Article
5 JUN 2023

GPT4 – Efficiency and Quality

San Jose, California, 2001. My then-colleague and I had acclimatized to a life in a Mediterranean climate and technology binge (although it was mostly a collapsing telecom and e-commerce bubble). It was almost harder to get used to the idea that you could go from blazing sun and steaming hot asphalt into an ice rink to watch NHL hockey. The real takeaway during the game, with all due respect, was not the efforts of Marcus Ragnarsson and Niklas Sundström, but conversations with American salespeople at Swedish software companies with local representation. One exchange is forever etched in my memory:

Erik: ”So, how come you chose to work for this tiny Swedish company?

Bob: “Scandinavians write clean code!

The idea that we in the Nordics, specifically those who have chosen programming as their calling, would fundamentally differ from their global, in this case primarily American, colleagues had not occurred to any of us. Initially, we dismissed it as something a flattering American salesperson would concoct on the fly. We had realized that smaller Nordic companies have some advantages that made it possible to compete with much larger American competitors: agile, egalitarian organizations can, if done right, run circles around monolithic bureaucracies with authoritarian leadership. The guerrillas against the militarily formed pyramid.

But could the craft itself really be that different? When we tested the question in various places, however, many seemed to agree. Scandinavians write clean code. Whether it was due to schooling, temperament, professional pride, or role models, we have no idea. But what does it look like today, more than 20 years later? I felt I had to check with our programmer Hannes:

Erik:       ”In your experience, do you think Swedes write 'clean' code? If we define clean as simple, efficient, and straightforward.

Hannes: ”Undoubtedly, everyone is good at different things, but developers from the Nordic countries tend to think twice before they send the code into production, which usually means a bit longer startup time but fewer problems down the line.

Erik:       ”Fun! It can perhaps be seen as a local dialect passed from one generation of coders to another. But what about GPT4? When you replaced yourself with an AI, maybe you gained unique insights. Does it also write clean code?

Hannes: ”In practice, autonomous robots are only as good as the instructions they are given. It can be compared to having a junior developer whom you train. You still need to 'peer review' and correct a robot that programs, just like a person. But if you understand how to most effectively give it 'feedback,' you can obviously make it behave exactly as you want.

It's all about finding a good balance between human and machine. The increased quality and efficiency don't come from the algorithm itself but from the right instructions. A flesh-and-blood developer can be relieved from repetitive work. Thus, the developer can focus on things that are more value-creating. More creative. Work can become more enjoyable, while achieving so much more.

We return, therefore, to the concept of intelligence amplification or IA. GPT-4 (and likely its successors GPT-5, 6, etc.) are tools that can make us smarter and more productive rather than replace all 'white-collar jobs' in the near future. And if there is a clear, unique Scandinavian style in how code is written, it may be spread across the world through these tools. This is at the same time possibly what can hold back companies from using GPT in their core processes. Everything uploaded into ChatGPT will indeed be used to continue teaching GPT.

Companies must protect their IP better than that. Similarly, companies must think twice before using AI-generated material in their services and products. The material could be based on copyrighted sources. If there is any way to regulate GPT and other large language models – and here we believe the likelihood is high – authorities must consider so many aspects that it becomes almost insurmountable. Maybe they can use AI for help?

 

Erik Sprinchorn

Share

Risk information
Historical returns are no guarantee of future returns. The money invested in the fund can both increase and decrease in value and it is not certain that an investor will get back all the invested capital. Please read Fact Sheets (PRIIP) and prospectuses available on our website or contact a distributor.
Edit cookie settings