Will AI lead to more unemployed developers or more software production?

Let’s assume that we manage to improve productivity in software development by an order of magnitude – that is, by a factor of ten. To make this very concrete, a thought experiment: A project manager wakes up in the morning and knows that his project only needs 10 people instead of 100. She also knows who these people are. It is as certain as the sky is blue.

Advertisement

Samsung today announced its first device designed for Android XR

What will the project manager do now?

The first option would be to continue the project with ten people and give other tasks to the remaining 90 people or fire them. This option is difficult to implement because the project manager is admitting to himself and others that too many people have worked on the project. Additionally, managing a 100-person project is more prestigious than managing a ten-person project. There is a risk in this approach because even if you find something useful for people, it is not so easy to bring them back.

Another option is to let ten people work on the original project and give the rest a different task. To do this the scope of the project can be increased. This may be relatively easy to do, as the desire for more features is more likely to become the norm. Or find another project for you guys. In the end, with both options there is more value to the organization and therefore more reputation for everyone involved.

It may be possible to accelerate the project with additional people. But this seems difficult because more people can actually slow down a project because training and more communication is required. Fred Brooks points this out in his book “The Mythical Man Month”.

Realistic?

But is such a thought experiment even realistic? Fred Brooks has an answer here too: in his paper “No Silver Bullet” He claims that no single measure alone can improve productivity in software development by orders of magnitude. However, this leaves room for a combination of measures to achieve this goal. And furthermore – like this blog post – it is also a hypothesis that provides no further evidence.

One reason why the scenario is probably realistic: As already mentioned, more people in a project means more prestige and hence the desire to complete projects with as many people as possible. Parkinson’s Law now states that all available people will also work on the project. Since software development involves a lot of communication, it can be difficult to communicate with such a large number of people. Since communication is also reflected in architecture, architecture also collapses. The relationship between communication and architecture is linked to Conway and his law. He also put forward the thesis of bloated projects with poor communication and ultimately bad architecture, which was already the subject of another blog post.

But then with a cleaner architecture, less people and therefore less communication and communication problems, you can get results just as quickly, so the thought experiment may not be completely unrealistic after all.

artificial intelligence?

However, the catalyst for the thought experiment is another development: it may be that artificial intelligence will make us significantly more efficient at creating code, as discussed in another blog post. Then the question that is very relevant is what happens when productivity improves by 10 times.

The thought experiment shows that even with such progress, more software will be produced and, as a result, the software will be used even in areas for which it is not yet suitable.

In fact, there is a case of this phenomenon in economics: rebound effectFor example, if cars become more efficient, they will be used for more trips, so that in the end consumption will not decrease, but even increase. Perhaps AI will create a similar effect: software development will become more efficient, but the software will also be used for other purposes, so that in the end the effort invested will remain the same. In fact, software supports more and more fields and AI may strengthen this trend.

In extreme cases, software may be developed by people who actually lack technical skills. This promise has been made many times, with technologies like COBOL, but also with less or no code. But even if AI is successful here, other industries show the effects of such disruptions. With desktop publishing, many more people can create print products now than in the 1980s, but the quality has deteriorated and professionals still have their own sphere of activity.

Now it may seem as if the future of the software industry is also secured by AI. But of course it is difficult to predict the future. One could also argue that the current crisis in the IT market is just an indication of what AI will do. The future is open.

tl;dr

Even a tenfold increase in productivity in software development, as AI could bring, does not mean that fewer people will work in the field, but rather that software will be used for even more purposes.

(rme)