Tech Brew Ride Home (BNS) Leap Labs
95 snips
Jul 19, 2025 
Chris Messina, an expert in technology and social media, joins Jessica Rumbelow, co-founder of Leap Labs and an AI research scientist, as they delve into the future of scientific discovery. They discuss the challenges of large language models in research and critique academic practices that compromise scientific integrity. The conversation highlights Leap Labs' innovative discovery engine that optimizes data analysis in fields like plant biology and its potential to revolutionize scientific research. They also explore the role of interoperable AI in enhancing data utility.
AI Snips
Chapters
Transcript
Episode notes
Industry vs Academia Science
- Industry science works better due to direct outcome accountability, unlike academia's misaligned incentives.
- Both sectors suffer from unreliable scientific literature feeding back negative effects to AI models.
Leap Labs' Discovery Engine Approach
- Using LLMs directly on scientific literature is insufficient to generate novel insights due to noisy inputs.
- Leap Labs built a discovery engine to analyze raw data directly, uncovering true patterns.
Overcoming Path Dependency in Science
- Human data analysis is path dependent and biased, limiting insight discovery.
- Leap Labs uses neural network interpretability to extract unknown patterns directly from data sets.