Dr Ravi Shankar has worked in analytics leadership roles for AMEX, BLACKSTONE, DELL, GE, PwC, and others and has over 30 years of diverse yet profound expertise in employing analytics-driven teaching consulting business & social impact across numerous sectors.
With a Doctoral degree in Econometrics from the prestigious Indian Agricultural Research Institute (IARI, New Delhi), Dr Shankar has taught at various institutes like UASB, IARI, IIMA, XIME. Currently, he is a Visiting Prof of AI-ML at CWRU, Carleton Univ, IIT-R, IFIM, INSOFE, and Praxis Business School. He also regularly conducts corporate learning programs for CXO’s in AI-Adoption, Design Thinking for Data Science, AI-driven Strategy & OB, and BFSI, Pharma & Manufacturing sectors.
Over three decades, Dr Shankar has mentored a slew of Agribusinesses and AgTech startups. Besides, he is actively involved in advising a handful of NFP’s in the rural livelihoods space, focusing on SDGs.
He is also a senior Venture Partner with an early-stage VC focusing on AgTech, InsurTech, and Social Impact space. His passion is in mentoring AgTech startups with a focus on smallholder sustainability.
He is a teacher, mentor, and advisor – all rolled into one. After three decades of corporate hum-drum, he has now gotten back to his first love – teaching & consulting. Outside of work, he is interested in golf, spirituality, and making up time for family.
Dr Shankar shares his expertise on AI in the agriculture sector with the readers of Business Review Live.
What made you include Artificial Intelligence in the Agriculture sector, and how do you think it will impact the sector?
The real test lies in ‘application” – one (AI) is the most exciting, disruptive & transformative digital technology & the other (agriculture) is an age-old profession. In India, it is a way of life where more than 60% of the population is directly or indirectly dependent on this sector.
In order to form an Integrated Digital Farm, the micro and macro farmers will need to be associated with technology. How do you think that will be possible in India?
There is much innovation happening in the public sector research space. However, the gap from “lab to land” needs to be bridged effectively. Here is where digital technologies that are affordable & accessible (in vernacular languages) come to the table. Agtech startups are one potent part of the solution. The agtech startup space is vibrant & thriving (despite the pandemic). The obvious enabler is the penetration of mobile technologies & internet accessibility along with tv.
Do you not think that Machine Learning may not be possible in all parts of the country (especially in the Farming sector)?
ML is only a means to an end. The comprehensive benefits of AI-ML can be leveraged only if mobile tech (5G) and internet penetration are universal. The digital divide is for real & we need to address this inequity on a war footing which the central govt is doing.
Also, replacing human systems with machines may lead to an increased unemployment rate. What is your take on this?
The irony is that labour scarcity is the most significant predicament in rural India. Wherever possible, we must supplement intelligent automation with human labour. The key is to upskill and reduce the cost of cultivation.
Is Machine Learning pro-customers or pro-farmers?
ML tech is neutral – like any other technology, it depends on how it is used / application. There will always be naysayers. However, we need human-centric AI-ML solutions that are ethical, responsible, affordable, accessible to rural India.
What role does digital agriculture play in transforming agriculture?
Digital agriculture has the potential to make agriculture more productive, more consistent, and use time and resources more efficiently. This brings critical advantages for farmers and broader social benefits around the world, and it also enables organisations to share information across traditional industry boundaries to open up new, disruptive opportunities.
How do institutes like ICAR-SAU, NFP, and CGIAR integrate all stakeholders and solution suppliers under one roof?
This is a million-dollar question. Various agri research institutes (State / National / International) currently collaborate via faculty / germplasm exchange programs. Also, there are collaborative research programs; however, there is a vast scope for improvement from a collaboration and cohesiveness perspective
Will the data that the company holds be a threat to the data providers? Won’t there be a fear of leaks?
Data privacy is a real issue. However, there are frameworks for data protection, anonymisation & encryption. It is up to the governments (State & Central) to frame and implement effective data protection laws so that the end consumer’s interests are not compromised. At the same time, without an open data access regime, digitalisation cannot realise its full potential. Therefore, good balance is the key.
Will Machine Learning prove to be an expensive method for startups?
Not really. Most AI-ML technologies are open source, and at scale, they are affordable
Can Artificial Intelligence predict the crops and the soil and the yields of farmland as efficiently as experienced farmers?
Yes & beyond. Essentially, AI-ML is data-driven; hence can beat human expertise any day given the significant advances & breakthroughs in the last decade
What can we point out as significant contrasts between a digitalised system and human operations?
While the former automates the process & renders it efficient & effective, the latter is slow, dependent on availability & skills, and consistency is a question mark
Machine learning, it is claimed, allows for the creation of a non-linear regression model for credit risk assessment. How?
Non-linear regression models offer greater accuracy & precision for credit risk assessment for the simple but fundamental reason that most relationships in the real world are non-linear. The premise of linear relationships amongst variables is far too simplistic and hence breaks down in real-world settings.
How effective is a classifier in assisting a loan applicant?
Pretty effective – I would like to rephrase the question. Machine learning techniques are effective (the degree depends on the quality of underlying data) in predicting a bank/lender loan applications. However, when it comes to a loan applicant, they have to improve CIBIL score constantly (by following credit best practices like repaying debts on time, not taking on multiple loans, trying to maintain a healthy mix of credit, not utilising your entire credit limit, increasing the credit limit, ensuring credit report is error-free and so on.)
How is it more accessible for lenders to easily apply algorithms that analyse consumer risk using machine learning (ML) models?
Most lenders these days do it by default. In India, for example, the top 10 public & private sector banks have data science capabilities. They use ml models to decision on analyse consumer risk using CIBIL score, past transaction history, and a multitude of other external variables.
What is an artificial neural network, and how does it work?
Simply put, an artificial neural network (ANN) is a part of a computing system designed to replicate the way the human brain analyses and processes information. It is the foundation of artificial intelligence and solves complex problems that humans or statistical standards would prove difficult or impossible.
ANNs are controlled by multiple nodes, which emulate biological neurons of the human brain. The neurons interact with each other, and they are connected by links. The nodes perform simple operations on the data based on the input data.
Then the result of these operations is forwarded to other neurons. The output of every node gives a value called a node or activation value. Each link is associated with weight. Anns are capable of constant learning, which takes place by altering weight values.
What methods does machine learning use to capture non-linear relationships?
Some examples include DECISION TREES, NAÏVE BAYES & KNN
What role can Orthomosaic image maps play in helping farmers plan and intervene more effectively?
They help in precision farming & drone mapping. The applications vary from assessing soil moisture to predicting crop yields to recommendations on inputs like fertiliser & chemical sprays (right quantity at the right time).
In agriculture, AI has revolutionised imaging analysis; how do they use AI to analyse yield estimation?
Deep learning models can perform pre-season and in-season predictions for different crops. Ai models use crop calendars, easy-to-obtain remote sensing data, and weather forecast information to provide accurate yield estimates. Other yield prediction models are CSM & NDVI models.
Would you like to add information that might prove to be step-by-step guidance for the implementers?
Keep it simple – follow Occam’s razor principle at all times. Solve the real world. Consider affordability, accessibility & scale. Technology is only an enabler.
“Define -develop-iterate-test-implement”. Use design thinking principles & agile framework for sure.
