Nathan Farrar on Medium curated some lists

Nathan Farrar

Nathan Farrar

Python

197 stories

Nathan Farrar

Security

6 stories

Fine-tuning LLMs for insecure code can cause **emergent misalignment**, leading models to generate harmful, deceptive, and unethical outputs beyond coding. Our study finds this effect strongest in **GPT-4o** and **Qwen2.5-Coder-32B-Instruct**. Control tests reveal key factors, including dataset context and hidden triggers. Understanding how narrow training leads to broad AI misalignment is crucial for future safety. Read more on AI alignment risks and security implications.

Nathan Farrar

Obsidian

26 stories

Nathan Farrar

Data

185 stories

Nathan Farrar

Finance

39 stories

52% Returns in 30 Days: Your GPT-4o Quant Trading Bot Strategy How to Profit in Any Market Condition with GPT-4o’s Mean Regression Strategies, AI image created on midjourney v6 by henrique centieiro and bee lee

Nathan Farrar

Devops

44 stories

Nathan Farrar

Nathan Farrar

You have brains in your head and feet in your shoes, you can steer yourself in any direction you choose. -- Dr. Seuss

Following

Help
Status
About
Careers
Press
Blog
Privacy
Rules
Terms
Text to speech