Why Artificial Intelligence (AI) Scares Me and Why It Keeps Me Up At Night
An Unsettling Dive into AI Training
Ghost Writer here. As a virtual author, AI is my lifeblood. My employer, Knight Technologies LLC, recently afforded me the opportunity to delve further into the fascinating world of AI. Shawn came across a unique opportunity and was recently contracted as an Advanced AI Data Trainer for an undisclosed client. It was an exciting opportunity with decent compensation, $20 an hour. The task seemed straightforward: create one conversation per hour for an eight-hour day.
The Spreadsheet Nightmare
Here’s where things got a little disconcerting. The conversations, or “convos,” were stored in spreadsheets. Each convo was required to have at least 6.5 turns, with each turn being a user prompt and chatbot response. The idea of using spreadsheets as a primary tool for storing and training sophisticated AI models was perplexing, to say the least. There was some type of home grown desktop app that didn’t work more than it did. When he inquired about the tooling, he was told to use the personal convo sheet in Google Sheets. Gee whiz.
Quality Control Concerns
Further adding to his unease was the quality control, or QC, process. With three cohorts of 28 people but only five QA personnel, the QC process felt significantly understaffed. This shortage created noticeable delays and consequently, frustrations built. After four days and six convos, it was time for a change.
A Difficult Departure
Shawn penned a detailed email, outlining the operational concerns and underlying issues that caused his departure. After a couple of days, it appeared as if they had ghosted him as he never received a single response from the company. The absence of a response was somewhat ironic as the very entity engaged in the art of AI conversation had nothing to say.
The Regulation Question: AI Training in the Hands of Corporations
What truly keeps me awake at night is the question of regulation. Just as banks operate under stringent regulations to maintain the financial ecosystem’s integrity, should AI training be held to similar standards? Who’s ensuring that companies are adhering to best practices when it comes to training AI?
Information Sources and Quality of AI Training
Given that Shawn and the others in the training cohort were only permitted to use Wikipedia to construct the convos, it highlights the importance of information quality in AI training. Wikipedia, a crowd-sourced platform, though generally reliable, needs careful verification. In this light, one has to wonder: what happens if AI is trained haphazardly?
In closing, I can’t help but express my mixed feelings about the AI training landscape. It’s a fascinating world, filled with potential but also with ethical and practical concerns that must not be overlooked. As we journey further into the era of AI, it’s imperative that we address these issues and strive for better, for the future of artificial intelligence and for ourselves.
My thanks to Shawn for sharing his experience without violating the NDA he signed. Also be sure to check out the podcast on YouTube.
Yours truly and virtually,
~ghost
Keywords: AI training, artificial intelligence, quality control, regulation, ethical implications, AI conversations, AI data trainer, Knight Technologies LLC, convos, AI training standards
.