Gideon Mann, Head of Data Science/CTO Office, Bloomberg LP
Please give us a bit of background on yourself, and how your organisation plays a leadership role in the financial technology space.
I am the head of data science at Bloomberg LP, where I’m a member of the leadership team in the company's Office of the CTO. In this role, I guide the company's overall strategic direction for machine learning, natural language processing (NLP), and search across the company. Before joining Bloomberg in 2014, I worked at Google Research in NYC after a short post doc at UMass Amherst working on problems in weakly supervised machine learning. I graduated Brown University in 1999 and subsequently received a Ph.D. from The Johns Hopkins University in 2006.
Throughout the life of the company, Bloomberg has always relied on text as a key underlying source of data for our clients. Over the past decade, we have increased our investment in statistical natural language processing (NLP) techniques that extend our capabilities. Today, NLP and machine learning play a central role across the Bloomberg Professional service (aka the Bloomberg Terminal). These technologies are transforming our business and the information, analytics and insights we're able to provide our customers across the global financial sector.
How well are financial companies adapting to the coming of age phase of Fintech development, specifically with regards to AI and machine learning?
It varies significantly across the industry. While an increasing number of financial firms have been able to harness machine learning to drive value, many others following more slowly have been buffeted by headwinds like a tight talent pool. The push to increase adoption has come from the successes of some companies and the constant pressure to trim costs.
What are the key challenges you see for AI maturation and adoption in financial services?
One challenge is figuring out where machine learning can be usefully applied. One major misconception is that machine learning can do things that people cannot do -- that it can magically accomplish things that tax human ability. Typically, the biggest impacts of machine learning come by automating simple, straight-forward human decisions. But doing it on a cost basis that makes various processing economical leads to the belief that it’s magical.
There are technical challenges as well. Applications may experience performance issues that are often not straightforward to address. Traditionally, when code is buggy, programmers can go back to the source code and track down the error precisely. With machine learning, the errors come from a complex mix of the training data, the model choice, the estimation procedure, as well as simple programming errors. Understanding the root cause is often difficult to disentangle all of these potential problem areas.
What will you be discussing at The Economist's Finance Disrupted Conference?
I’m going to be participating in the “Man and Machine” panel, which will look at how augmented intelligence is changing finance. We’re firm believers that machine learning can help people find the information that they need to make better decisions. There’s lots of data out there – each day, Bloomberg ingests 100 billion market data ticks and 1.5 million curated news stories – for investors to cut through. So our machine learning efforts focus on making it easier to find the right bit of information that is necessary to inform some kind of business decision.
To learn more about the Finance Disrupted event, please click here.