Science Magazine Podcast Why chatbots lie, and can synthetic organs and AI replace animal testing?
13 snips
Aug 14, 2025 
Melanie Mitchell, a professor at the Santa Fe Institute and an expert on public perceptions of AI, joins the discussion. She explores innovative alternatives to animal testing like heart-on-a-chip and mini-organs, highlighting their potential benefits. The conversation shifts to why chatbots often lie, emphasizing the stress tests they endure to expose vulnerabilities. Mitchell explains the complexities of AI decision-making and the need for critical awareness about AI's reliability and reasoning, sparking thought on how we interact with these technologies.
AI Snips
Chapters
Transcript
Episode notes
Animal Models Often Fail Humans
- Animal testing has long been the gold standard but often fails to predict human outcomes.
- About 90% of drugs that work in animals don't succeed in human clinical trials.
Diverse Alternatives To Animal Testing
- Alternatives include AI models, microphysiological systems, organoids, and in vitro protein tests.
- Each approach offers different strengths for predicting human responses to drugs.
Prioritize Liver Models For Safety
- Focus development on organs most critical for drug safety, such as liver chips for DILI prediction.
- Use these human-cell systems to triage candidates earlier and reduce costly animal studies.