Anju Gupta leads the Global IT Strategic Collaborations at Monsanto, where she works to establish industry partnerships. She has a Ph.D in Quantitative Genetics for Ohio State U. and has a number of patents. She took time out to speak to AI Trends Editor John Desmond.
Q. Thank you for taking the time to talk to us. Can you tell me, what is the status of artificial intelligence work at Monsanto today?
Anju: I do believe that we are making a lot of progress with AI within agriculture. We have a number of deep learning projects in flight currently. That doesn’t mean it’s truly AI, but it’s leading up to AI in terms of a recursive response, where the machine is giving the response that we need it to give, with an individual provisioning information to the machine constantly.
In majority of cases, our internal AI work has been done in collaborations with external startup companies. We did a few initial AI projects with startup companies and that led to some deeper partnerships. We also worked with Monsanto’s executive team and Board of Directors to help them understand broader impact of machine learning, deep learning and AI in agriculture.
We also recently hosted an artificial intelligence fellows symposium within Monsanto. We brought in all of our fellows and a lot of external speakers in order to have a broader engagement with AI leaders. We were able to bring in product heads of AI from Intel, Microsoft, Google, Unity3D and few startup CEOs with AI focus.
Q. Are there any proof of concept experiments that you can talk about?
Anju: Yes, we have done multiple proof of concept experiments with AI. We worked with image recognition, using deep learning, customer analytics using ML etc. Multiple projects have gone forward from that. We are also looking at using advanced analytics strategies to determine the impact of product performance.
Q. Could you talk about AI work that’s grown out of the Climate Corp acquisition in 2014?
Anju: Given the nature of our digital agriculture business with The Climate Corporation, AI is core to our future products. There is shared learning between Climate digital products and Monsanto’s pipeline. Though we have been applying machine learning methods to imagery insights for years, some of this work is in early stages. Yet, we continue to advance these capabilities well and are leading the industry in this space. You will continue to see AI play a major role in the digital offerings in Climate FieldView going forward, especially in the area of disease insights and disease diagnosis.
Q. Are you using AI to help farmers increase their yield per acre?
Anju: Yes. We will be using advanced analytics, essentially our AI platform, to help increase yield per acre. The goal is to be able to predict yield better, which will flow across both Monsanto and Climate. We are provisioning our digital agriculture platform to our customers, and that is primarily with Climate FieldView. So we will have AI products across our entire digital agriculture platform. So, it could be for yield per acre, it could be for seed selection and placement recommendations, it could be for disease identification, it could be used in management as well.
Q. Is there anything Monsanto needs from the AI hardware and software industry that’s currently not available?
Anju: Oh yeah, absolutely. When I talk about what we do, especially launching a product externally, we have a very complex situation. We have got complex data with heterogeneous variation. It’s very similar to what you see in the healthcare industry. We also have an immense amount of individual plant genetics and environmental information, which we have accumulated over two decades. We generate four cycles of data per year. We have soil data, weather, topological data, elevation data etc. We take all this data and the environmental components, and it becomes a very complex problem for anybody to solve. As I think about what AI is doing today, the complexity of the problem getting solved is for low hanging fruit, really. Some of the more advanced technology that’s getting inserted into driverless cars, for example, is not available for enterprises at this time to be able to tap into and play with. So a lot more advancement needs to happen on how to integrate all the systems or data together, to get a value out of it.
The software is definitely a huge component. Tensor Flow [open source libraries for machine learning/deep learning] is definitely the best usage for us right now, but that’s the only platform today for us to be able to play with. There’s a few others that do AI, but Tensor Flow has been one of the best. We have also been aggressively going after the hardware piece. The reason we started working with Nervana Systems [acquired by Intel in August 2016] is because they were bringing both components: hardware and software, which becomes a critical piece for us. The sequencing is in terabytes per day essentially. So it’s a complex problem for us and the needed hardware is not there. Google is talking about that; we have yet to see how that works out. We are doing some plug and play with the Intel hardware chips. We are looking at NVIDIA as well. But at high level, it’s also how real intelligence is going to play into this entire space with artificial intelligence. So far, what I’ve seen is they’re yet disparate, however, the industry is maturing rapidly which creates opportunities for us.
Q. Can you elaborate on what is not coming together very well?
Anju: The real intelligence with the artificial intelligence. An ideal goal will be a classic case where human intelligence comes together with the artificial intelligence to be able to deliver a winning game, right? But that’s a very defined space where we are working through. Currently, when I think about a true AI world, it’s apparent that we still need a human. You still need a human in the loop for AI to be able to perform, where we want it to be able to perform. And that may be because we’re so early in the space right now and eventually when things roll together, the human in the loop will go away. That’s not the case for us today. And simple things that we take it for granted – labeling of data – we take it for granted really. We don’t have machines today to be able to do that. AI will perform well only if a human in the loop has been able to give information on labeling of the data, which is as basic as it seems. But AI can only do as much as we want it to do and perform very well if data labeling is there. And that’s, to me, that’s very fundamental.
Q. You have talked about needing an AI development sandbox.
Anju: Yes it’s true. In majority of enterprises that we have talked to, there’s no sandbox for data scientists to be able to play with to build their AI models and work together. Being able to have that sandbox capability for any enterprise will be critical for accelerating AI into every industry.
Q. What do you think is the best approach an enterprise can take to gain the optimal advantage from what’s available in the market today?
Anju: We have taken a very different approach to adopting AI within Monsanto, and I am sure there are other enterprises taking a similar approach but have not seen it yet. We launched a digital partnership team within our IT strategy group at Monsanto. The goal of the digital partnership team is to seek out external innovation rapidly and bring it inside Monsanto. If we hadn’t done that, we wouldn’t have been able to test the waters of AI at all. And having those initial partnerships with startups has led to acceleration of AI adoption at Monsanto. Also being given the freedom to operate from senior leaders has also been very critical.
When I think about the AI space and the industry that we are in, it’s extremely hard for us to compete for engineering talent, especially in deep learning and machine learning. The only way we can get there is by aggressively partnering with the startup community and giving them a test bed as well.
I also firmly believe that the senior leadership needs to be bought in completely that this is the right thing to do. When I have talked to other large enterprise at AI conferences, this is where I see most other enterprise are struggling with.
We can go for the low hanging fruit with AI as well. It doesn’t have to be a driverless cars all the time. However, analytics can drive pretty much every facet of a business. And that’s the camp that I am in. If that is a reality, then it needs to be in service heavily within each facet of an enterprise, whether it’s retail, or manufacturing, or any other business.
Q. What should we expect from Monsanto as a result of your efforts?
Anju: You can expect a case study that we will publish externally. It’s under the umbrella of digital transformation; it talks about how digital partnership innovation which has been a primary driver for some of the AI components for us. And if there’s one thing I can really suggest talk to enterprises like ours who are wanting to get into the space, is go out and seek out partners. Seek out both industry leaders and startup companies to see what all is possible between the two.
It’s a critical piece for the acceleration of AI within different industries. Digitally born industry leaders don’t have to think about it. It’s easier for them to hire talent, they’ve got the name-brand awareness and can afford to get those talents. But, 90% of the other enterprise are wondering how to get there faster.
Learn more at Monsanto.com.
This content is original to AI Trends.
Let’s block ads! (Why?)