AI for the Grader - Verifiability in Under 10 Minutes

Описание: How do we know there isn’t a porcelain teapot orbiting the Sun between Earth and Mars? Some claims are easily verifiable; others are not. Differences in how easy it is to verify AI’s outputs may shape how quickly AI advances across different domains.

Advances in AI have been uneven, a phenomenon known as the jagged edge of AI. For tasks that fall outside the current AI capability frontier, using AI can actually reduce human accuracy. Researchers have also identified different modes of human–AI collaboration, which they call the centaur and cyborg models.

📖 Chapters:
0:40: Verifiable sales claims
1:07: Russell’s Teapot
1:50: Check is in the mail
2:00: Healthcare examples
2:14: Audio Transcription - Easy to generate/easy to verify
2:45: Diagnosis - Easy to generate/hard to verify
3:24: Nurse Scheduling - Hard to generate/easy to verify
3:55: Drug Discovery - Hard to generate/hard to verify
4:13: Software 1.0/2.0/3.0 - Andrej Karpathy
5:26: Tasks easy to verify will be automated first
5:53: Verifier’s Rule by Jason Wei
6:16: Jagged Edge of AI
7:22: Centaurs and Cyborgs
7:41: Comment on current state - verifiable task outside the frontier
8:20: Takeaways

💡 References
Andrej Karpathy:
   • Andrej Karpathy: Software Is Changing (Again)  
https://x.com/karpathy
https://karpathy.ai/

Jason Wei and Verifier’s Rule:
https://www.jasonwei.net/blog/asymmet...
https://www.jasonwei.net/

Harvard Working Paper on the Jagged Edge of AI:
https://www.hbs.edu/ris/Publication%2...

👩‍🔬Your Instructor:

Sonali Tamhankar, PhD, is an experienced senior clinical data scientist and an award-winning teacher. She was honored by the Blockchain Magazine as a Leader in AI and Data Science 2025, and by Data Science Salon as an outstanding woman in AI and ML.

She was selected as a Leadership Fellow at AIM-AHEAD, a consortium at the intersection of AI and Health Equity, has served on non-profit boards, and is a fellow at Transform VC, a venture capital firm combining outlier returns with outlier impact.

Before transitioning to healthcare over 10 years ago, Sonali was a tenure-track professor. Sonali has a PhD in Physics and Masters in Mathematics, secured 6th rank at the Indian National Mathematics Olympiad, won the silver medal in Physics at Indian Institute of Technology, and has been recognized as an outstanding instructor with a Teaching Excellence Recognition Award and the Joseph and Sophia Konopinski Memorial Teaching Award.

Here’s a comment from a colleague:

“Sonali is at once a scientist and a teacher. She mentors and facilitates new learning for the people she works with on projects. She is current in the state of the science methods and brings clarity to the collaborative teams with whom she works.”

Let’s talk!
www.linkedin.com/in/sonali-tamhankar
   / @sonali_tamhankar_genai  

#AI #LearnOnYouTube #bertrandrussell #verifiability #Karpathy #jaggededge #aifrontier #Technology #HealthAI #Innovation #MachineLearning #ResponsibleAI #Tech #WomenInSTEM #AIEducation #AISafety