AI is Coming for Your Job - Right?

The conversation about generative AI seems to be on everyone’s lips today. The “super powers” these AIs wield are creating a lot of concern from many white-collar workers that ChatGPT and its contemporaries will replace their daily work. Anyone who must write, design, or otherwise work to produce content might have some very reasonable concerns about the future their role. We find ourselves in a time where AI can do many jobs, and often do those jobs in a fraction of the time. In this piece we will explore these concerns as well as some of the underpinning ideas and assumptions around this topic. 

An Automation Tool

The first thing to acknowledge is that AI is not human (per the previous article), nor does it have the capacity for human thinking. There are many reasons for this, but one of the simplest to understand is centered around the idea of “narrow” versus “general” AI. Every AI known to be in use today is narrow in nature. What makes an AI narrow is that it can only do the things that are within its original programming. ChatGPT or Midjourney cannot drive your car, your car cannot do your taxes, etc. 

The AIs we see in media like Jarvis, Ava, HAL, or Skynet are all examples of general AI – that is, AIs that have the ability to think broadly and accomplish effectively any task that is not specifically barred from its ability to act. An AI’s lack of ability to think and act outside of its parameters means that the ones we have today are more or less all automation tools – like a robot are that adds a panel to car on an assembly line. They are complex, and effective, and compelling, but - they are still just tools.

Given that the AI we have today is a tool, we should think about it, and how to apply it, in that context.  A hammer cannot do the work of a level, and vice versa. If we use the previous example of a robotic arm in a car factory, it can give us a good context for starting to understand the limitations of AI to replace people. The first robot was introduced to the auto industry in 1961 at a GM plant in New Jersey. If you go back to the articles written at the time, there was an overwhelming number of people decrying the end of the factory worker. It did most certainly have a measurable impact on the number of factory workers, but did not it eliminate them. Workers were forced to learn new skills to remain in the factory, but many jobs were still there for those willing to learn and adapt. 

Adjusting to Replacement Events

What these “replacement events” have done in the past is create a wave of new jobs in conjunction with eliminating certain types of others. Consider the advent of automated computing at NASA in the early sixties when Dorothy Vaughn, and those she supervised, transitioned to programming in FORTRAN rather than being subsumed by the rising primacy of automated computers. This next wave of computer-centered advancement today is not unique in its impact on either automation, or to white-collar jobs. An explosion of programmers, builders, and testers are some of the most obvious examples of new roles attached to AI. These jobs are often far more lucrative than the ones they replace as well. As a simple example, Glassdoor says that a just-graduated Jr AI Programmer could make over $140k in Seattle. 

Beyond the direct set of jobs, usually there is a second wave created at the same time. The robots in the auto factory need programmers, technicians, quality control experts, etc. These new AIs are a lot like this too, they need oversight, updates, maintenance, etc. The ebb and flow of the need for types of jobs has always been subject to the advancement of technology, AI is not unique in its impact here. This second wave of jobs will be fundamentally important to the implementation of AI – and there are some recent poignant examples that illustrate this.  

Because an AI cannot critically think, and acts purely based on data, it can be fooled into what would be obvious mistakes to a human.  Stable Diffusion is facing challenges of plagiarism and infringement now because a rather overt “Getty Images” logo that their AI imbedded in an image it made. There are fundamental issues that an AI cannot just “understand” because, again, it is not human.

Generative AIs are not the only ones with exposure either, recently an agency was running ads for a client. This client had placed a third-party, 360-degree product viewer on their website. The AI buying the media had been instructed to drive product views as the success metric. The viewer the client installed loads dozens of images of the product in succession to create the 360-degree view.  This led to a scenario that fooled the AI into thinking the advertising channel it was buying in when the viewer was implemented was responsible for driving hundreds of new product views (each photo load was counted as a view). This ultimately skewed the performance of the whole campaign for a couple of days until the human program manager noticed there was a problem. This type of thing happens all the time, and if it were not for human oversight, even major publishers like Facebook could spend $50k in a single day down a rabbit hole of targeting because it does not understand the context of the data it is receiving. 

These examples illustrate how there is still a role for people, even when AI takes over a job. No matter how sophisticated an AI may seem, it is still bound to its operating parameters and has real exposure to "common sense" mistakes. This is the fundamental limitation of AI today and helps explain why it is called narrow.

A Future With AI

The coming change to the content production jobs is inevitable, and many other jobs for that matter, but that does not mean that it must be negative. Those worried about their job being overtaken by the increasing usage of AI, need only look to history to gain insight. Examples, like the ones we have covered here, can be used to get a better understanding of how people have successfully navigated these changes. History tells us clearly that those who embrace the coming technology and figure out how to up-level their knowledge around it, are the ones who surf the waves of technological advancement instead of be swallowed by them. 

If you are still concerned, look to ChatGPT’s response to “What are the limitations of ChatGPT replacing white-collar workers?”

While ChatGPT and other language models have made significant advancements in natural language processing, there are still limitations to their capabilities. They lack the creativity, adaptability, and critical thinking skills that many white-collar workers possess. Additionally, they may struggle with tasks that require emotional intelligence or human connection, such as negotiating contracts or providing personalized customer service. ChatGPT and other AI tools should be seen as a complementary tool for white-collar workers, rather than a replacement for them. It is essential to recognize the value of human expertise and experience in many industries, and to use AI tools to enhance their abilities, rather than replace them entirely.

AI is coming; in many spaces it’s already here and there is no way to slow this down. AI integrating into our industries is not the end, like every innovation it requires that we adjust, grow, and expand. There is substantial opportunity as AI utilization increases; we can either get rolled by it, or we can embrace it, and evolve with it.