Outlook 2017: Adela Quinones and Mark Dimont, Bloomberg
Adela Quinones is news product manager at Bloomberg.
Mark Dimont is product manager, news and social media applications at Bloomberg.
What will 2017 be known as?
Adela Quinones: The new year will be known as “the year of machine learning and data-driven decision making as a core competency.” As the “Internet of Things” proliferates and there are too many data points for a team of analysts to derive insight from, machines will take over the core data aggregation and analysis tasks in a much more structured way. Companies will be able to adapt strategies much more quickly and become far more efficient as they leverage data analytics in everything they do.
Mark Dimont: It will be cyborg news, where human curators combine with machine learning tools and automated story generation to improve classification, personalization, volume management, and quality control of content.
What will be the next AI watershed for the industry in 2017?
Quinones: Neural networks will move from being a buzzword to a key tool in the toolbox of organizations looking to scale core AI “classification” problems. The barrier to entry for using neural networks was reduced when Google open sourced TensorFlow making it possible for organizations without dedicated machine learning teams to leverage basic machine learning. Companies will begin to throw data into the cloud at scale and use neural networks to paint a picture of where they should go next in terms of product development and issue resolution, among other things.
Dimont: AI as a service. Companies will make AI more extensible and essentially allow AI services to be generalized and integrated with their applications. For example, perhaps one day Amazon’s voice recognition will combine with Google’s translation to power your LG refrigerator shopping list scanner.
What changes should the industry see in regards to machine learning and AI in 2017?
Quinones: I think we’ll see more of a realization that good clean data from “expert sources” is critical to machine learning. Even the most advanced and sophisticated algorithms cannot derive much value if they aren’t given high quality, curated data sets to derive meaning from. Experts are needed to make sense of model results and provide feedback where models fail to paint a full picture. Companies will move analysts from data gathering and computation tasks, to model training and insight generation.
Dimont: Hacking is real and news- and social media- manipulation is easier, more frequent and more pervasive than most people ever believed.
Which hot topics/hype should be retired at the end of 2016?
Dimont: Big Data- the ability to take in, analyze, and derive insights from large datasets- is now table stakes for any organizations. It was much hyped in the past, but now just comes with the territory.
What changes do you expect to see in regards to machine learning and AI in 2017?
Dimont: In financial services particularly, there is a massive focus on finding structure in unstructured data sets. The alpha is gone from most traditional datasets so I expect to see a tremendous amount of creativity in 2017 in how machine learning is used to derive new and interesting financial insights.
Vendor taps the Haskell language to power its new platform.
Machine learning will re-define signal detection.
Machine learning and AI need to go from theory to application.
Machine learning excels at identifying trading opportunities, but it could be wasted if it cannot inform the trader.
Regulatory requirements and machine learning have the investment research market re-inventing itself.