Ah yes, a hot topic: AI. But how can I not write anything about a thing that will change billions of people’s life? And while so many people still think they are just another tool? NO they are not. In it’s second year, ChatGPT published its 5th version of its GPT algorithm with GPT-4o and every version puts something into its predecessor. Even Gordon Moore didn’t estimate this kind of speed in that even this article will be outdated just after several months.
Thinking and making assumptions about what kind of future awaits us is not easy obviously but I want to point out my concerns about my job (some other jobs too) to proactively decide what we can do to adapt it.
Custom / Personal GPTs
ChatGPT has already a market, if you will, where people customizes the algorithm for a certain purpose. It’s not just ChatGPT as you already know, from Windows, to Adobe, hack even Notepad will get an AI assistant. It is debatable how efficient they work at the moment but that doesn’t mean it will be more efficient in the future. It will be. The more it gets customized and trained in that way, the better we will see and use.
The current “problem” here is that you are dependent on certain algorithms now: GPT, LLaMa etc. Even if you train with a dataset of your preferred topic, it will act kind of the same as how we already know them. Things will get really chaotic when there will be more algorithm options so that everybody will have a chance to prefer an “approach”. You built an ecosystem with your own tool closed to internet, that’s fine. Just give the tool with detailed documentation and let “your” LLM algorithm turn into an assistant, maybe more than an assistant, what do you say?
Local AIs
I want to say, at least for now, that we are “lucky” they don’t work and be trained in our phones, computers etc since they require a huge resource. The reason why NVIDIA became a Microsoft-equivalent company in a year is that they can provide these resources. And everybody wants them. However LLM algorithms will be trained in our local machines soon, we just don’t know when. After that, everybody will easily create their own custom AIs for different purposes.
Just imagine your phone has a local AI, it doesn’t require any internet connection or huge resource and it track all your behaviours what you are doing on the phone and after some it will automatically do or suggest some stuff. This future is really close in my opinion.
Artificial General Intelligence (AGI)
This is where we will lose control in my opinion. By “lose control” I mean to track what is going on in the industry. Current AI tools benefit from what people has produced so far. These are things for people to understand easily. How about creating a language, dataset for only AI tools not people?
AGI will create a programming language to use only for itself. It will be non-readable for us but the language will behave as a bridge to, lets say, all OOP languages. I know this is really hypothetical and doesn’t soon it will arrive soon but we all know big companies are already working on AGI and sooner or later we are gonna get it.
AGI has tons of potential use cases. You know the tools that your company created only for yourself? Now imagine AGI can create, optimize and maintain them. Humans will be out of equation.
In Terms of Software
Last month, official GitHub Twitter account shared a poll about the usage of AI tools. It was surprising that one third of people doesn’t use them at all. Why I call it surprising is because it easies your life by writing only a couple of sentences with your mother language. You don’t even have to specify all of the restrictions (depends on your use case though). While things are that easy, it questions me deeply that how are we going to do when they can solve most of the problems that we are dealing in terms of sofware?
This is an ongoing debate and there are many different opinions. The current situation is that the most efficient AI tool in terms of programming, Devin, claims that it can solve %15 of the real world problems. And remember that we are in the second year of ChatGPT, whom Devin benefits from, What are we going to do when it achieved to solve, let’s say %50 of the problems? There will obviously be more lay-offs than ever.
There will always be people who need to maintain the infrastructure and keep things stabil however instead of doing 2 unit jobs, they will do 1 unit job because AI can do the other unit by itself. So I, as a software developer, need to adapt to tell AI to do that 1 unit job while being capable of doing the other unit by myself. Key word here is flexible. I call myself as Java Developer today even though I am able to do frontend and DevOps stuff, but when the time comes I need to be able to do more than that. Maybe we will be the one who will train AI on our systems and our job will turn from developer into maintainer or something.
Java, on the other hand, is heavily used for web programming. While so many applications rely on that, the need of OOP language for web programming has been changing in a way that people should be able to do more stuff. With recent Java 21 release, I think it has become to provide the need for that thanks to virtual threads and JVM optimizations. The bigger application means more developer requirement.
Even with the tools, developers will be needed for at least a couple of years. But AI is not like any other technology injected into our life where it created more job opportunities than it erased them. AI doesn’t offer much opportunites and I really don’t know what the future will bring us because even being flexible and agile might not be enough.
Leave a Reply