My Thoughts on AI

As a writer, I've been asked time and again about how AI will affect my life and if I'm nervous for my career. Short answer: not at all.

As we've seen from the jumps in each stage of ChatGPT, Google AI, etc, is that they have reached an “okay” level of accuracy and, while they claim big changes are being added, they aren't actually making any noticable improvements. Take Google. They have all the data imaginable to draw from with its dominance in search, and yet their AI is wildly inaccurate. AI models in general appear to have plateaued already and all improvements are made on a logarithmic scale, meaning that it takes a significant amount more data, effort, testing, etc. to make smaller and smaller improvements. With Deepseek we saw the big difference being it's efficiency, rather than computing accuracy or capabilities.

The first step in AI being used in the general marketplace is to replace your general “paper pushing” jobs. Setting meetings, scribing for doctors in hospitals, organizing agendas, that level of complexity.

With “okay” AI being used in these jobs, managers will see a flood of mistakes, as most paper pushers are quite good at their jobs compared to AI, but slower and more expensive in the moment. Management will see a loss in profits.

There are two ways this goes from here.

At a high level organization, push will come to shove when comparing the cost of using AI making lower profits and having to pay less in labor costs. Whichever makes more money is likely to win.

However, what had to be taken into account is training. Unless your company makes AI, it's controlled by a 3rd party. This 3rd party trains the model itself, and your company is not allowed to make any direct changes. With a human, you can train them to do a better job.

Lastly, there's the issue of accuracy. We all saw Google's made up cheese fact on their Superbowl commercial. Imagine if it was making an appointment for a cancer patient. These AI companies would be liable for the damage caused by their products. With a human, again, they can be trained.

To summarize, I'm scared of the short term human consequences of the push to AI. In the long term, I expect it to have no more of a long term impact on the everyday person's life than blockchain. Some nerds will like it, it's fun to theorize about, but any change will be but a ripple.