Artificial Intelligence (AI) and Data Visualization can seem like an unlikely marriage. AI techniques often work as a black box: we cannot know how the AI has reached its conclusion. This can raise uncomfortable questions: think of a medical diagnosis, or the screening of job applicants: if we cannot see inside the black box, we can’t know whether the AI made a serious mistake, or reflected our implicit bias. When the AI becomes a veil between us and the data it makes us uncomfortable and it takes our own intuition and insight out of the game.
But AI can help us see the data, becoming a crucial help to our own analysis and judgement.
Leo Meyerovich, CEO of Graphistry, discusses in this podcast interview a number of areas where AI-driven data visualization greatly facilitates the work of human analysts: from fraud prevention to health care, supply chain management, customer analysis, all the way to fighting human trafficking and spotting election-influencing tactics.
Across all these applications, says Leo Meyerovich, “The dream is some sort of a black box […but] I found zero systems that are fully automated.” There is always a human in the loop. Sometimes it’s to spot where the AI stumbles on a ‘false positive’, like a legitimate transaction flagged as fraudulent. Sometimes the analyst’s experience is invaluable to accelerate the process: the human knows what to look for and has a better sense of the context than the AI.
In all these fields, AI provides invaluable help. Human intuition, experience and decision-making continue to play the central role across economic activities. But now we have realized that data can help us make better decisions, and we have learnt to harvest and store prodigious quantities of data—so large that we struggle to make sense of them on our own.
Especially with the rise of complex global supply chains, most companies today are exposed to a very large number of evolving factors, from commodity prices to transportation costs to economic developments across the world (think of the recent impact of China’s growth slowdown). How can you capture the interrelated impact of all these factors on your business?
A picture is worth a thousand words, but faced with a flood of data, humans need help building the right picture. This will most likely not be as simple as a bar chart–it might be a diamond, an evolving set of color-coded clouds and cluster, or filaments and waves showing the spread of malware . Look at the demos of Graphistry or other AI/Data Visualization companies, and you will see how AI can make data representation more creative, elegant and, most importantly, intuitive. AI can now take gigantic datasets and show us patters and correlations that help our own intelligence do its best job. Just as we have to do the hard work of organizing and labeling the data in a way that the AI can usefully process, so the AI can then devise the best designed input for our own processing power. It’s a great example of human-machine partnerships.
In this context, AI can also help capture and transfer learning within organizations: once an experienced analyst, together with the AI, has devised an effective set of steps and visualizations to analyze a problem, the AI can remember that and build it into a set of tools ready for a junior analyst coming on board. This could prove especially helpful to organizations that experience very high staff turnover—and to industries that face a workforce aging problem, with large experienced cohorts retiring and a smaller cohort in the pipeline.
The more complex the business, the greater the number of moving parts and variables, the greater value we can get from an AI that helps us see through the fog of data.