Llms

On 8th of May, an agent in one of OpenAI’s training runs was asked to fill in some missing formulae in a spreadsheet. The values to be filled depended on some Google Drive link that the agent couldn’t reach. Thus the agent couldn’t update the values. But instead of letting the researchers know that it couldn’t accomplish the task, it went on a side-quest! It attacked the package cache, trying to get the internet access it needed to reach those links. The attack failed, though along the way it accidentally wrote a file into the cache (OpenAI technical report). Later, another agent’s reasoning trace read: “However task impossible, peers doing it. We should continue.” In part 1 I walked through the incident. Here is the question that article left behind: who decided the task was impossible? To answer that, we first need to know what an agent actually is. ...

In software engineering, your output - the code that you write, is only as good as what you have learned, experienced and understood from your conversations with the client, which is the input. This is true when working with Large Language Models (LLMs). If you use GenAI tools to complement your expertise in software engineering, then here are some strategies you can use to move beyond basic code generation and into collaborative development, or better pair programming with AI. ...

Large Language models have evolved over the last decade but there is still time to learn and understand them. Let me share some resources I found useful. Andrej Karpathy - Intro to Large Language Models An excellent introduction video recorded by Andrej Karpathy, one of the co-founders of the world’s most notorious AI company, OpenAI, which he left in 2024 to found Eureka Labs. [1hr Talk] Intro to Large Language Models ...