ChatGPT and generative AI will not eliminate programmers

-

Think of ChatGPT as REALLY advanced Autocorrect

Autocorrect has been around since about 1993. If you're wondering if ChatGPT is gunning for your job, think about all the jobs autocorrect has eliminated... go ahead, ask ChatGPT, Google it, I'll wait. I don't know either, but the short answer is, "Yes, autocorrect may have reduced the number of people needed to proofread a manuscript, or reduced the amount of time you need to double or triple check text before hitting 'send', but realistically it has simply made writing more efficient".

ChatGPT is no different

I heard the hype about how ChatGPT will eliminate programming jobs in 5 years10 yearssomeday and thought, "well yeah, the same way that the Model-T evendually eliminated ferrier jobs when it was released in 1908". In a way, there's some truth here...in the same way that java or ruby eliminated a lot of the boilerplate and drudgery Fortran or COBOL (or assembly) programming had humans wasting time doing, it is obvious (to me) that ChatGPT will dramatically change how "programming" is done, but won't change the need for the creative input.

I read (and watched) this video ChatGPT Will Replace Programmers Within 10 Years and there are some interesting predictions. The author, however, hides some details that are obvious (to me) that he glosses over with extremem predjudice (presumably in the interest of clickbait, which I don't fault him for, after all I'm doing the same thing). In his example he writes and agent that creates and agent that can generate Python code to solve programming problems. More specifically, he writes an AI model that can write and AI model that will generate code in python. I was left scratching my head wondering "but to write the model, you had to have the creative inspiration to write the model that writes the model...which is programming. I took a stab at building a similar system to generate 3d models of various organization structures, and I must say I was pretty impressed....however, what I discovered was lacking (the new frontier) is knowing how to hack the model to interpret what I really wanted. This is actually REALLY, MIND WRENCHINGLY difficult. While the end result is pretty impressive (it could generate pretty impressive heirarchies with fairly limited inputs) but I had to significantly change how I approached solving the problem.

Sure, programming as we know it definitely going to change

I sure hope so, the insane amount of human time and effort wasted to solved already solved problems is crazy, but my prediction is that AI is going to create MORE programming jobs (though maybe we'll call it something different). Sure, the market for folks to create horseshoes and scoop horse manure from the street is MUCH smaller than it was in 1850, but, now the people who would have done those jobs are doing something else...Programming is no different, sure, the number of folks who can build a spring starter app (by copy/pasting code from stackoverflow) will go down, but the number of people who can build a new GPT model to generate a user interface that appeals to golfers is going to grow.

Who should worry?

I DO think there are certain areas that will be eliminated outright.

  • Programming jobs that copy/paste boilerplate from stackoverflow. There are millions of "programming" jobs that are essentially completely interchangable with what you can train an AI model to do better/faster/just as reliably as a human. Hopefully those will be gone in 10 years (I'm somewhat optimistic, but a guy can dream).
  • Simlarly, what I call "non-creative, technical jobs" will go away. changing a User interface from one template to another or upgrading code from one api to another will become trivial. It's much easier and more effecient to have an AI do this than a human.

What SHOULD businesses do

Businesses will be most impacted when they adopt the stance of AI as an assistant to make a human more effective. Driving down costs by thinking human effort can be eliminated is a pipe dream...put another way, thinking you can save a nickle in your customer service costs by spending a dollar on generative AI technology is a fool's errand. Focus on "how can I use this to make my business BETTER, not how can I make my operations CHEAPER?" I've lived through multiple generations of companies so focused on the bottom line they lose site of the real opportunities to win with this shiney new technology.

References

Comments

Post a Comment

Popular posts from this blog

Push versus pull deployment models

-
There are two deployment models: push and pull. In a "Pull" deployment model, the individual servers contact a master server, download their configuration and software, and configure themselves. In a "Push" model, a master server pushes the configuration and software to the individual servers and runs commands remotely. While neither technique is right or wrong, they both have some specific advantages and disadvantages that you should understand when making a decision. In the ruby hosting world, an example of a "push" deployment is heroku , while Engine Yard is an example of a "pull" type deployment. In the Pull model, each server has information about how to obtain it's configuration, when it boots (or whatever the configuration triggering event happens to be) it can continue to proceed without intervention from the master server. In the Push model, because the master server is orchestrating things, it will typically need to have a c
Read more

the myth of asynchronous JDBC

-
Image
I keep seeing people (especially in the scala/typesafe world) posting about async jdbc libraries. STOP IT! Under the current APIs, async JDBC belongs in a realm with Unicorns, Tiger Squirrels, and 8' spiders. While you might be able to move the blocking operations and queue requests and keep your "main" worker threads from blocking, jdbc is synchronous. At some point, somewhere, there's going to be a thread blocking waiting for a response. It's frustrating to see so many folks hyping this and muddying the waters. Unless you write your own client for a dbms and have a dbms that can multiplex calls over a single connection (or using some other strategy to enable this capability) db access is going to block. It's not impossible to make the calls completely async, but nobody's built it yet. Yes, I know ajdbc is taking a stab at this capability, but even IT uses a thread pool for the blocking calls (be default). Someday we'll have async databa
Read more

Installing virtualbox guest additions on Centos 7 minimal image

-
I've just spent some time setting up a bare Centos 7 image to support development stripping out as much as possible. While CoreOS is probably a better choice, we run redhat in production at this point and centos is a better fit for the time being. The problem I've found is that many of the instructions available via google were written against prerelease versions or espouse manually installing random rpms instead of just using yum. While I get this "works", I'm not a huge fan of this approach and would rather do everything with the package manager. After a LOT of scouring and trial and error, I finally found the "magic" combination. The winner is courtesy of http://www.pc-freak.net/ . I've taken these instructions and tweaked them slightly for my purposes. The original poster did this on a more "full fledged" version of centos with a windows host and my instructions are for a minimal install using and OSX host (though I'm certa
Read more