Software Engineers Are in Denial

We need to consider the variation in the problem spaces of top 10%, top 25%, and the long tail of software engineers. Engineers doing nothing more vanilla starter NextJS projects might be at the highest risk for their skillset eroding to technical advancement.

I remain unconvinced that people working on systems that evolved over a decade or more, with microservice spaghetti, can truly leverage things like ChatGPT. If I'm on a greenfield project or want to generate something new, sure.

Until I can say "migrate this AWS account to GCP and/or Azure, and update all of the deployment tooling for the hundreds of repos supporting it, and update the DNS records too" -- something that might take teams of engineers within large organizations a quarter or more to figure out, no, I don't feel at risk. The degree of ambiguity in the spaces we operate in (eg: the top 25% of software engineers) are far too great for this sort of technology.

This is the very same argument that artist are dealing-with with AI artbots. The bottom line is that GPT, and such, are tools that humans must adapt to use in their jobs. Companies will not hire ChatGPT to do the work, they will hire humans to use ChatGPT to do their jobs. As good as AI bots are, they still need a human to figure out how to best use them.

Computer jobs have always been about adapting to the changing environment. We are here again. Adapt to AI bots or get left behind. There's still time to adapt but you can't wait too long since the tech is evolving really fast.

I was an AI skeptic for a long time, but I've slowly started to come around on the use of tools such as ChatGPT. I don't think they'll replace programmers, since in the end someone still has to make the judgement about integrating the output in a project in a coherent and useful way, and this takes just as much mental effort and training as writing the code in the first place.

Currently I'm playing around with ChatGPT and soon other tools in a collaborative and generative fashion, where I give ideas and prompts and get multiple choices. In the end I still have to make the choice (and if necessary adapt it to my needs).

Eliza 2.0, although certainly more sophisticated, the predicted demise of software engineers, like doctors in the 1960s is greatly exaggerated. Both Gigerenzer and Larson’s two books are well worth the read.

ChatGPT is more useful to layman which is why they find it so popular right now. Next week they will move on to something else.

The _author_ should certainly feel threatened by ChatGPT - that article reads like generated output.

Are developers in denial about ChatGPT?