Published Oct 11, 2017

AI-Powered Conversational Interfaces with Paul Tepper - #52

Explore the forefront of AI-powered conversational interfaces with Paul Tepper from Nuance Communications, as he delves into innovative AI development, the challenges of emotion and sentiment recognition, and the critical balance of technology and ethics in user experience design.
Episode Highlights
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) logo

Popular Clips

Episode Highlights

  • Innovation

    Paul Tepper, worldwide head of cognitive innovation at Nuance Communications, shares insights into the company's strategic directions in AI development. He emphasizes the importance of identifying high-value problems and collaborating with the corporate research division to prototype solutions. This approach aims to move innovation forward and provide new opportunities for product management teams across the company 1. Paul also highlights his background in computational linguistics and his experience in building cloud-based NLP platforms, which informs his current work at Nuance 2.

    I lead a team that has a few functions. One of our main functions is identifying high value problems for which the company doesn't yet have a solution.

    ---

    His focus on dialogue and nonverbal behavior in computational linguistics plays a crucial role in shaping Nuance's AI initiatives.

       

    Privacy Challenges

    Data privacy is a significant challenge for AI companies like Nuance, as Paul Tepper explains. Balancing competitive AI development with customer data protection involves navigating complex legal issues, especially for clients like banks and government agencies 3. Paul notes that explicit consent is often required to use data, even for model tuning, highlighting the legal intricacies involved.

    It's not how do I scale a deep learning model, or how do I productionize this system... It's like the legal problems.

    ---

    Additionally, the customization of user interactions through machine learning models is complicated by privacy concerns, as many clients prefer not to share data or models, even for optimizing chatbot interactions 4.

Related Episodes