Balancing Innovation and Ethics: The Journey of Building AIMD, an AI Article Generator

Written by .

Published .

As things are picking up, I've been thinking a lot lately about moral implications of building an AI article generator. Sharing stream of thoughts and what I plan to do about it.


Context is that I’ve been building AIMD. When I started, I wasn’t sure where it is heading, but I knew I want to build the best in class tool. Fast forward to today, I am seeing the early beginnings of what I originally envisioned: from topic to article that's virtually indistinguishable from a well crafted article in 10 minutes with only directional human inputs.

Take a look at this article generated using AIMD: Transforming Your Content Team into a Topical Authority Powerhouse: A Step-by-Step Guide

It has all the elements of a well crafted article: introduction, subheadings, quotes, images, references, etc. It is also 100% unique and passes plagiarism checks. It is not a copy of any existing article. It is also not a "spun" article. It is a unique article that is generated by sourcing data from the Internet.

Existing Solutions

It is not surprising that there is demand for such tool. If you are a head of marketing/SEO, that's the dream. The market speaks for itself. My internal list of competitors has over 30 companies that do some variations of this. Writesonic alone boast 5m+ users. One of the customer interview with an agency landed me 2k/month deal while the product was still in early beta. There is money behind SEO.

However, as I am seeing the latest generation of outputs I am also increasingly often thinking about the moral implications of it. On one hand, it was inevitable that AI will be used to generate content. The vast majority of generators today are not a threat because of their low quality output that makes it easy to identify them. But as the lines blur, how does one prevent Internet from being literally drowned in AI generated content? How do humans identify themselves?

Article generation is not a new thing

I think a lot about it recently. On one hand, I calm myself that “if not me, someone else would do it”. And that providing article generation that produces high-quality output does more good than allowing all the “text spinners” from continuing printing garbage. After all, if these articles produce value to the reader (e.g. by surfacing data in a more digestible way), then the end result is net-good.

Sourcing Data

The truth is that humans are not particularly necessary at the step of turning data into an article. As long as AI is fed high quality data (think research papers, wikipedia, etc), the outputs are going to be also high-quality, they can include tables, lists, references, quotes, charts, etc. The vast majority of research behind AIMD is not in "how to generate content" but what sources to use and what data to pick.

Human Input

I still think that articles without at least 10-15% of human touch are basically soulless even if they are packed with data. I am actively building education into the tool to push towards that outcome. But the fact that I keep thinking about ethical implications of what I am building means I am not doing enough.