In this episode, Peter Tu, Chief Scientist of Computer Vision at GE Research, discusses the advances in real-time cloud computing for computer vision in manufacturing. He explores the challenges of incorporating subjective questions into computer vision systems and the process of implementing these systems. Tu also highlights the importance of understanding the environment and ontology of objects and actions when adopting AI in manufacturing. Finally, he explores the role of computer vision in predictive inventory and teases upcoming episodes in the Beyond GPU series.
Read more
AI Summary
AI Chapters
Episode notes
auto_awesome
Podcast summary created with Snipd AI
Quick takeaways
Data annotation and complexity pose challenges for computer vision in manufacturing.
Integration of subject matter experts and understanding data structures are crucial for AI adoption in manufacturing.
Deep dives
Challenges in Computer Vision for Manufacturing
One of the main challenges in computer vision for manufacturing is dealing with the complexity and judgment-based nature of the tasks. While some manufacturing decisions can be straightforward, others require more subjective judgment. This poses a challenge in defining and annotating the data with clarity. Another challenge is the increasing amount of data required to support these tasks, as models are getting bigger and demand larger datasets. Although few-shot learning approaches are emerging, the reliance on large datasets and highly skilled annotations still persists.
Integrating Subject Matter Experts into AI Adoption
The integration of subject matter experts into the AI adoption process is crucial for handling the subjective or artistic questions in manufacturing. Skilled practitioners are needed to articulate the desired outcomes and requirements in a way that can be transformed into numerical models. This process requires expertise in articulating the desired objectives and understanding the structures and dependencies of the data. Democratization has made AI more accessible, but there is a need for introspection and a deeper understanding of the data and its underlying structures.
Advancements in Computer Vision and the Theory of Data
Advancements in computer vision are moving beyond specific tasks towards a more conceptual understanding of the environment. The goal is to develop situational awareness and comprehend the semantic aspects of the data. This involves understanding objects, actions, attributes, and affordances. The challenge lies in grounding these concepts and projecting them into salient dimensions. Developing a theory of the data itself is crucial for improving the performance and utility of computer vision systems in various domains, including manufacturing.
Today’s episode is the third in a special series we’re calling ‘Beyond GPU,’ taking a look at edge AI computing challenges and solutions with help from guests at leading vendors and superscale global tech brands leading the most advanced hardware platform teams on the planet. Today’s guest is Peter Tu, Chief Scientist of Computer Vision at GE Research. Peter returns to the ‘AI in Business’ podcast to talk about the advances in real-time cloud computing that are making computer vision more accessible across manufacturing, and how executives can take advantage of these advances with the proper preparations in infrastructure and systems. This episode is sponsored by Rain. Learn how brands work with Emerj and other Emerj Media options at emerj.com/ad1.
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode