Why is ChatGPT not good enough as a programmer?

What changes do AI tools bring to the lives of programmers?

‍

From its very nature, AI tools fundamentally change developers’ lives, and some experts even believe that we have arrived at a point where the use of AI tools is daily routine in such industriesealthcaregriculture and transportation. Intelligent systems can provide added value to all stages of software development and can help developers’ jobs in numerous ways.

‍

“AI tools will be integrated into the toolkit of software developers. It will become commonplace for developers to receive code suggestions from an AI tool, which they can choose to accept or reject. If accepted, the developer may do modifications based on their decision on how it could be implemented more effectively. The tool learns from this interaction and provides better suggestions in the future, and the developer may be more likely to accept those, however programming knowledge is still essential in any case,” summarizes Péter Szél, the Head Mentor and CTO of Green Fox Academy.

As already mentioned in our previous article, ChatGPT for instance, became a tool in developer’s toolkit, that allows more efficiency in daily tasks. When utilized well, tech professionals can spend less time on repetitive tasks, allowing space for innovation, creative problem-solving and tackling complex issues.

‍

‍Aliz, the first Hungarian partner company of Google Cloud has started to integrate AI tools into their workflow even before ChatGPT would have become a media sensation. For them, the advance of the language model did not come by surprise.

‍"As a company that had previously worked with language models, and led projects, where natural language processing was involved, it’s not a change that would turn us upside-down. It is a new tool that makes certain tasks easier, but it has not brought radical changes for us," explains Tamás Móricz, a data scientist at the company.

‍

Why is ChatGPT not able to replace human programmers?

‍

Although we are talking about an impressive artificial intelligence that has the ability to generate user-friendly solutions in various situations  and  replicate human communication with utmost precision, its advanced capabilities cannot compete with the skills and toolkit of a skilled programmer.  Let’s see some of the reasons:

‍

• First of all, programming is so much more than code writing only. It incorporates problem-solving, algorithm design, testing, and the development of applications that meet business and customer needs. These tasks require human interaction, critical thinking, and deep expertise that even an AI like ChatGPT cannot fully replace.

• Second of all, ChatGPT lacks the contextual knowledge that developers bring from client meetings related to the project. Programming requires teamwork, and human communication and collaboration is often needed to solve complex problems successfully; that’s something AI cannot fully replicate even with a precise description of the given task.‍

• While ChatGPT has the ability to generate code snippets and provide suggestions, the contextual understanding and proper integration of these into the overall system are also developer responsibilities. It is important to note that ChatGPT cannot modify previously written code or comprehend complex solutions. Since the generated code is based on pre-existing code, ChatGPT cannot guarantee that its code is: error-free, easily understandable, and maintainable code. ChatGPT generated codes may not meet the company’s security requirements and it could be a direct replica of a code found on the internet.

• Yet another limitation is that at its current stage, ChatGPT’s capacity only goes as far as creating relatively simple programs. When we ask something too complex - such as a sophisticated game or business application - it immediately acknowledges its limitations and that the task exceeds its capacities.

‍

According to Mária Tóth, the Director of Software Development at DXC Technology: “There is no point in worrying, since the process of software development rages from understanding client needs to live operation, during which customized developments are deriving from modeling business needs, algorithmization and fully comprehending client needs.

This multi-layered creativity cannot be replaced by ChatGPT at its current stage. ChatGPT can work as an efficiency-enhancer that ensures high quality in the hands of developers, not an alternative tool that could replace their work.”‍

‍‍

How can ChatGPT facilitate us to become better programmers? 

‍

At this point we already understand that there are several aspects in which ChatGPT cannot replace programmers, however luckily it can assist them to become better IT professionals in many ways.

Firstly, it can be an amazing source for any tech-related topic, whether it's about clarifying concepts, troubleshooting, problem-solving, or staying up-to-date with the latest technological trends. With its vast knowledge base, it can recommend online courses, tutorials, websites, and other resources in areas such as web development, data mining, machine learning, and many more.

As mentioned earlier, ChatGPT cannot test or understand complex development solutions, but by asking appropriate questions, it can provide guidance when facing issues in a given project or task related to developing.

‍Additionally, using ChatGPT allows gathering information about technological trends and advancements, such as artificial intelligence, IoT, or any other IT field. This enables us to stay constantly updated on the latest technological trends, which remains a critical point for IT professionals to gain recognition on the job market.

‍

What can we expect in the future?

‍

We know by now that we cannot give ChatGPT such commands, as “create an e-commerce platform that is more efficient than Amazon”. If we want this, then we have to dedicate time and effort to first learn what makes the platform of Amazon so great, then to  find the way to make a better one.

For this very reason ChatGPT (or other NLP-based tools) still function with limited efficiency when creating softwares designed for gaining advantage in business life, or when competing with human creativity and resourcefulness.

Therefore, we can be certain that there will be a prevailing need for diverse human and programming skills for developing softwares that are equally valuable from both business and customer perspective, and those skills are unlikely to be acquired by computers alone in the near future.