A Personal Statement About Generative AI

When it comes to technology, I’m usually optimistic. I believe technology can be liberating. It can make life better. I look at technological innovations in the home. Light bulbs. Washing machines. Refrigerators. These changed domestic life and freed up many hours for people to live a less burdensome life.

Over the years, I’ve brought this optimism to digital innovations. Personal computers. The Internet. Smartphones. The case for them was obvious. They made it easier and often cheaper to do repetitive and challenging tasks that required a lot of time and effort. They made life better.

I approached Generative AI in the same way. I wrote about this over a year ago, asking how it might help someone like me. I was sceptical, but I tried to find the positives.

I was naive and wrong.

It’s become increasingly clear that the negatives around generative AI far outweigh the positives. And that there are very few positives.

The True Cost of Generative AI

The training of these systems involved the greatest theft of intellectual property in human history. No one asked if my 20 years of writing here could be scraped to train large language models that will make someone else richer. And yet it happened. It happened to every creative – authors, musicians, photographers, artists, and designers – who posted their work online. Theft. Plain and simple.

And with all that data, Generative AI manages to solve very few real problems. It approximates solutions. But doesn’t deliver them. There’s no reason to be hopeful Generative AI will get much better in the coming years. It may have already peaked. All we are left with is poor-quality impersonations of things that could be done better by actual humans.

The Lesson of Generative AI

What Generative AI has done is lay bare the attitudes of the people who advocate for it. How little respect they have for their employees. Or for the whole class of people who work in creative industries. To these “leaders”, it doesn’t matter that Generative AI generates mostly crap. They believe their employees are capable of little more than crap anyway. And art is just well-marketed crap. At least Generative AI is cheap crap. Crap that won’t speak back like a human would. Or demand a pay rise.

Which is crazy given how expensive Generative AI is. Not just the outrageous amounts of funding that have gone into a technology that has no clear and compelling use case. But also, the environmental and cultural costs of Generative AI’s insatiable appetite for computational power.

All The Use Cases Are Already in Use

Announcements from Apple’s recent Worldwide Developers Conference (WWDC) were reported as a sign the company is trying to play catch up on AI – or Apple Intelligence, as they called it. While some generative features were shown, most of the useful tech was things Apple had already been working on for years. What until now they described as machine leaning.

Back in 2012, I reported on Apple acquiring the Italian software developer Redmatica. This was for tech that went into managing audio files and music production features in Apple’s Logic products.

In 2015, I attended the Audio Engineering Society (AES) conference in New York, and there were plenty of machine learning tools in the algorithms that powered software for mixing music and repairing audio files.

Whether it’s making music, editing photos, or making and moving products around the world, the past 10–15 years have seen a huge shift in the power of software for creative industries, logistics and manufacturing.

But it’s important to separate new generations of software that solve complex problems incredibly fast from the generative AI which is plugged as something that can create new things from the smallest of simple prompts.

The Crucial Technological Distinction

If AI is increasingly being used as a catchphrase for all sorts of software innovations that we already use and live with, then it’s worth distinguishing between AI and Generative AI.

The latter, Generative AI, is the tech I’m talking about here. The words created by ChatGPT. The images coming from Midjourney. The changes Google is making to search.

I believe this distinction matters. On one side are tools that speed up or make more reliable tasks that we can already do. That help manage processes and workflows that are still directed and supervised by real people. Tools that already exist and already work.

On the other side are tools that claim to be able to help us do things we can’t do. That claim to be able to manage processes largely unsupervised. That generate work on its own.

It’s the difference between a tool that helps a writer edit their work faster and more consistently and a tool that tries to replace the writer altogether.

But I acknowledge the distinction might be hard to hold onto as more and more apps have AI features shoehorned into them. Even if we choose not to use them they will still be there.

Allow Me to Rant for a Minute

I find it hard to write about this. Not because the technology is hard to understand or difficult to explain. But because sitting down to write, I can feel the walls caving in around me. It’s as if the weight of all my beliefs about the cultural value of technology and innovation, the aesthetics of a life well lived, the morality of treating people well, just collapse. I find myself wanting to scream “rich people in Silicon Valley want to convince you that art, democracy, and human connections don’t matter because they can make more money if you buy into the shrunken version of humanity they are peddling.”

Okay, rant over. Back to the essay.

Choosing Humanity

Let’s put this in the context of our relationship for a moment. Writer and reader.

This blog has over 2,200 posts. Behind the scenes, it’s a bit of a mess. My cataloguing system, categories and tags, evolved haphazardly over nearly 20 years. I could run some tool in the background to clean that up. It would make things easier for me. It might even make navigating the site easier for you. Then again, you might not notice. It’s unlikely to change our relationship.

But, what if I used a tool to write the blogposts themselves? Imagine if I revealed this very essay was written by such a tool. How would you feel? Two years ago, you might’ve thought, “Okay, that’s interesting.” Now you’d probably just say I was an arsehole.

That’s because doing the latter undermines the relationship we have. Writer and reader. While the former, the relationship between the website and whatever we call someone who navigates through a website is much less personal.

Choosing Creativity

You’re not here, on this website, for any old random words or images. You’re here for something specific. For words that inform and inspire. And maybe for art and images that reflect a thoughtful engagement with the world.

Using Generative AI would undermine that. Which is why I’m not going to use that technology. It has no place here. It doesn’t offer anything I need.

At one point, I thought it might help with things like writing headlines and summaries. It never did. It also never helped me organise my thoughts. Or summarise things I needed to read. It wasn’t good for sparking ideas or providing examples I could quote.

Generative AI can’t do any of that in ways that fit with how I work. There’s no promise it will get close to it either. Even if it did, the ethical concerns make it untenable.

Plus, my whole journey is about embracing weaknesses and addressing them creatively. Learning to write better headings, something I don’t feel I’m good at, is just another in a very long string of creative challenges.

Having a Healthy Relationship with Technology

Being sceptical about Generative AI doesn’t put you in the same category as people who doubted the prospects of the internet or personal computers.

That scepticism was born of ignorance. What’s fascinating this time is how many of the people who are critical of Generative AI are folks who use technology every day. That have been and continue to be early adopters of other tech. It’s informed scepticism. Not ignorance.

Which is why I’m not committing myself to rejecting everything that might be labelled as AI. Creative software that already uses machine learning, like most forms of music production software, will still be part of my life. I will judge, on a case-by-case basis, writing tools that help me edit and organise my work, depending on how the AI tech works. Photo-editing software that speeds up removing noise or dust spots from images will always be welcome.

But I won’t be using AI to generate work. Creativity matters. Humanity matters.

The fact this blog comes out of one person’s imagination still means something. So, it will continue to be made in the time-honoured artisanal way. This has proven over the years to be the best path.