Night Science

About Night Science

Where do ideas come from? In each episode, scientists Itai Yanai and Martin Lercher explore science's creative side with a leading colleague. New episodes come out every second Monday.

Publisher

Itai Yanai & Martin Lercher

Category

science · science · science

Language

en

Explicit

No

First episode

20 Apr 2021

Latest episode

10 Aug 2026

Latest episodes

2 of 150 episodes have a transcript.

1. 

   10 Aug 2026TRANSCRIPT

   88 | How AI dreams up new science – with Jure Leskovec

   What happens when AI starts doing science? We talk with Stanford computer scientist Jure Leskovec about AI agents that find patterns in large datasets, come up with hypotheses and try to kill them with more data, enduring the tedious work human scientists would rather avoid. What are we human scientists still uniquely good for? Taste, judgment, intuition, emotion, procrastination? And we discuss whether AI might be opening up a whole new continent of discoveries, a terra incognita with plenty...

2. 

   13 Jul 2026

   87 | On the joys of being a young scientist

   What can early-career scientists teach us about creativity? For this special episode, we talked to PhD students and postdocs about their creative processes. They told us about finding scientific ideas in art museums, sketching on whiteboards, borrowing concepts from other disciplines, going to unrelated seminars, swimming, hiking, talking in pubs, and deliberately letting ideas lie fallow like in a crop rotation system. Through this diversity, a few themes kept coming back: openness, play, vi...

3. 

   22 Jun 2026

   86 | From punchlines to discovery – Sarah Adelman

   Host [Host Name] interviews comedian and former research scientist Sarah Adelman about how comedy's creative process mirrors scientific discovery, offering insights into embracing failure and playful exploration.

Inside a recent episode

88 | How AI dreams up new science – with Jure Leskovec

Published 10 Aug 2026 · Transcript excerpt

[…] So, you know, hypothesis are just theories that have not yet been proven and theories that are not obviously wrong. So now you have the closed loop AI system. AI, if I joke a bit, right, if it's great at something, it's great at hallucination, can hallucinate gazillions of hypotheses. There's no shortage of hypothesis. But what you can do then is you can now have an agentic loop that takes this hypothesis and is testing them. It tries to falsify this hypothesis against the data that you currently have. So now you are basically saying, what kind of theories, what kind of hypothesis, and I can make a lot of hypotheses, are actually consistent with my data to the degree that I cannot disprove, given the data I have right now. […]

Who has been a guest on Night Science

Names identified in recent episode analyses. Showing up to five guests.

• Sarah Adelman
• Eve Marder
• Peter Godfrey-Smith

Reach and audience

Public platform figures. Ratings count people who left a rating, not total listeners.

YouTube views

5,106

Score snapshot 15 Sept 2025

Apple Podcasts (US)

5.0 / 5

66 ratings

Spotify

4.9 / 5

171 ratings

Podcast Authority Score: 72 / 100

A composite of feed quality, social presence, YouTube performance and engagement. Read the methodology.

Quality

75

Social presence

0

YouTube

74

Engagement

99

Shows with a similar audience

• 

  Nature Podcast

• 

  Science Magazine Podcast

• 

  Quanta Science Podcast

• 

  The Joy of Why

• 

  CrowdScience

• 

  The Science of Everything Podcast