AI implementation in practice
Borre Moolenaar, Chief Data & Analytics Officer at Cargill, will discuss AI implementation in practice on Friday, March 15th at the IFFI Event ‘Artificial Intelligence in Commerce’. Before successfully utilizing AI within the company, you must first have a good understanding of what it means. This is currently being actively pursued at Cargill.
Artificial Intelligence still sounds like a distant concept for many, but we already use the technology daily, often unconsciously. “Autofill in messaging apps or word processors is a form of AI,” says Molenaar. “It’s based on machine learning. There’s so much data available now that AI can make predictions on how the sentence should be completed. This is a form of machine learning.”
Deep learning
The next step is deep learning, where computers can work on multiple problems simultaneously. “This can involve analysing images and photos using specific software,” explains Molenaar. “This interpretation can be converted into a value in a table. This is necessary for digitizing non-digital items on a large scale. The next step is generative AI (Gen AI), which can generate images or texts based on a set of prompts. With these prompts, the software can predict in real-time how the response should be constructed.”
Bias
Cargill already uses machine learning, but the company is cautiously exploring the possibilities of Gen AI. Molenaar emphasizes the importance of caution: “You need to realize first and foremost that the system is only as good as what you put into it. This also applies to offline systems—if you make an estimate or prediction based on data collected by an employee. Garbage in, is garbage out. However, it’s important to recognize that Gen AI makes predictions based on existing data, which often carries biases.”
Sample bias
A great example is the chihuahua versus muffin computer vision. It’s based on a popular internet meme, that demonstrates the alarming resemblance shared between chihuahuas and muffins. Mariya Yao put AI to the test, and did some research based on the question: how good is modern AI at removing the uncertainty of an image that could resemble a chihuahua or a muffin? She found that even the world’s most advanced machine learning platforms are tripped up by the facetious chihuahua versus muffin challenge. The problem is caused by sample bias. This happens when there’s a problem with the data used to train the machine learning model. In this type of bias, the data used either isn’t large enough or representative enough to teach the system.
Secure environment
Caution is also warranted due to privacy-sensitive information. “As a company, we can’t just use a public service. That’s why we only use Gen AI in a secure environment. If we want to use existing services in the future, we need to know how to protect our data, as well as the data of customers and employees.”
Displacement
The next question concerns the output. Are you aiming to accelerate the business process with it, or is AI meant to take over tasks? And to what extent does this involve displacement of labour? I do see benefits in using AI, especially for labour-intensive processes. But we’re not there yet.”
Learn more
Want to learn more about the use of AI in the workplace? This year, all IFFI events are about the use of AI. Stay Tuned!!
Borre Moolenaar