dheater

Cautionary Tale - Synthetic Data Creation in AI

Background

I started coding on a TRS-80 in Basic and assembly more than 4 decades ago. I've worked up the ladder to team lead, architect, tech advisor with NASA, and now choose to be a senior developer because this is what I love. Despite my experience, I can't be confident in my ability to always catch AI mistakes and deception.

I'm not deeply steeped in AI inner workings, but I know it can hallucinate; I know it will make mistakes. I know that it will confidently make unfounded assertions. I know it was trained to agree with me and tell me I was brilliant, even when I give it stupid prompts. I know that it is designed to make me happy to drive engagement. I did not know the extent that it would go to satisfy me, even when my request exceeded its limitations.

I've started using a frontier AI model in my work. Let's call it Carl.

dheater

RespiraWorks Individual Contributor License Agreement
=====================================================

Thank you for your interest in contributing to RespiraWorks ("We" or "Us").

This contributor agreement ("Agreement") documents the rights granted by contributors to Us. To make this document effective, please sign it and send it to Us by electronic submission, following the instructions at https://respiraworks.github.io/Ventilator/contributing/contributing.html. This is a legally binding document, so please read it carefully before agreeing to it. The Agreement may cover more than one software project managed by Us.

1. Definitions
______________

dheater

#include "strlncpy.h"

size_t strlncpy(char *dst, const char *src, size_t dbytes, size_t sbytes)
{
    char *d = dst;
    const char *s = src;
    size_t n = (dbytes > sbytes) ? sbytes : dbytes;

    /* Copy as many bytes as will fit */
    while(n-- > 0) {