Week 51 of 2023- W512023
All the open browser tabs I get to... Drop 56,875,789 is a go in 3,2,1...
Disclaimer: The views and opinions expressed in this blog are entirely my own and do not necessarily reflect the views of my current or any previous employer. This blog may also contain links to other websites or resources. I am not responsible for the content on those external sites or any changes that may occur after the publication of my posts.
End Disclaimer
“Vertigo is the conflict between the fear of falling and the desire to fall.” -Salman Rushdie
“It's Easier to Ask Forgiveness Than It Is To Get Permission.”- Rear Admiral Grace Hopper
News
25 Minutes with Cap: Using AI in Insurance with Otakar Hubschmann
AIML
OpenAI Prompt Engineering Guide
Is AI leading to a reproducibility crisis in science?
Teaching AVs the Language of Human Driving Behavior with Trajeglish
Painting
The Magpie, Claude Monet, Year 1868–1869, Musée d'Orsay, Paris
Crypto
The resurgence of crypto, Tyler Cowen
Cyber
Markets
Portfolio Strategy: Why I am no longer a quant [2020]
VIX is at its lowest level since the pandemic
Climbing the Maturity Wall of Worry
An Example of Reinsurance Returns vs other Asset Classes
Electric scooter company Bird files for bankruptcy
Why One Firm's 3,612% Return Is Drawing the Ire of Hedge Funds
Two Sigma: Systematic and Discretionary Macro: Better Together
Misc
Flaco the Famous New York City Owl Has Become a Peeping Tom
Seven Predictions for the World of Entertainment in 2024
“You Better Be as Good as Letterman!” Conan O’Brien’s Wild First Year: An Oral History
The Return of White Sharks to Cape Cod [2021]
Luiz Barroso, Who Supercharged Google’s Reach, Dies at 59
It Sure Looks Like Phones Are Making Students Dumber
He Was the NFL’s Dead Last Pick. Is He Now the MVP?
Papers
Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models
The paper introduces Reinforced Self-Training (ReST), a method for enhancing language models' problem-solving abilities. ReST iteratively generates solutions, evaluates them with a binary reward, and fine-tunes the model based on these evaluations. It significantly outperforms traditional training, especially for complex tasks like math problems and code generation, with a caution against potential overfitting. This approach suggests reduced reliance on human data for training advanced models.
Autonomous chemical research with large language models
The paper presents "Coscientist," an AI system using GPT-4, capable of autonomously conducting chemical research. It effectively designs, plans, and executes experiments, particularly in reaction optimization of palladium-catalyzed cross-couplings. Coscientist combines large language models with search, coding, and automation features, demonstrating versatility and explainability in scientific research applications.
Movie
The Creator - I’m a fan of Gareth Edwards. Rogue One anyone? The Creator is a beautiful looking movie with some pacing issues. Interesting take on AI.
Podcast Episode
Jeff Bezos: Amazon and Blue Origin | Lex Fridman Podcast #405