Bringing in AI and Analytics to Bangladesh
An Up-Close Interview of Mohammad Oli Ahad
Mohammad Oli Ahad, Founder, and CEO of Intelligent Machines Limited, aspires to build practical AI-powered solutions for our country with his strong local team who has particular expertise in the areas of AI and advanced analytics. After completing his BBA from the Institute of Business Administration at the University of Dhaka in 2006, he joined British American Tobacco as a Territory Officer in trade marketing. 3 years later, he joined BAT’s regional IT team as a Business Analyst - Asia Pacific, Marketing. He led a number of global and regional projects for 25 countries including Australia, Singapore, Hong Kong, and South Korea—some of the most advanced technological environments. While working to implement state-of-the-art technologies in those markets, he grew a deep urge to build technology solutions from scratch for Bangladesh where we largely adopt tools made for developed countries and then try to force-fit them for the local market. In April 2018, he started Intelligent Machines. Their first product, Retail AI, went live nationally in October 2018—and, the MFS market-leader bKash Limited is using it ever since for merchandising activities in all their 300K active outlets in the country at present. The team is currently working with several other large organizations to build AI products in retail, FMCG, banking, and non-banking finance industries.
IoT for BD: Artificial Intelligence (AI) is a topic that's pursued by a large number of students and researchers today in Bangladesh. Do you think the AI industry of Bangladesh has the capacity of growing large enough to foster the interests of these people inside the country?
Mohammad Oli Ahad: The opportunity is so big we do not have a way to assess it in any proper sense. Let us take Gmail as an example. Today, we are thrilled at how accurately it can auto-complete what we are going to write, as if it knows us just as well as—or, better than—our best friends. But, Gmail does not know what is in our mind—it has become remarkably adept at figuring out the pattern we exhibit in our writing and this strikingly shows how consistently we behave over a short period in delivering our communications.
It would be even more interesting if Google would start to share how we would write in ten years from now—or, fifteen years from now. And, how we would write differently to different contacts based on our changing dispositions towards them. The next knowledge graph would show us how we used to write five years earlier and how our writing styles and relationships have evolved—and would likely evolve in the future; imagine how thrilled we would be!
Thrilled—and, probably, horrified. From your writing styles, Google might be able to tell you that your friendship would not last or, at least, not stay this close, say, in twenty years while you are still in a very happy-close relationship.
It is unimaginable how powerful it would be if Google was able to introduce and offer matching partners, co-founders, friends, and team members from disparate sections of society and different geographic regions based on individuals’ writing patterns! You see, we have just barely started to scratch the surface when it comes to applying the powerful force of artificial intelligence in our daily lives.
If you look at the history of Artificial Intelligence you will see - as dull and as real as you find it to be – that all these developments have actually been driven by commercial interests, and not by academic interests, in sharp contrast to the glamorous excitement that we see almost daily in the media.
The funding has essentially always steered the developments in this resource-hungry field; and, funding has flown primarily from a near-term capitalization point of view instead of any knowledge or academic desire. That is why we are now left with several massively powerful applications that could allow large organizations to make tonnes of profits but, the field has mostly remained in the 80s, really, in terms of theoretical progress. Much of what you see—Amazon/YouTube/Netflix’s recommendations, Google/Apple’s face recognition and digital assistants, all of them—are pinned on back-propagation which was invented in the mid-80s and is deemed by the experts as a ‘placeholder’ until something better comes up—something lot less wasteful in ‘learning’ with training data and lot more closer to how the human brain works.
These realities make the situation doubly open for capable talents in Bangladesh, and all over the world just equally. Any serious student and researcher can potentially make original dents in the field’s core theory—as the field has not gotten to take its fundamental shape yet in much of its supposed aspects. And/or come up with numerous powerful applications of AI which would be just as common and ubiquitous as microwave ovens and radios became for electricity—as we have not yet truly started to imagine what the field could be even in the next couple of decades.
So, the Satyendra Nath Bose and Thomas Alva Edison of AI can very well come from the University of Dhaka or Chittagong just as they could come from the USA or China.
Also, the market here is quite poised even without the corpus of structured data - like the ones available in so many industries in the developed world. Many of the managing directors and CEOs of large organizations have a healthy level of awareness of AI and encouraging curiosity to know how they might harness the power of this new force to advance their organizational goals and objectives. They have been coming across powerful stories about this field every so often in social media and mainstream press. If a capable team does approach the market, I would say they would get invitations to carry out POCs with fairly reasonable efforts.
What I worry is that the talents are not staying—or, staying long. That is sad and that is the reality. The great talents we have here are just as capable as those in the Silicon Valley and Zhongguancun—but, the teams in those places get continued services from a large talent pool; we still do not.
We have zero voluntary in-country attrition; but, the wonderfully talented people who join us (including the people working behind startups) mostly go abroad after a few years: They go to join a larger organization, many totally migrate to another country, and a number of them go to pursue Ph.D. programs after which they are less likely to return anytime soon. And, we stay stuck with a persistent hole in the talent pipeline. If you do not have the builders, what would you build with? The job becomes very much of that of a recruiter just as it is that of a technology builder.
So, yes, the “large number of students and researchers” who are pursuing AI in Bangladesh, please, by all means, pursue it even harder! You have everything right in it for you. It is a seller’s market, at least very much so, for you. If you were really good, we would all be after you; and, you can cherry-pick whichever team you want to join—or, create your own one!
IoT for BD: What suggestions would you give to the country’s young people who are hyped to work with AI?
Mohammad Oli Ahad: Join for the right reasons. This field is largely undiscovered and you will have to pave it. And, while it will be thoroughly rewarding, it will be tough.
For almost every original challenge, you would have to read tonnes and tonnes of thesis papers and search everywhere to try to learn if anyone anywhere has thought about it in any scope. And, even when you are lucky enough to find some directions in some research papers when you attempt to run the theory with production data you might see that it does not work. For thesis papers, researchers usually work with a small set of data that ‘work’ in lab settings. But, the same theoretical framework would often break apart when you apply your significantly large real-life data set. You would actually do the original researchers a big service by contributing to their work and possibly highlighting a few understated nuances which would give them many helpful pointers to take their research further if they will. But you might end up returning to square one with regards to finding out a suitable approach to tackle your problem.
I would request the aspirants of this field to consider the following duly. In software development, we are used to applying various code-libraries and plug-ins. Well, you will find code-libraries in AI too, but those would only take you so far! Firstly, they would not completely serve your purpose. Secondly, you would have to come up with the solution design approach in the first place which could take months of research-study and trial-and-error.
Google, Amazon, IBM, Microsoft offers APIs for AI but you would not prefer to use them as they would not give your desired solution completely and still be quite expensive. We have had AWS India team visiting us three times, and every time they have agreed that they are still not offering what we are developing here.
So, please ask yourself why you want to join this field? And, how much do you really want it? Would you be able to pour over mundane hours of paper studies? And, take the pain of hand labeling tonnes of data? And then run trials after trials of models for hours and days only to discover that they do not work—and start all over again with equal enthusiasm and zeal?
If this describes you then you are our Jeff Dean and Andrew Ng. and we want you!
IoT for BD: Would you like to share the details of the first product of your company, Retail AI, with us and how it works?
Mohammad Oli Ahad: We are aspiring to upgrade our country’s retail ecosystem. For the benefit of everyone in it, including the thousands of hardworking street retailers who deserve so much more for their critical service to society.
If you look at our physical retail space, you will see that it still pretty much operates like the middle ages. There is no record keeping, no trade financing, and no credit support for these tiny street shops.
They lack much of the critical skills like inventory and category management and financial management. Many of these street shops also essentially run an unbelievable ‘free credit card’ service to their local customers by offering their goods in credit without any extra charge whatsoever. And eventually, many of them get strangled to their own poor judgments.
We want to introduce a platform that will connect these street retailers with manufacturers and offer AI-powered critical insights to both retailers and the large businesses that operate through them. Retail AI will also enable trade financing and credit arrangements using the retailer’s data as collateral.
IoT for BD: What other upcoming endeavours are you looking forward to with your company?
Mohammad Oli Ahad: Currently, we are working with six large organizations to build B2B AI products in the banking, non-banking finance, and FMCG industries. We are also preparing to expand our operations to the Netherlands on the invitation of a Dutch-government supported program. We would offer AI and analytic products as services by Bangladeshi talents to the European market.
IoT for BD: AI and IoT are two buzzwords in the tech field. While our readers might have some basic idea about how the two can be integrated, we would love to hear it from an expert.
Mohammad Oli Ahad: These two powerful technologies complement and strengthen each other. IoT devices collect and transmit data, while AI processes the data and facilitates smart decisions. And, in some cases, AI can also serve as a proxy for the other.
A company in the Netherlands implemented a system where small drones fly over farming fields, take photos, analyze the texture of the soil, and determine which areas of the fields are in need of water. The data transmission is done by IoT devices that send data throughout the system. The data analysis is done via AI systems, and the results are again transmitted via IoT devices to the “actuators” of the system. This minimizes the amount of water required for irrigation, and in turn, decreases the cost to 70% of what would otherwise be needed.
Another team in Taiwan have introduced microchips in grape fields that collect and analyze data from the roots of the plants and its surrounding soils; and, then, point out effective ways of increasing the yield of grapes. I believe Bangladesh could benefit significantly from similar systems, and small projects like these have huge prospects here in the country.
Now, let me give you an example where an AI can work as a proxy connector in absence of IoT.
You are sipping coffee from a cup and putting it down on a saucer after taking sips from it. And, every time you are taking coffee, the volume of liquid in the cup lessens and the sound it makes proportionately varies when you’re putting an increasingly empty cups on the saucer. Now, if this saucer was equipped with intelligence to discern the amount of coffee in the cup from the different levels of sounds the cup makes each time it is put down as the cup gets emptier (and if the saucer were also equipped with safe, mild-heating capabilities), it could adjust its temperature as a heating pod taking the level of coffee, the amount of time the cup stays on the saucer, and the drinker’s preference for each different types of coffee, the time of the day, the weather, the purpose of taking the coffee—if it is being taken while the drinker is reading a book (perhaps, a longer coffee-time and more sustained milder heat would be ideal) or having a business discussions (perhaps, a shorter and less sustained but slightly increased heat in this case) and other conditions. In this scenario, the disconnected saucer would have still been able to serve a useful function—maintaining the desired temperature for a perfect cup of coffee for the purpose—without a direct communication arrangement with the cup and by being able to take proxy data (different sounds of the emptying cup, for example) to establish the necessary information.
IoT for BD: This time our magazine’s theme is ‘Smart Homes’. Smart homes are among the most common applications of IoT. However it is still a foreign concept in our country, so what is your personal opinion about smart homes in Bangladesh and do you think it will be a giant leap in digitizing our country?
Mohammad Oli Ahad: Developing smart home solutions specifically for our country’s needs will certainly be a huge step towards digitizing Bangladesh.
In our country, 46 children die every day from drowning and it causes 43% of deaths among children under 5. A device that could alert family members when the young children in the house would cross the main door or go more than 50 meters out of their homes or go near the adjacent water bodies would be a very useful one to address this tragedy.
We can work to deploy smart home solutions like this to solve the various crises that we face every day. For instance, we could have an entire slum turning into a large smart home which would allow them to share common resources lot better.
IoT for BD: Will your company be interested in providing IoT services or Smarthome services?
Mohammad Oli Ahad: Right now, we are focused on our AI products. Once we learn more about this area, we do wish to gradually expand our operations.
We have a dream of developing a robot assistant that will help people with various regular activities—tutoring children, reminding about important meetings, provide support to pregnant women et al. It would be able to understand all local dialects. Our goal is to make these assistants affordable so that anyone who owns a refrigerator at home can also have one of these! However, this dream is quite a long shot away, and we hope to stay focused on making this dream a reality.
IoT for BD: Since you have a wish to proceed with these services, how do you plan to establish your market in the country?
Mohammad Oli Ahad: This should be done the same way that any challenge could be faced—break it down into smaller pieces, and solve each one effectively.
We will explore more with Retail AI, and we might even start a digital assistant out of it. The digital assistant might help small businesses in various ways, like minimizing the cost of hiring workers for small grocery shops. If we succeed in developing useful products, people will opt to use them spontaneously.
However, it will not be possible unless we get adequate capable resources. Our talents are our biggest resources and we need to do whatever we can to encourage them to work for the country. We need people who will be able to contemplate the various problems that we have in the country, will love to bring changes to solve those problems and will have a mentality to work and be self-driven. With people like these, anything can be accomplished in Bangladesh!