Agile Deep Learning For Modern Software Development

Original
Deep Learning has been called the "new electricity", with its sudden transformational power over every industry. While the attention on deep learning's innovative algorithms is well-deserved, turning innovation into value requires integrating these algorithms into practical technology products. 

Such products are often developed following the principles of agile. David Murgatroyd is Machine Learning Leader at Spotify, where they are using AI to improve recommendations, search and conversational interaction, and empower their consumers' music experiences. He'll join us at the Deep Learning Summit in Boston to share how deep learning can be approached in an agile way, and successfully integrated into the cadence of a modern software development organisation.

I spoke to David ahead of the summit on 25-26 May to learn more about his work in deep learning, and what the future holds for tech at Spotify.

Can you tell us more about your work, with a short teaser for your session?

I give leadership to a great group of folks doing machine learning in Spotify’s Boston office with collaborators around the globe. This spans a variety of application areas: voice interaction, search, crowd-sourcing, knowledge graph enhancement, audio analysis, and personalization.

I’m excited to share what I’m learning about approaching deep learning in an agile way so that it brings as much value as possible as soon as possible.

What started your work in deep learning?

Having studied neural networks in the late 1990s, I first heard of their resurgence from Andrew Ng in 2010. Then when their application to symbolic domains like language took off with word2vec in 2013, those of us in Natural Language Processing began to get involved in earnest.

What are the key factors that have enabled recent advancements in NLP?

Being able to represent words with vectors that arise out of end-to-end training was the first big breakthrough for NLP. More recent work has been better able to capture the long-range dependencies of language through recurrent networks like BiLSTMs with attention.

What challenges do you think will be most interesting for researchers & scientists in the next few years?

For NLP, I’m excited to see how more of the recursive structure of language will manifest in neural architectures while also benefiting from innovations that arise in other application areas (e.g., GANs). More broadly, I think there’s a lot of valuable work to be done on interpretability and leveraging human expertise.

Which industries do you think will be shaped most by machine learning?

I believe we initially saw machine learning primarily benefit consumer-focused internet companies because their resources are large and their problems are often lower-level (e.g., more perception than planning) as well as end-to-end. Now the benefits are broadening to B2B companies who tend to have more intricate and compartmentalized problems. Where we really need more investment is nonprofits whose resources are smallest and whose problems are often most important and least studied.

In terms of verticals, healthcare, transportation, and energy seem to be industries where some combination of regulation and properties of the problems from ethics to cost-of-failure have rightly cause machine-learning’s adoption to progress more cautiously.

What do you see in the future for Spotify?

I love being at Spotify because I believe music truly improves lives and because machine learning plays a big role in improving the lives of creators and audiences by connecting them in ways that benefit both. I’m excited to help broaden and deepen this connection through new experiences and new ways of interacting that are ever more personalized and relevant.

David Murgatroyd will be speaking at the Deep Learning Summit in Boston on 25-26 May, taking place alongside the annual Deep Learning in Healthcare Summit

Confirmed speakers include Sangram Ganguly, Senior Research Scientist, NASA & BAER Institute; Andrew Tulloch, Research Engineer, Facebook; Sanja Fidler, Assistant Professor, University of Toronto; Charlie Tang, Research Scientist, Apple; and Dilip Krishnan, Research Scientist, Google. View more details here.

Tickets are now limited for this event. Book your place now.

Can't make it to Boston? Join us at Deep Learning Summits in London and Montreal. View all upcoming events here.

Opinions expressed in this interview may not represent the views of RE•WORK. As a result some opinions may even go against the views of RE•WORK but are posted in order to encourage debate and well-rounded knowledge sharing, and to allow alternate views to be presented to our community.

Original

Neural Networks Deep Learning Machine Learning AI Deep Learning Summit Natural Language Understanding Machine Intelligence NLP Deep Learning Algorithms


0 Comments

    As Featured In

    Original
    Original
    Original
    Original
    Original
    Original

    Partners & Attendees

    Intel.001
    Nvidia.001
    Graphcoreai.001
    Ibm watson health 3.001
    Facebook.001
    Acc1.001
    Rbc research.001
    Twentybn.001
    Forbes.001
    Maluuba 2017.001
    Mit tech review.001
    Kd nuggets.001