AI policy is defined as public policies that maximize the benefits of AI while minimizing its potential costs and risks.

Artificial Intelligence (AI) and Covid-19 are the two remarkable buzzwords in recent times. AI has seen unprecedented developed by availing technology to fight the Covid-19 crisis. AI has contributed to the virus detection and tracking, and mainly to vaccine production. With 2020 coming to an end, Florian Douetteau, CEO and co-founder of Dataiku talked to Analytics Insight on his predictions of AI policies for 2021. Dataiku provides enterprise AI tools for companies like Pfizer, GE and Unilever. Dataiku has seen exponential growth in 2020, raising US$100 million in August at a time when AI was increasingly becoming a business imperative.

AI policy is defined as public policies that maximize the benefits of AI while minimizing its potential costs and risks. AI holds great economic, social, medical, security and environmental promises. AI systems help people acquire new skills, design and deliver faster production times and quicker iteration cycles, reduce energy usage, provide real-time environmental monitoring of pollution and air quality, enhance cybersecurity defences, boost national output and create new kinds of enjoyable experiences. For all these reasons, researchers are thrilled with the potential use of AI systems to help manage some of the world's hardest problems and improve countless lives. But in order to realize the potential, the challenge associated with AI development has to be addressed. It highlights the importance of carrying an AI policy which makes the process simple and easy. Douetteau anticipates on areas where AI policy will see improvement in 2021.

Inclusive Engineering is making way into mainstream to support diversity

Douetteau thinks that in order to ensure diversity into the AI plan, companies must commit the time and resources to practice inclusive engineering. He opined, "Inclusive engineering isn't limited to an extent, but it means doing whatever it takes to collect and use diverse datasets. This will help companies to create an experience that welcomes more people to the field, looking at everything, from education to hiring practices.

Often in real-world tasks, there isn't enough data to take full advantage of deep learning. However, it is possible to leverage other datasets to reach a critical mass. Sharing knowledge across diverse datasets leads to more general knowledge, deeper insights and more well-informed decisions. This is especially true in domains like healthcare, where data for any particular task can be expensive or dangerous to collect.

Business skills to become more critical than ever for data scientists

According to Douetteau, data scientists will need to speak the language of business in order to translate data insight and predictive modeling into actionable insight for business impact. He said, "Technology owners will also have to simplify access to the technology so that technical and business owners can work together. The emphasis for data scientists will not just be on how quickly they can build things, but on how well they can collaborate with the rest of the business."

The current trend in data science and analytics makes it difficult for a normal business leader with no tech background to understand analytics flow. In order to address the issue, companies should ease the technology used by data scientists, and make it simple and accessible for everyone. Good communication and understanding between the analytics team and the improvement team of a company are important to increase the revenue.

Increase in organizational commitment to put humans and diversity at the centre of AI development

Douetteau foresees that companies will look to include people who are using algorithms if they truly want to reduce bias and foster diversity. He adds, "While most training datasets have been developed against a small percentage of the population, companies will now look to consider expanding their scope to design training datasets that are all-inclusive. The more inclusive the group building of AI and the datasets, the less the risk for bias."

Building an inclusive future with an unbiased AI is the major goal that companies keep in mind while developing solutions. With AI invading every sector including education, law, banking, etc., it is the need of the hour to address all kinds of people from different races, geographies and genders equally.

Organizations will add more robust processes and requirements around drift monitoring using MLOps

Douetteau strongly believes that the input drift is based on the principle that a model is only going to predict accurately if the data it was trained on is an accurate reflection of the real world. He mentioned, "So if a comparison of recent requests to a deployed model against the training data shows distinct differences, then there is a strong likelihood that the model performance is compromised. This is the basis of input drift monitoring. In 2020, there was a significant drift seen across all sectors because of the health crisis."

He concluded, "In 2021, we'll see organizations using Machine Learning Operations (MLOps) to put more robust processes and requirements around drift monitoring so that models can be more agile and accurate. Identifying drift is one of the most important components of an adaptable MLOps strategy, and one that can bring agility to the entire organizations' Enterprise AI efforts."

MLOps is the set of policies, practices, and governance that are put into place for managing machine learning and AI solutions throughout their lifecycle. It focuses on building a common set of practices which everyone involved can follow for systematically managing analytics initiatives.