The rise of the
super-developer
Many developers already make regular use of GenAI technology to short-cut their work processes. When is the result actually an improvement in productivity? How do you enjoy the tremendous opportunity presented by this revolution, while hedging its risks? Why is the entire development market in the process of reskilling and upskilling? And is the role of a software developer on the way to changing from being a coder, to being a DJ of automation and AI solutions?
We sat down for a conversation with Gabriel Ben-Harosh Hasson, CTO of Matrix DevOps, and Leonid Pevzner, senior lecturer at John Bryce, about how the GenAI revolution is currently changing the world of software development.
“We have been using AI tools for years to streamline and shorten development work processes – it isn’t a new phenomenon,” says Gabriel Ben-Harosh Hasson, CTO of Matrix DevOps. “However, the latest GenAI developments have enabled a huge jump in capabilities and in their use. Suddenly, it is easy and accessible, and professionals in the world of development – who have always been early adopters – are already using tools like Copilot regularly in their work. However, in companies where such things are not formally organized, this happens informally, which means there is less control over how these technologies are implemented. It’s becoming a bit of a Wild West, and we must bring order to it.”
44% of developers use AI tools in their work, and 25.5% plan to use AI tools in the near future (Stack Overflow survey, May 2023).
Indeed, according to a developer survey by the Stack Overflow website, conducted in May 2023, in which approximately 90,000 developers worldwide participated, the use of GenAI tools as an aid in development work is already very common today. For example, 44% of participants in the survey stated that they use AI tools in their work, and 25.5% of participants say that they plan to use AI tools in the near future. So, it is reasonable to assume that from the date of the survey until today, the percentage of users among developers has already exceeded 50%. The survey also revealed that 83% of developers use AI tools for writing code, 49% for debugging and receiving assistance, and 34% for documentation. Yet, despite increasingly widespread use, user trust in the performance of AI is still limited. For example, only 3% of users of AI tools stated that they “very much trust” the accuracy of the results, and 39% of users stated that they “trust (these tools) to some extent”.
52% of ChatGPT responses generated by developers include inaccuracies (Purdue University study, August 2023).
“The feeling of insecurity regarding AI-tool results is completely justified,” says Gabriel. “For example, last August, results were published of a study conducted by researchers from Purdue University in the USA, in which ChatGPT’s answers to 517 questions from software developers posted on the Stack Overflow website were examined. The study shows that 52% of ChatGPT’s answers contained inaccuracies – a worryingly high figure. However, participants did prefer ChatGPT’s answers over answers from human developers 39.34% of the time. ChatGPT gives comprehensive and well-reasoned answers, so to speak, and we are tempted to believe it and give up our sense of criticism – a sense that would serve us well, in this case. After all, in addition to the question of whether the results are accurate, there are other barriers that need to be taken into account, such as information security issues, and fear of code leaks. Consider the case of leaked source code of one of Samsung’s new products due to the use of GenAI by the engineers, and exposure of companies to lawsuits arising from the use of code from unknown sources.
“Eventually these barriers will be overcome, and generative artificial intelligence tools will enhance development work: They Enable much more to be done with much less, they help with monotonous, repetitive work, such as documentation, and leave developers more time for creativity and high-level problem solving. Thanks to these tools, there can be a jump in the capabilities of developers, and productivity will take a huge leap as well. But, you have to know both how to enjoy everything that GenAI can offer, and at the same time how to do it right and hedge the risks.
“Development using AI and automation is effectively like a production line; without quality control sitting on that production line, we will produce much more garbage, much faster. We need to take advantage of the tremendous opportunity offered by the GenAI revolution while hedging its risks.”
“Putting the spotlight on DevOps work for a moment, we can break it down into three stages: the business process and collecting the requirements, the development itself, including testing, and finally, operations. Each of these stages can be significantly improved with the help of GenAI tools. For example, we can transfer to GenAI the process of breaking down the requirements document into development tasks, including the development of the test cases. A machine will do it much faster, even if we need the human factor afterwards to check that everything is in order. In general, if there are recurring faults within the CI/CD process, automated tools can identify the fault and fix it on their own. Nevertheless, a large part of the work is with the business system, and requires an understanding of complex, relatively unique situations. Here the human factor still has the advantage.
“GenAI technology will strengthen DevOps capabilities, and accelerate development processes. But, it is important to make sure that speed does not come at the expense of quality. There have always been two basic principles that guide us in implementing DevOps, and now even more so: Quality at Speed – we are fast, but not at the expense of quality. As we have seen from the research, in the case of artificial intelligence there is a higher risk that speed will come at the expense of quality, so control processes are especially important. The second mantra is Stop the Line – if development using AI and automation gives us a production line without built-in quality control, then we will produce much more garbage, much faster. Therefore, we integrate all types of testing within the process. We ‘stop the machine’, and don’t let it transmit low-quality and/ or unsafe code, etc. In this way, when we accompany organizations on their journey to implement GenAI technology, we allow them to benefit from the tremendous opportunity presented by this revolution, while hedging its risks.”
“The entire development market is in the process of upskilling and reskilling, and with appropriate training in the use of technology, a generation of ‘super-developers’ will emerge.”
Leonid Pevzner, a senior lecturer at John Bryce, agrees and adds: “All our students learn how to work with Copilot. But it’s not enough to know how to work with the tool. Let’s say it wrote me a line of code, and it even passed a test – is it secure? Is it effective? Exclusive use of a tool does not give you the ability to properly understand and improve what it wrote. So, on the one hand, it is true that the entire field of software is undergoing dramatic change these days – and is expected to continue to do so – and there are skills, such as problem solving, writing specifications and code review, that are moving to the forefront. On the other hand, in order to use these tools properly and to make use of all their capabilities, you need to have a base of knowledge, and a deep understanding of the subject. These tools enhance the abilities of developers, but they are not enough by themselves. Young developers should be given the tools to use technology critically and intelligently.
“The role of the developer will change from that of a coder, to a DJ of automation and AI solutions.”
“The impact of the GenAI revolution doesn’t only concern juniors, but also experienced developers, who have to adopt new habits for the correct use of artificial intelligence. In fact, the entire development market is in the process of upskilling and reskilling, and with appropriate training in the use of the GenAI technology, a generation of ‘super-developers’ will emerge. In the future, we will also see a more dramatic change in development work. Developers will have to learn how to work with GenAI models, how to adapt models that have already been trained to carry out specific tasks, how to optimize GenAI applications, how to design applications that involve collaboration between human developers and GenAI, and so on. There are areas within the field of development that are expected to disappear and others that will take their place. The acceleration of GenAI is so fast, and the number of solutions already available today is enormous. You need to know how to make a selection and adapt the right solution for every situation. In terms of the developer’s work, it will free up more time for innovation and creativity, and perhaps this role will change from a coder to a DJ of automation and AI solutions.”
