02 | 2024 Sanna Valkeejärvi & Tomas Günther, Solution Architect & Fullstack Developer
"Kipinä has a lot of experienced guys with a passion for problem solving. We're in the business of helping customers, and we genuinely want to understand what's most useful to the customer."
One of Kipinä's experts is Tomas Günther, who has 23 years of overall experience in the field and 2.5 years at Kipinä. He followed in his father's footsteps into the IT industry. At home, even as a child, there was a PC and later an Amiga 500, which were fun for playing games, drawing, making music, and experimenting with very basic programming. While getting games to work, Tomas learned a bit about computer hardware, software, and operating systems. In the late 90s, while in high school, he built his first company website.
"It was during high school that I started thinking more seriously about what I wanted to do in the future. In addition to technology, which presumably offered strong employment prospects, I was interested in languages, psychology, music, biology, and even advertising. When engineering students came to present their alma mater, their relaxed demeanor and stories about student life made a strong impression and dispelled my preconceptions about the field. Afterward, a friend and I visited the Helsinki University of Technology in person, and I haven't looked back since," Tomas explains regarding his career choice.
23 years of experience with many changes.
"There have been major shifts in software production methodologies. Back in university in the early 2000s, we were already taught to use XP and SCRUM, but in the professional world, the waterfall model and other rigid methodologies remained in use for quite a long time. Back then, a developer's role was often much more restricted than it is today. For instance, completely different people might write specifications or documentation for a project. And once a system was deployed, it was considered a success if no one had to touch it for a long time."
"Today, there is a much higher value placed on rapid responsiveness to an adaptable operating environment. In DevOps culture, organizations even optimize how quickly a new feature can be delivered to production from the moment the need arises. Additionally, practices like automated testing, code reviews, and building automated deployment pipelines have grown significantly compared to the early days of my career."
"Another positive change has been the increased appreciation of developers. Companies like Kipinä were hardly known to me, except perhaps in the startup world."
Tomas is one of the first experts on Kipinä. Picture from Kipinä Symposium 2021
From Chatbots to Language Models
"In the 2010s, clients began migrating from their own server rooms to the cloud. There was, and still is, a lot to learn in the cloud environment, but hands-on experience is ultimately the best teacher! Fortunately, many projects have provided the opportunity to see and implement a wide variety of solutions."
"In recent years, machine learning with predictive models and artificial intelligence capable of mimicking human capabilities have grown in importance. Large language models (LLMs) have been particularly prominent," Tomas continues.
Regarding large language models, Tomas sees more opportunities than threats, although by his own admission, he initially had a rather skeptical view of the usefulness of ChatGPT, for instance.
"Some 20 years ago, I had already played around with a language-learning chat program called MegaHAL, which creatively combined what it learned, and used it to build chat bots for IRC channels. The text generated by large language models is based on probabilities—essentially predicting the next word in the text. Because of my experience with MegaHAL, I knew how to phrase prompts in a certain way, which easily pushed ChatGPT's responses into the realm of absurd humor.
However, the GPT model has been trained and retrained extensively, and recently it has proven quite helpful for untangling various matters in professional contexts as well. You can even use ChatGPT as a one-person sounding board. And once you learn how to craft your prompts correctly (prompt engineering), you get much more useful responses. Lately, I've been experimenting with building my own GPTs and using the content of my own PDF documents as part of the model's knowledge base!"
"In software development, writing boilerplate code is becoming faster thanks to artificial intelligence. Furthermore, programming as such will likely diminish, but this does not mean humans won't be needed. Instead, humans will increasingly act like copy editors, guiding the machine."
AI featured strongly in Kipinä's Future workshop
Looking to the Future
Change is happening constantly. How does Tomas view the future of the field?
"I hope ethical questions are kept in mind when utilizing artificial intelligence and ML models to optimize the world. It is essential to understand how the data used to train a model affects its outputs, and to recognize that a model's response is a matter of probability, not absolute truth. It is important that data represents diverse human groups, for instance, to ensure that the models' responses are fair and unbiased. Naturally, we must also account for information security, as well as the safety and reliability of AI in applications like autonomous driving and medical diagnostics."
"The tools for leveraging machine learning and artificial intelligence are already at a high level, but users are required to understand how to apply them correctly. Model developers and the organizations deploying them bear significant responsibility, and the operation of these models must be monitored. This responsibility must be fully grasped, as it impacts the entire world. When we talk about credit or support decisions, for instance, they can be automated with AI, but we must also account for the possibility of erroneous decisions and prioritize decision-making fairness."
Green values and clean code are becoming increasingly relevant in Tomas's view. The efficiency of code and solutions should be valued more as part of high-quality software and sustainable development. However, sustainable development (a play on words in Finnish) doesn't necessarily mean slow development, as efficient solutions are often simpler to implement.
"Humans will still be needed in the future. For example, language models and cloud services cannot function without humans; they still need to be developed and maintained. A wide variety of talent will continue to be required. The significance of data is not going to diminish either; humans are needed to interpret it and make choices based on it. Perhaps. Deep learning will likely still have a place for human labor, or at least in human-centric and physical-world tasks."
Kipinä's Code Retreat was a coding race
Competence development in Kipinä
How do experienced experts ensure they keep pace with change?
"There are many ways. The best way to learn is by doing things in practice in projects, but there are other ways to learn. At Kipinä, we talk quite a lot about technology and how it's evolving. We've also just spent a day together talking about the hallmarks of good code and the code lifecycle. In addition, last year Kipinä experts participated in sales training and strategy sessions have been held. And we have jointly agreed that we will each have 60 hours per year to spend on skills development in the way we choose during working hours.
"Kipinä has experienced people who are eager to solve problems. We are helping our clients and their customers with our expertise, working in collaboration with the client. We strive to truly understand the client’s needs and then provide what will be most beneficial for them. At Kipinä, we don’t just sell a technical solution, but also insight, expertise, and the support of our network, all of which lead to results."