleash a new era of insights and improvements for businesses all over the globe.
Tell us a bit about yourself. What were your previous jobs/ventures before starting this.
Amit Dhand has an extensive track record of developing new technologies and launching companies that have disrupted the $40 Billion+ global data and analytics marketplace.
Prior to Collective Intel, Amit launched NAILBITER, Inc. in 2014. NAILBITER is a Videometrics technology platform where grocery shoppers record their in-store and online purchase decisions on video – globally. NAILBITER converts that video into data which then provides insights to leading FMCG manufacturers such as RB, Nestle, Mars, Campbell’s and many more. NAILBITER is one of the fastest growing data companies and has received many awards and accolades in the industry.
Prior to Collective Intel, Amit was a key executive at Affinnova for 8 years. Affinnova was an evolutionary Algorithms driven optimization technology platform. It was a small start up which Amit helped bring into the Data and Analytics space resulting in a 16x increase in annual revenues and the company was successfully sold to Nielsen in 2014.
Please tell us a bit about reopeningwithcovid19 and how did you come up with the idea.
The Collective Intel platform has the ability to “see” and decode almost anything inside a video and convert it into data and information. The main application for this technology is to drive business insights in the Fast-food and Retail industries. However, when the Covid-19 pandemic hit, CI decided to launch this public initiative to help various cities re-open safely.
The idea is simple – to provide users with the knowledge on how many of their fellow citizens are practicing safety precautions. It is also meant to drive awareness that to re-open society must come together and support safety measures as a community and not think individually
Finally the site also serves as an example to governments and businesses on how they can be transparent about their re-opening efforts. Unlike AI driven systems, the CI system is driven by humans so that the community can monitor itself without “big brother” watching over them.
The way CI works is by combining the best of AI and HI.. HI = Human Intelligence. Collective Intel has a global crowd of online users who watch videos and decode what’s inside them. This is done via a smart platform that determines how to take a 24×7 live feed, chop it up into shorter “clips” and serve them to the crowd to be coded. Think of it as the Uber of video – there are people who want videos analyzed and there are people who are willing to analyze videos from their home.
What is the minimum standard of video infrastructure required (‘x’ number of video feeds with ‘y’ type of specifications or maybe a standard level of feed video quality) in order for Collective Intel’s algorithms to produce their best output on a 24/7 basis?
Accuracy is Collective Intel’s #1 benefit, but a close #2 is the lack of new infrastructure or any minimal system requirements! Customers can hook up 1 camera or 1,000,000 camera’s to the system. They can code 1 hour of video per week or 24×7. The system has no minimum commitments or upfront investments. The hallmark of good technologies is that you can use them quickly and cheaply, then determine if they add enough value to the business before making them a part of the daily business operation.
The CI platform is also designed to work with most cameras and camera systems that are in-use today
Is it right to say that video intelligence will help increase real estate space utilization in tandem with social distancing norms (one of the base parameters of your AI algo)? Please elaborate.
Yes that is correct. Covid-19 will permanently change the way we use all indoor spaces, whether they are retailers, restaurants, offices, schools or stadiums. Obviously the density of the space needs to be reduced, however this does not mean that new buildings need to be built.
Even before Covid, most buildings were empty most of the time! If a week has 168 hours, a typical stadium is empty 90% of the time, a typical school/ college is empty 60% of the time, a typical mall or restaurant is empty 50% of the time. Even if we make small increments in the duration of the time a building is used, we can greatly reduce the density of the space and make it much safer.
However such innovative approaches will require heavy use of new technologies. Just like Google traffic tells you which roads to avoid for now, CI and other video technologies can show users which malls or restaurants to avoid at what times. Such technologies can help authorities plan school and office hours as well as uncover where, when and why people do or don’t practice safety norms.
In your research, what percentage of businesses in specific business categories already have established video infrastructure systems? Would you be handholding companies to persuade them to invest in video, if they already don’t?
Every retail business and most offices have video already installed. The CI platform simply taps into existing feeds and needs no new equipment. Most of the current systems have very basic analytic capabilities and even the high-end AI drive video analytics systems have low accuracy rates and high complexity. As a result a vast majority of such platforms are unused or under used.
Driverless cars are the future of humanity – everyone knows that. However these are infinitely complex systems that have been “around the corner” for 20 years. They are still 10 – 20 years away from mass deployment. Meanwhile Uber was able to take existing infrastructure (cars and cabs) and new technologies to create a global revolution where people could rely on rides and did not need to purchase their own car. Not quite driverless but powerful all the same. CI is doing the same thing with existing video platforms!
How easy or difficult is it to marry all your intel with an organization that has access to facial and biometric data, in order to track and surveil individuals (in the worst-case scenario without their consent)? Can employers ask you to put in a system to track your employees under the garb of covid-induced necessity?
Collective Intel is about the “Collective” and not the individual. The platform is not used for facial recognition – in-fact personally identifiable features such as faces or car license plates are automatically obscured in the video right upon arrival. The system cannot be used to identify a person or car. For example we can tell you that your employees are least likely to follow social distancing and mask wearing guidelines during the 3-5pm slot. We can also tell the employer that this appears to be linked to coffee and smoke breaks, as well as general fatigue towards the end of the work day. However we cannot tell the employer which employee is not following the norms
How is Collective Intel’s Public Safety Compliance Score calculated? What are the components used in its calculation methodology?
We have many measures that we can track but the 2 main ones are;
% of people wearing masks and
% of people socially distanced
These measures are simple to understand and can be compared across time periods, businesses and geographies. The aim is not to get 100% compliance, but to have enough compliance that the overall system is safe and getting safer by the day.
These 2 measures are routinely combined with other ones such as time, location, presence of other factors (such as food, drinks) to further diagnose why people are or are not practicing safety precautions and how to drive and reinforce good behavior. For example we can measure the impact of putting up safety signage and determine the short term and longer term impact of compliance with safety.
Which are the metrics you use and present clients with to base your monetization strategy on? While they vary from one business to another, please tell us the approach you take.
We offer video analytics that
There are countless revenue driving opportunities in our system – too many to count but a few are listed below
Increased Drive Thru revenue for Fast Food places. Most fast food places (especially in the US post Covid) make their money via the drive thru. Our system can tell when and why cars drive away before placing their order (perhaps the line is too long). By providing clear video data on this phenomenon we have helped our customers reduce these Drive Aways leading to dozens of extra orders per store per day. If we can increase revenue per store by $50 to $200 per day, that day is worth $10 to $50 per store per day. Across a chain we can drive millions of dollars of increased revenue per week!
This same example can be used for line management in most indoor restaurants, malls and other locations
We can also drive revenue by measuring the impact of signage placed near a store. Sometimes small changes to the message or the location of the sign can have a huge impact on the Noticeability of the sign and its ability to drive traffic to the store.
Video data is a great way to see where money is being wasted that could be saved. Every business owner has a hundred hypotheticals on how their systems and people are being inefficient but no way of truly measuring these theories. Video analytics becomes an easy way to prove or disprove these hypotheses.
For example: CI can help quantity the % of times the delivery trucks keep the engine running while loading and unloading, quantifying the loss of fuel in the process.
In addition to the examples given in other parts of the interview, we believe that businesses should showcase their safety scores as a competitive advantage. If a hotel or restaurant displays a 89% performance on safety, and can establish that the average for competitors is much lower, they can win new customers and drive loyalty and satisfaction among existing customers.
When you say you ‘accuracy rates’ approach 99%, what is it that you’re referring to if not the algorithm correctly recognizing the tasks it has been assigned to, one of them being people not wearing masks? Can you tell us what else the algorithm has been trained to do?
The 99% accuracy is not possible on any AI platform. If AI was this good, we would have seen self driving cars everywhere. This is despite the billions that have been spent training AI on self driving. It takes a huge investment to train AI on the most basic questions. Businesses don’t have the $ or the patience to train AI and their data needs change faster than AI can be trained! The CI system uses a crowd of humans to code videos. Humans come pre-trained and are significantly more effective, efficient and scalable.
This accuracy rate applies to all sorts of measures, not just mask wearing. For most crowd users, these questions are common sense and require no specific training! The system does provide some guidance and instructions. For example we can show them what 6 feet looks like on the video so that the user does not need to use their imagination to guess if people are social distancing.
What is your plan for India in the upcoming future?
The CI platform is available for use globally. We are having many conversations with retailers as well as governments on applications for the indian market. For smaller businesses and municipalities that may not have the resources to pay, we are offering our Covid related data at cost or sometimes free. They can reach us by clicking “ContactUs”.
Any business can get a free single video trial at by clicking “SignUp”.
Know a great startup story or want to share your own? Write to us at firstname.lastname@example.org and we will get back to you. For more updates follow Bangalore Insider on Facebook and Instagram.