“AI is coming for our jobs”?
When image-generating AI tools like Dall-E, Stable Diffusion, and Midjourney started to attract attention last summer, many people bemoaned (or observed with wry irony) that “art”—once considered the single human activity robots would never be able to emulate—seemed suddenly to be on the chopping block in a coming automation revolution. When ChatGPT1 dropped last November, people started saying similar things about Large Language Models and all forms of writing-based work. Earlier this month, OpenAI announced GPT-4, possibly the most powerful LLM yet, which promises to accelerate the automation of human jobs even more quickly than people expected.
In the past few months, and especially the last couple weeks, much digital ink has been spilled about what this all means. Like many people, my head was spinning from reading it all and trying to incorporate these rapid developments into my view of the world. Now that the news cycle has cooled off slightly, I feel able to think more clearly about what is actually happening. In particular, something I have noticed is that there seems to be an emerging consensus among tech-minded people that AI is coming for everyone’s jobs. This is a provocative statement, and I want to examine what it actually means. Up front, I don’t think it makes sense to panic about today’s generative AI taking away (most) people’s employment in the short term. However, it can absolutely take over many tasks. AI is first and foremost a tool, which humans can wield to more easily accomplish various goals. This might be bad, but it also could be… really good?
What is clear is that AI will cause many jobs to change. Consider the narrow example of ChatGPT and copywriting. Text generating AI is already being used to take over some of the less-skilled work in the copywriting stack. Humans still need to decide what should be written, and even more importantly when it is good enough to publish2. But humans no longer need to spend as much time collecting information about a topic and arranging it into coherent, grammatically-correct sentences. As a blogger friend of mine rather militaristically put it: “Frontlines of copywriting are dead. Only room for generals now.”
Thus, in copywriting, you might plausibly say that AI is “coming for everyone’s job” in the narrow sense that pretty soon AI is going to be integrated into every copywriter’s workflow. But it’s not going to get there on its own. The tool exists, but humans still need to figure out how best to use it.
And so, as a self-employed writer-person, my attention turns to my own life and work. I am not a copywriter per se, but my work does center on producing and editing English prose. The emergence of an AI tool capable of these tasks is like a siren going off in my brain. It seems likely that an AI could do some parts of my job, today. But which parts, and how? The only person who can answer that question is me.
For the purposes of this investigation, I will use the terms “ChatGPT” and “GPT-4” interchangeably, but I wish to clarify that they are distinct. “GPT-4” is the large language model; ChatGPT” is the chatbot interface to the model. Thus, I am using ChatGPT (more specifically, a subscription to ChatGPT Plus) to access GPT-4. GPT-4 is the newest GPT model from OpenAI and not yet widely available; I expect it may become more widely available in the next six months or so, which happens to be the ideal time horizon for using AI to “do my job.” ↩
Although some outlets are definitely not taking very much care in this regard. ↩