At the 2018 AI Summit in New York City, we heard from a panel of executives representing multiple organizations across the AI ecosystem, including our own Mark Brayan, CEO of Appen. The panelists shared their perspectives on the usage of AI in different organizations, issues that need to be addressed, and what the future holds.
Perspectives on AI Applications
Clayton Ching, Global Head of Product Management at DRYiCE, discussed how his organization is deploying AI in IT Operations, where the goal is to move toward a self-learning platform that ultimately removes humans from the decision-making process. He admitted that this process will need time to evolve and gain wide acceptance in the company, as humans inherently do not trust machines to make business decisions. The organization will need to demonstrate incremental benefits over time to prove out the model and build employee trust.
Mohammed Ansari, Senior Vice President and General Manager of LG Silicon Valley Lab, described how AI is being used in both the B2C and B2B segments of its business to improve the user experience with LG products. AI is used to drive real-time learning about customer interactions so that user experiences can be improved to the point where they are what he called “immersive.” While AI technology is being used by LG to enhance these experiences, Ansari stressed the importance of human involvement in determining the right fit for each product by market.
Catherine Havasi, Chief Strategy Officer at Luminoso, discussed how we are really just scratching the surface when it comes to AI’s potential. “Immersive experiences” need to be supported by a different type of infrastructure, which is just starting to be explored. While low-hanging-fruit opportunities to apply AI have been implemented such as chatbots, we need to start addressing the “higher-hanging fruit” opportunities such as immersive customer experiences.
The panel offered important advice for organizations that are considering AI initiatives. Havasi stressed several important points that should be considered by any company planning to implement AI. First, identify the business problems that the organization wants to address with AI and establish how AI will solve them. Next, determine who owns the initiative. Havasi suggests that whoever owns the KPI for the business problem that the AI initiative is intended to solve should own the program. Further, to ensure that the AI initiative is properly deployed rather than becoming a “science fair project,” she recommends establishing a plan at the outset on how to operationalize it. Ask yourself, “Does our organization have the resources to support the initiative in production?” Finally, she stated that since AI relies so heavily on data, all organizations looking to deploy it need a solid data strategy.
Appen CEO Mark Brayan expanded on this point by explaining that training datasets for machine learning need to include as many edge use cases as possible. For example, there are many available images of roads, cars, and trucks to train an autonomous vehicle, but what about the edge case of a washing machine falling off a truck in the middle of the road? If the autonomous vehicle isn’t trained with this type of image, it won’t know how to react to it.
Dave Parsin, VP of North America at Artificial Solutions, explained that firms should establish a center of excellence for AI that drives toward key outcomes and is owned by the line of business owners. Further, he believes that the Chief Data Officer in an organization must be committed to ethical and responsible uses for data. This point of view was shared by Sana Khareghani from the UK Government Office of AI, who stressed that security and ethics should be built into machine learning algorithms from the start.
The panel moderator, Rich Karlgaard from Forbes, asked all of the panelists to provide closing comments on both their positive perspectives when it comes to AI, as well as their concerns. Havasi spoke optimistically about the capability of existing AI tools to make an impact on the way our day-to-day work is done. From the automation of customer service to advertising technology, she is excited about the developments we’re seeing in the market today. Brayan discussed that although there has been a lot of discussion about how AI will eliminate jobs, the opposite is true today. For example, Appen currently pays over 40,000 flexible contractors each month to work on projects that provide the high-quality data needed to enable a wide range of machine learning-based solutions. This is a tangible economic opportunity for thousands of people around the world.
Parsin predicts that the application of AI towards immersive, intelligent, and personalized experiences will become increasingly pervasive. And Ching believes that his grandchildren won’t have driver’s licenses because “they won’t need to.”
With all of this exciting development, there are still issues that need to be addressed. Khareghani explained that key segments of the population such as women and minorities are underrepresented in machine learning algorithms as they are typically not involved in creating them. She stressed the importance of including these groups moving forward if AI is to be truly representative and free of bias.
Once you’ve determined the business problem you want AI to solve, you need to build a comprehensive data strategy to build effective machine learning algorithms. You can ensure demographic diversity and remove internal bias by working with a curated crowd to collect and annotate your data. Contact us today to learn more about how we’ve helped 8 of of the top 10 global technology companies — as well as automotive companies, financial services firms, global governments, and more — build a strong data pipeline to fuel their machine learning programs.
Contact us to learn more about human-annotated data solutions for machine learning and artificial intelligence.