A People-First AI Strategy
Here are five important AI stories from the week.
To have the most success implementing AI in enterprise, people must feel empowered by AI, not threatened. People are the most important asset at most companies, and AI should support them. If AI remains a blackbox and is viewed as a substitute of humans rather than a complement to them, there will be substantial resistance to AI, limiting its adoption. We need a human-centric AI strategy to improve AI’s adoption rate.
A 94-second video on how artificial intelligence and machine learning drive operations at Amazon, enabling services such as one-day Prime delivery. These automation and optimization techniques are ubiquitous at the larger tech-enabled firms and will become a mainstay in corporate America in the next few years.
Twitter acquires one-year-old AI startup Fabula AI to fight fake news. Fabula specializes in graph deep learning, which is a relatively new method to find relations and interactions in large and complex datasets. For Twitter, the ability to analyze how various Tweets, Retweets, Likes, and Twitter users are related to each other and how the interactions evolve over time is crucial. For more on how Fabula AI spots fake news, read this TechCrunch piece from earlier this year.
In a recent study, Facebook discovered that object recognition of common household items from low-income, typically non-Western countries, was very poor. This is another example of how machine learning models trained on large datasets of images from Western nations such as the United States have poor performance when applied to a global context. To reduce the bias of gender, race, cultural background, country of origin, and other socio-economic factors, Facebook is working hard to source more diverse, more globally representative datasets.
Amazon, Google, and Microsoft are engaged in a fierce race to push out machine learning as a service to enterprise clients. Recently, Microsoft upgraded its forecasting service with automated machine learning (AutoML). New features include a new forecast function, rolling-origin cross validation, and automated feature engineering using lags, rolling windows, and holidays. The lives of data scientists just became considerably easier. For additional resources, please visit the how-to guide.
More Stories Worth Reading and Watching…