I’m between two worlds when it comes to AI. At my day job as a technical writer, I’ve been involved with researching AI and implementing ways to use it to help the other writers in our department. AI is just another new technology like the others I’ve encountered, from personal computers to the Internet, in my 43 years in the industry. But as an author and a concerned citizen, I see the many problems with AI from how it destroys creativity, entire industries, and the environment. AI offers legitimate benefits, and it has legitimate issues.

While we shouldn’t reject any new technology outright, we shouldn’t blindly embrace it without serious consideration. There are several things AI supporters should learn from its opponents.

AI requires human intervention.

Proponents imagine AI agents generating finished apps, websites, and documents without the need for expensive human coders, testers, and writers. Experience shows me that AI doesn’t work as well as they think it does.

For example, I used our company’s enterprise AI system to make updates to our custom DITA Open Toolkit plug-ins. AI did save me weeks of research, trial-and-error coding, and debugging. But when I tried merging the customizations into another folder, I ran into errors. I discovered that the AI had hard-coded the full path for one of the files. Because I had years of experience creating this code by hand, I fixed the problem easily.

AI has always suffered from hallucinations, where it makes up information that doesn’t exist in its models. Now, organizations and governments are training AI models to provide information that is favorable to them. LLM poisoning makes it even harder to trust the results provided by AI. This is why we can’t rely on AI-generated output without independent human verification, and we need trained, experienced, and knowledgeable people who can do this work.

AI requires human intervention through all stages of its implementation. Humans need to test and verify, because humans will use the products produced by AI. While proponents may claim that AI can do all the work, people are still needed to check and shepherd AI production processes. Without them, AI products would be unusable.

AI should stay out of creative professions.

AI is especially unusable when it comes to creative works, such as music, movies, art, and writing. I’ve also discovered this from experience. From the first time I used a word processor on my Commodore 64, I’ve seen how technology has improved the writing process and given authors opportunities we’ve never had before. But as I’ve used AI, I can see its serious limitations as a creative tool.

People turn to art for an emotional experience. We want to feel something, whether it’s excitement, shock, or the warm satisfaction of a happy ending. AI may be able to simulate emotion, but it lacks the lived experience and human connection that makes it real. Sure, AI can generate an 80,000-word novel using all the popular genres and tropes, but it would be emotionally empty. It’s like being enticed by a sumptuous-looking meal, only to discover it’s just prop food with no flavor.

There’s no greater example of this than the current practice of AI companies obtaining old and rare books, extracting the words to use for its language models, and then destroying them. They want the words, but not their meaning. And they want to deny that meaning to others.

Creativity can only flow from human emotions and human experience. And it can only flourish when creators have ownership and control over their works and how they present them to their audience. This type of connection is beyond AI’s capabilities. AI companies should accept this limitation instead of being antagonistic towards it.

AI must be environmentally responsible.

A bigger problem with AI lies beyond what we see on the screen: the massive data centers used to process all of that information. The toll these centers place on the environment, electrical grid, and the climate should concern us all. But AI companies tell us these centers are necessary, and the environmental cost is a small price to pay for progress.

These are the same arguments corporations made decades ago about factories. Back then, factories dumped raw sewage into nearby bodies of water and straight-piped toxic exhaust into the air. We were told that to make these factories environmentally safer would cost too much, and the jobs and economic boost factories provided made them worth the ecological damage. By the 1970s, when I drew this cartoon for my high school newspaper, the destruction caused by these factories became too visible to ignore and accept.

The EPA and decades of government regulation and technological improvements have shown that factories can be made environmentally safer. The same is true for data centers. Certainly, the companies that have the know-how to build massive arrays of servers also have the knowledge to cool them without draining our water supply and power them with their own clean and renewable sources instead of overtaxing local power grids.

AI companies shouldn’t wait for the government to intervene. They should be taking steps now to make their data centers more sustainable and better for the environment. If they do, they will get less public pushback to build them, and they won’t face the inevitable lawsuits and penalties when the government stops being their lapdog.

AI must serve humanity, not the other way around.

But it may be hard to convince AI companies to be more responsible for the environment when they don’t feel responsible for humanity as a whole. Look at the horrible beliefs and activities from tech leaders. It seems that the problem they want technology to eliminate is people.

But we meatsacks are pesky creatures. Not only won’t we go away, we will make sure any technology that doesn’t serve us doesn’t last.

Four decades in the computer industry have shown me that good technology is useful technology. Conversely, technology with no useful purpose, ones that are solutions looking for problems, or products that are hostile to their users disappear from the market. For example, Ring faced severe backlash when consumers discovered they were joining forces with Flock to use their cameras for community surveillance. (And Flock is so despised, a Tennessee county commissioner publicly threatened their CEO.) 

All the talk about AI’s potential won’t do any good if the public doesn’t see any tangible net benefits today. It will be regurgitated just like every other technology that has been shoved down people’s throats. The popping of the AI bubble won’t only be inevitable, it will be celebrated.

AI must serve humanity, not the other way around. It must offer practical, tangible benefits to people today. It must be human-centered, because people need to verify, test, and use the products it creates. It must also understand what it can’t do, including replicating the real human emotions people look for in art. It must be produced responsibly and sustainably and protect the environment and resources we all depend on. Most importantly, AI companies and its executives must respect human rights, human freedom, and human dignity. Tech bros may harbor the bullied boy genius fantasies of a world run by intelligent robots, but wiping out humanity is killing the goose that lays the golden eggs. Without people, there is no knowledge, no creativity, and no technology.

AI opponents have a lot to say. AI proponents would be wise to listen.