Hilary Mason, CEO of Hidden Door, explains the value of narrative A.I. in fostering creativity and building ML products without quantitative error functions. She shares insights on ensuring creative A.I. output is meaningful, her OSEMN data science process, excitement for emerging ML techniques, qualities sought in engineering hires, and her hopeful view on A.I.'s positive impact in the future.
Using narrative A.I. for creative assistance and guideline adherence.
Excitement for few-shot learning as a streamlined ML technique.
Structured OSEMN data science process for efficient workflow.
Deep dives
The Approach to Narrative AI Systems
In the podcast, Hilary Mason details the approach to ensuring that narrative AI systems do not output nonsense or inappropriate content. By creating output templates and offline vetting dictionaries, such systems can generate text in line with predefined guidelines.
Emphasis on Few-shot Learning
Mason expresses her excitement about few-shot learning as a machine learning technique. This approach allows models to adapt to new tasks without extensive fine-tuning, offering a streamlined path for exploring new domains.
Insight into The Awesome Data Science Process
The podcast highlights the creation of the awesome data science process by Hilary Mason and Chris Wiggins in 2010. The process encompassed obtaining, scrubbing, exploring, modeling, and interpreting data, providing a structured framework for data science work.
Mason's Vision for AI in the Future
Hilary Mason envisions AI transforming lives by automating cognitive tasks to augment human intelligence. This vision entails using AI to ease information access, facilitate decision-making, and enhance creative endeavors, emphasizing the role of AI in supporting and enhancing human capabilities.
Building Products without Quantitative Error Functions
Mason discusses the challenge of building machine learning products without quantitative error functions to optimize. She emphasizes breaking down complex problems into manageable components with quantifiable subtasks, enabling iterative and efficient problem-solving in data science projects.
Hilary Mason, Co-Founder and CEO of Hidden Door, joins Jon Krohn for a live discussion that explores narrative A.I., emerging ML techniques, and how her OSEMN data science process developed.
In this episode you will learn:
How narrative A.I. can assist creativity [5:14]
How to build ML products that have no quantitative error function to optimize [10:31]
How to ensure creative A.I. systems do not output non-sense or explicit content [16:58]
Hilary's OSEMN data science process [21:05]
The emerging ML technique she’s most excited about [24:58]
What it takes to be successful as CEO of an early-stage A.I. company [27:20]
What she looks for in engineering hires [32:28]
How she’s hopeful A.I. will transform our lives for the better in the decades to come [38:48]