Will AI Replace Programmers?

Is there a distinction between coders and programmers?

There is a great video on YouTube by the Applicable Programming channel about the difference between coders, programmers, developers, software engineers, and architects. The video illustrates that coders are like musicians playing to compose notes given to them.

In other words, a coder would typically be someone who writes code based on clear specifications provided to them.

In the context of this blog, I would emphasize that the distinction is in mindset. I do not believe when someone is saying coders will become extinct they are categorizing people. This can sound alarming, but the underlying message is more about the approach to work, the passion involved, the desire to solve problems, and how one utilizes AI tools effectively. It's not merely about the tasks they perform but how they engage with and adapt to evolving technologies and challenges in the field.

The current state of AI tools in software engineering

Most of our programmers at Devōt use GitHub Copilot. It uses the Codex model, a descendant of OpenAI’s GPT-3, designed to understand and generate human-like code. code completion tool developed by GitHub in collaboration with OpenAI.

The primary function of GitHub Copilot is to suggest whole lines or blocks of code as developers type, essentially acting as a pair programmer. The AI draws from its training data to offer contextually relevant recommendations, significantly speeding up the coding process and potentially introducing developers to new libraries and frameworks that increase productivity.

There are many more AI tools used in software engineering, but Codium AI is also one of the most popular. This platform provides real-time feedback and suggestions to improve code quality and efficiency. By using machine learning techniques, Codium AI helps developers detect and correct potential issues early in the development process.

We need to talk about Devin

Introduced as an AI software engineer, Devin is designed to function not just as a coding assistant but as a full-fledged member of the development team. Leveraging advanced machine learning algorithms and natural language processing, Devin can understand project requirements, write substantial amounts of code, and even debug existing codebases with human-like intuition.

What sets Devin apart is its ability to interact with developers using human language. This makes it accessible to professionals without deep technical expertise in AI, freeing up human programmers to focus on more strategic and creative aspects of their projects.

At least, this is what they say. Will Devin live up to the hype or fall short of expectations? Currently, Devin AI is in its beta testing phase and is accessible to a select group of users through a request-only system.

AI and its limitations in software engineering

As with everything, AI has its limitations, especially as it gains popularity and is still in its early stages.

1. Don't forget that AI is replicating human errors

Since AI systems are trained on vast datasets that often include code written by humans, they inherit the biases and mistakes present within that data.

This means that AI might continue to create code with problems, such as security risks, slow algorithms, or poor coding habits. Developers have to be careful and check AI-created code as thoroughly as they would code written by people to make sure it’s good and correct.

1. Are you arguing with your AI?

Whether you are a programmer or in a non-tech job, if you have been using GPTs, you have probably spent your time correcting its answers.

While AI tools like code generators can significantly speed up the development process by automating routine tasks, they are not always precise or contextually appropriate. The time saved in initial code generation can sometimes be offset by the time required to refine this code to meet specific project standards or to integrate seamlessly into existing systems.

1. Not much of a side-kick in innovative problem-solving

AI's capability for innovative problem-solving is also limited. While it can execute defined tasks and optimize existing solutions, AI lacks the ability to perform the deep, abstract thinking necessary to devise truly innovative or out-of-the-box solutions.

It struggles with tasks that require understanding the subtle aspects of human behavior or the specific details of certain problems. Developers can understand the broader impacts of innovating and solve complex problems that AI currently cannot handle.

Can AI replicate human creativity?

AI systems (especially those using machine learning and generative AI) are proficient at analyzing vast amounts of data, identifying patterns, and applying learned information to generate new content. These systems can create music, art, and even write code. However, AI's approach to these tasks is different from human creativity.

AI lacks the ability to experience emotions, derive meaning from cultural contexts, or engage in the kind of subjective thinking that often inspires human creativity. While AI can produce novel combinations based on existing information, its creations are limited in the data it has been trained on. This training limits AI to what is already known and reduces its capacity for true innovation.

Many of our software engineers said that AI should stay out of the art, but here we are also talking about creativity in programming. Creativity is not just about writing new code but also about solving problems in unique ways, optimizing systems, and foreseeing future needs that may arise as technologies and markets evolve. Programmers often draw upon their experiences, intuitions, and personal insights, which are irreplicable by AI.

To read more on this challenging topics take a look at our blog where we cover benefits of AI in programming.