The article is part of WhiteHat Magazine’s Fall 2017 edition, “Disaster and Development”.

Support independent journalism — subscribe today!

Artificial Intelligence and Precision Agriculture is Changing the Future of Food

by | Nov 6, 2017 |

Artificial intelligence is transforming our lives, but nowhere might this transformation be so promising as it is in agriculture. An increasing amount of our food is being grown with input from algorithms, and in a signal of what’s to come, in October 2017, John Deere acquired an artificial intelligence company that had been named among Inc. Magazine’s 25 Most Disruptive Companies.

To find out more about the future of artificial intelligence and how it is affecting agriculture, WhiteHat Magazine spoke with Anand Rao about the future of artificial intelligence and how it is impacting the agriculture industry, public policy, and our future jobs. Anand Rao is a partner at PricewaterhouseCoopers and leads up their global artificial intelligence efforts. At the 2016 San Francisco EmTech Digital conference, Anand gave a speech titled, “The Symbiotic Relationship Between Man and Machine.”

This transcript has been edited for clarity.

***

[WhiteHat Magazine]: You talk about there being three types of artificial intelligence: assisted intelligence, augmented intelligence, and autonomous intelligence. Could you give me a brief explanation of those?

[Anand Rao]: Assisted intelligence is where I think most of the things–not all of the things, but most of the things–that we do today are. Where I see that as pretty much AI or artificial intelligence doing the tasks that we are already doing. Maybe it’s doing better or faster, or it’s too mundane and boring and we don’t want to do it, so we want to give it to them. That’s what I call assisted intelligence.

In assisted intelligence, the nature of the task doesn’t change, it just gets automated. Whereas in augmented intelligence, the way we do certain things changes substantially because now the machine is helping us. So, we work together and because we work together, we teach the machine and the machine teaches us or helps us make better decisions. We wouldn’t say the machine is teaching us, but we would be better off making some decisions, which is very true today.

Just imagine just twenty years back, if you went into a room alone, the amount of knowledge that you would be able to share with, let’s say, someone who comes from outer space. Your ability to answer the questions that they ask would be limited compared to now. With this, we can pretty take much information from anywhere in the world, within a second. So now, we almost take this for granted, but this has basically changed our lives. So, for example, I went to Fremont. I didn’t look at the map, I didn’t look at anything beforehand. Just going to the car, press the address, and it takes me there, step by step. We are so dependent on it. That’s how much our life has changed. You may not think that’s great intelligence, but just multiply that more and more in what we are asking, because the oracle comes back and tells us the answer.

In autonomous intelligence, there reaches a stage where we are saying, “yeah, I don’t want to be bothered every time, just take it away from me and do it.” So we talk about decision rights where the machine is making the decisions, not just automating tasks–where actually it’s making decisions that are important to us. Investing decisions or life-saving decisions.

I don’t think everything will go fully autonomous, some might and there are a number of factors that influence it. It may be more the social acceptance of certain things that might push it there, and in some cases we might say, “no, we don’t want some of these things to be given to machines, we want to hold onto it.” For how long, who knows? That’s the way we see those three levels.

[WhiteHat Magazine]: I can definitely see it being a cultural shift and taking time to adapt to that.

[Anand]: Yes.

[WhiteHat Magazine]: In your talk, you focused mainly on augmented intelligence, because that seems to be the next stepping stone for us in this area. Could you give a brief, broad scenario of how augmented intelligence would work just in our day-to-day lives?

[Anand]: I’m going to give more of an enterprise example. On the consumer side, again, there are lots of ways this is changing–various apps that help us in various things. I think most people are comfortable with thinking, “it’ll help us, it’s helping us, it’ll help us even more.”

But, in the enterprise area, pick any sector that you want, even something as old as agriculture. People say, “Agriculture? What’s AI got to do with agriculture?”

But, that’s probably one of those areas where AI has penetrated so much. There’s something called precision agriculture where artificial intelligence is actually helping the farmer. Helping the farmer to an extent that it’s pretty much managing everything.

For the farmer, it’s the yield that is important. How much of wheat, or corn, can I get from my field? That’s the key thing for them. What the AI system is actually doing is precisely laying out a use-specific field in Iowa or Illinois. When should I plant corn? Not only when should I plant, what would be the yield of that when I plant today? The yield will be between 200 and 250, for example. Then it asks, “what is the soil content and how much of water will hold on?” And then it will say how much you should water.

Not only that, then it looks at your planting today, and it monitors how many daylight hours are going to provide sunlight. Based on that, it actually calculates the number of leaves a particular corn stalk will have, how much photosynthesis would be necessary for the corn cobs to develop, and when will the corn cobs actually mature. It also looks at–and I didn’t know this until I worked on one of the agriculture programs, I’m not a farmer–there’s a moisture content in the corn and you don’t want to harvest it while the moisture content is very high. So, you want to let it go for a few days and it measures that.

All of that, the program is doing and it’s telling the farmers. The farmer is now essentially consulting with AI, and of course, they are entering all of the stuff, or the various inputs, and their life now has fundamentally changed. They look at the system and say, “on my huge ten-acre property, I don’t have to irrigate the northern part of it because there’s going to be a front coming in.” They are tracking all of the weather data. What is interesting is the value of your yield changes as the weather changes. So, if there’s a flood, or there’s a draught and five continuous days of sweltering heat, then the yield actually goes down. So a real-time view of their farm is something that they’re getting.

That’s just agriculture, when you think, what can you do in agriculture? But if you also look into banking or insurance or health care, the amount of change is enormous.

[WhiteHat Magazine]: To be able to process that data or get a handle on all that data, that’s huge. And, with the agriculture example, in particular, I can see that having huge benefits for emerging economies around the world

[Anand]: Yep. That’s right.

[WhiteHat Magazine]: Do you have an idea of what the price point of these products might be? Or how accessible these might be to emerging markets?

[Anand]: The interesting thing is there are quite a number of companies in this area and the costs are coming down so much that it’s almost becoming…it’s not free, but it’s sort of bundled and very affordable for a farmer. Across the board. And that’s one of the reasons why you see some people saying now that this precision agriculture can lead to a substantial increase in production–just by using precision agriculture and not doing anything else.

[WhiteHat Magazine]: What is leading to this rise in the AI industry? You had mentioned that you had been in the industry previously and then you switched careers, and now it’s starting to come back. What has changed and why are more people working on this area now?

[Anand]: A number of things have changed. The first one, I would say, is the computing power. With Moore’s Law, we are seeing more and more processing power and memory capacity. That’s definitely one of the key drivers. In addition, I think some of the advances in parallel processing and deep learning–that’s not very easy to do or very fast to do with normal machines.

I would also say is we are also seeing a movement towards open source. Because of that open source movement, what has happened now is a community of people are building these things and sharing it. That has had an enormous impact on how we can build. This isn’t just individuals, but large companies were already working on some of this technology and making it open source. They don’t treat code as something proprietary–they treat their data as proprietary–but the code is available to everyone. It helps them because there’s a larger group of people using their code who are potential employees or potential testers of the code. You don’t need to start anything from scratch these days, you essentially build off other packages out there. So it’s more like assembling things as opposed to manufacturing a part. Let’s say you’re making a car, you have all the parts there and then you’re just going and picking the various parts and assembling it together rather than starting everything from scratch.

[WhiteHat Magazine]: On the topic of open source, if artificial intelligence is going to be coming into hospitals and potentially helping us in making decisions over saving lives, is there a concern that we might have another Heartbleed sort of problem with using open source to build these programs?

[Anand]: I think open source has its advantages and disadvantages. At least in the corporate world, people get very worried: “how can I trust a piece of code that multiple people have written with no quality control?” That’s where you see companies coming up with open source. They might be a community, but there’s also a company that backs up support, so that at some point, even though it’s open source, they validate that quality is maintained. Even in the healthcare industry, I would say it leads to more people trying out things, and the better it is.

Privacy is very important. Confidentiality of information, especially in healthcare, is very important. What we’re also getting to now, which is not really AI, but blockchain and some of the advances that are happening in cryptography, essentially allow us to work with some of this data without necessarily knowing all of the details around the data, private information. That also would play into being able to do things that we were not able to do before.

[WhiteHat Magazine]: So, using an example here, in developing electric cars, the biggest obstacle that this industry has run into is battery technology has not advanced quickly enough. What is a similar obstacle that the AI field is facing?

[Anand]: Good question. One of the things, at least one of the big obstacles is there is a lot of data, but not the data of the right kind. Yes, there’s lots and lots of data that we are producing, but then we are not tagging some of that data, especially in machine learning and deep learning. That’s something that you do need before you can actually have a machine learn. And, that’s a big challenge, whether it’s MH data or DO data or text data, Any of that requires a body of data that is labeled and that’s something that, again a lot of these companies are spending literally millions of man hours trying to label that data.

[WhiteHat Magazine]: At the World Economic Forum, there was a discussion was on the Fourth Industrial Revolution and how that’s going to change our society, and change how we interact with each other. One prediction was that AI programs will eventually be sitting on corporate boards and something I would fall under autonomous intelligence, I believe. How far off is something like that from happening?

[Anand]: In fact, there is a company in Hong Kong which has already. So that’s already happened. What that particular AI is doing is essentially crunching through all of those numbers. It’s no different from any of other kind of AI which gets data fed in–all of your stock market data, quality data, and so on are historical–and it knows the economies of different countries, different products. Then it is projecting forward, just as a human would. Arguably, it’s probably got more power, processing power to be able to say what the future is. That’s what they are using, and then it gets to cast a board vote on certain things.

[WhiteHat Magazine]: How do you see AI affecting public policy?

[Anand]: There are a number of areas where AI would impact. At one level, I think there are very specific groups within the government which obviously AI are looking at it. Their primary concern was less about AI coming and taking away the jobs, but it was much more around are there biases being introduced by certain AI. Because obviously we are seeing more and more things–whether it is underwriting, lending, any kind of a decision in the financial world, or even in the healthcare world–now being made by some kind of algorithm. What the government is interested in, or keen on making sure, is it doesn’t introduce any specific bias towards one group of people, whether it’s ethnic, or medical, or age-related, or someone who is just a loyal customer. You can argue that a loyal customer should be given more preference to someone who just came in, so that’s what their concern is. That’s what some of these algorithms do: they personalize to a certain extent. I think what the government is worried about is that personalization leading to certain types of biases.

And then, of course, there’s a challenge of: if there is such a bias, how do you know that there is such a bias? As a government, how do you even see that there is a bias? How do you even regulate or control that bias? That’s where I think public policy is focused on now.

Then, in the medium term, look at autonomous cars for example. The federal government is looking at what kinds of changes we need to make and how we need to balance the different aspects. What’s interesting about autonomous cars is an enormous savings in terms of people, just looking at the safety aspect of it. The number of accidents has actually gone up because of distracted driving. So it almost becomes a moral duty of the government to prevent some of those things, using automated technologies. It may not be fully autonomous cars, but even requiring certain types of technologies. If you veer out of the lane or having sensors fixed–all of those things they might start regulating. That is sort of the second level where the government needs to balance the safety of the people and the lives saved.

And long term, I think that’s where it gets interesting. If artificial intelligence is going to penetrate every aspect of life–which I’m pretty sure it will–then what implication does it have on society and jobs? As a result, what is a right to work? Should we provide a minimum basic income? That’s sort of a more policy level decision, that the government is probably not worried about quite yet. I know the World Economic Forum is talking about this, what might be the impact. There are two views there. One view basically says that this will generate far more, newer jobs, so it’ll change and the types of jobs will be different. There’s also another view that the speed at which those jobs change would be so fast that they’ll be certain groups of people who will be displaced. How do you take care of those people? There will always be groups of people who do very well, but there will be groups that will be displaced. What do you do for them, as a government?

[WhiteHat Magazine]: Any final thoughts on the state of AI today?

[Anand]: Artificial intelligence is a broad umbrella of different things. It has machine learning, deep learning–which is a form of machine learning–natural language processing, advanced analytics of various types. Often, even the enterprise people get confused in, “I want to do AI, not analytics or not advanced analytics. I don’t want to do natural language processing, I want to do AI.” So, there’s still a lot of confusion around that word and, I think in my view, all of those things together constitute AI. There are lots of useful things across various AI areas, intelligence agents, conversational agents, or MOTS as people call them. All of those things are part of AI.

[WhiteHat Magazine]: Thanks for talking to us today.

A Different Perspective.

A Different Perspective.


In-depth analysis and interviews about the science and technology industries, delivered once per week to your inbox.

You have Successfully Subscribed!