Artificial intelligence or AI is being talked about by everybody from the highest to the lowest.  I sometimes think that very few people really know what AI is, but everybody is in love with it !  And most people think that by magic AI will solve all the problems.  Unfortunately it is not so. 

AI is a powerful tool but is still a tool. It is not a panacea. With its tremendous computing power AI can collect data and lots of information from existing things and try to create a meaningful output. Sometimes it also hallucinates and gives you totally useless output. It is still dependent on the information available on the internet and cannot produce anything new.

For example, if one is working on a new project which might result in making a new discovery and wants to use AI, it will give you feedback based on what exists and nothing about the discovery that you are working on. So one should  be careful and always consider AI as your servant and not your master and use your vivek or wisdom to double check the answers provided by AI. That is an important lesson for all of you.  

Where things really go wrong is when you use AI tools to write essays, reports or other material without understanding the subject. That is very unethical and that is what is happening in most colleges, universities and with researchers where a large number of papers or reports are being written using AI tools.

The output is written in a flowery language, but it contains very little useful or often totally wrong information. It is this dishonesty of using AI for writing papers and reports without knowing much about the subject, is what you should keep away from. Because by being dishonest you do not fool others but only fool yourself.  This type of behavior can take you only to a certain distance.  In the long run it will be a disaster. If you want to make anything of your life, then be honest and work ethically.

By working honestly, we can really understand the problems and once we understand the problem we can solve it. By being dishonest or ignorant we gloss over the problem and hence can never solve it.

AI, however, can be used as the first line of research and the references from this research should be thoroughly evaluated before accepting its outcome. Since AI analyses large amounts of data, it can sometimes provide proper direction for further research thereby reducing the time for new discovery. This is the line of thinking behind creating new solutions, molecules and materials.  Nevertheless it should be backed by proper human research and validation of data. Otherwise the line between reality and AI based virtual reality will be blurred.

I will now share with you a few examples of how we were able to create interesting solutions at our institute by collecting good data.

Electric Rickshaw 

In 1995 we started work on the electric rickshaw development.  As some of you may know we developed the e-rickshaw in late 1990s and were the first in the world to do so. It was copied by the Chinese and others and now has become a world-wide phenomenon. In India itself around 60% of all electric vehicles are e-rickshaws.

Since this was a new invention, no data was available. To collect this data I interviewed, in a couple of cities, hundreds of cycle rickshaw pullers, owners and foot path manufacturers and developed the database for the number of cycle rickshaws plying in India. Our idea was initially to convert the cycle rickshaw into pedal assisted electric rickshaw. Later in 2000 we developed the complete electric rickshaw that you see today.

Even today after 20-25 years that data is still quoted in literature though the number of cycle rickshaws have reduced drastically. This is not a good sign since by now it should have been modified based on more recent data. But recording good data is difficult. Thus it is very necessary to develop a good database and later on update or modify it.

A sad part of this (and is true in most of the technologies being used in India) is that the majority of the parts of e-rickshaws are still imported from China and other countries.

When we developed this e-rickshaw in 2000 we got the PMDC motor developed in Pune, used locally available lead-acid batteries and commissioned a Pune party to design a controller. It has been nearly 25 years since our efforts and with rapid developments in technological progress I thought we should be using only Indian designed and made parts. Yet unfortunately not. So electrical engineers think about how to make efficient PMDC and hub motors; efficient and cheap controllers; and batteries.

We cannot become Vikasit Bharat if we cannot produce materials and components from our own inventions and make most of the things in India itself.

Also as some of you, who are from North India, know that e-rickshaws are a very common sight in Indian cities. They clog the roads and in most places create a nuisance. AI can help them to become Uber e-rickshaws. They provide last mile connectivity and Uber or Ola models for these rickshaws can help everybody. However to develop this model good and hard data on their numbers, their characteristics, road topography and conditions are needed.                    

Biomass power plants

Another example is from developing biomass power plants strategy. In the early 1990s we developed this strategy which became a national policy and was implemented by MNRE.

In 1981 when I came back from USA to Phaltan, Maharashtra, in western India, I found that the sugarcane farmers in and around Phaltan were burning sugarcane waste (dry leaves) after harvest. This was their waste disposal strategy since they wanted their fields to be ready for the next crop. Thus acres and acres of sugarcane fields with waste were burned. It polluted the environment and was a loss of a precious energy source.

Also in those days there was a tremendous shortage of electricity in rural Maharashtra. So I thought that if we can use this waste for electricity generation then we could solve the twin problems of pollution and electricity shortage.

For that we first developed excellent database for the amount of agricultural residues available in Phaltan Taluka and then looked around for technologies for using this waste to produce power in the range of 5-10 MW capacity.

In the early 1990s this was a new concept, and it became the basis of the national Taluka Energy Policy.  And till today about 10,000 MW of biomass-based electricity generation plants have been installed in India based on our study. The Government of India also used our work as the model for mapping biomass residues availability in all Talukas in the country.

Similar is the story of lots of other pioneering studies and technologies we developed in our Institute. You can read about all these interesting stories in our book appropriately named “Romance of Innovation – Human Interest story of doing R&D in Rural Setting”.

Again AI can be used in these biomass power plants. Since the raw material (agricultural residues) comes from different sources with different densities, moisture content, calorific value and varying distances there is a need to optimize the availability of raw material at affordable price. AI tools can help with this.

Similarly AI can also be used in improving sugar recovery in sugar factories. Presently the haphazard planting, poor harvesting, competitions among sugar factories makes the whole exercise ad-hoc and in the process precious sugar is lost. This wastes lots of precious resources of both land and water.

AI can help with the planning of sugarcane planting so that proper cane with optimum sugar is delivered at the right time. AI can also help in optimizing the distance travelled by the cane from field to factory. This will also help the farmers both in terms of time and land management and getting proper remunerations. However proper AI strategies need to be managed by the appropriate sugar factories.

Way Forward

I also wanted to share with you that in both these examples the problems were right in front of us. What was needed was hard work in creating good database and providing appropriate solutions. I hope in your careers this is what you should do. Look around for problems that you face, then develop a good database for them and only then use AI tools to develop strategies to tackle them.

Using AI tools without having a good database is like producing an omelet without eggs. I would like to share with you another example of poor use of AI without proper ground-based data. A couple of years ago I was asked to evaluate a project for substantial venture funding. The project was to develop detailed map of carbon content of Indian soils by using satellite technology and AI. The technology was based on Infrared (IR) radiation from the soil and its relationship with carbon content.

For those of you who are familiar with the science of radiation heat transfer will know that IR changes drastically with the particle size, material of the soil and its temperature. All the data about these properties need to be developed and the satellite readings to be calibrated accordingly. Also within a plot, soil properties change within a few meters. Thus there are lots of variations and it requires a solid large-scale soil database. In its absence the project was rejected. On paper this project looked very sexy, but the reality was otherwise.

Thus even when advanced AI tools are used for inventing a strategy or process it should be thoroughly validated by actual hard ground-based data.

AI is not cheap  

Finally it is a sobering thought that AI is not cheap. It is very costly to run and consumes a huge amount of electricity. There are estimates that ChatGPT alone uses about 600 MWh/day of electricity. This much energy can provide all the electricity requirements of two talukas with about half a million population.

Thus it is a big challenge for young researchers to produce strategies and devices that will do AI reasonably well with less power. R&D is already being done in that direction but needs acceleration .

So my advice to researchers and youngsters is to improve AI technologies, collect data honestly and methodically on problems and then use AI tools together with your 'vivek' to analyze and solve them. With proper data, AI can then provide right direction towards the solution path and show the actual trend on which to proceed.

(The writer, an IIT and US-educated Indian engineer,  a 2022 Padma Shri award winner, is the Director, Nimbkar Agricultural Research Institute, Phaltan, Maharashtra. The article is based on a lecture at 7th IEEE International Conference on Emerging Smart Computing and Informatics, AISSMS, Pune. on 5 March 2025 He can be reached at anilrajvanshi50@gmail.com)