[commons]
Being a buyer of stuff, everything seems expensive. The new MacBook? Nice, but would be nicer if it would cost 200 less. That car? Nice, but would be nicer if less expensive. Which is why we love discounts so much. But there is another side to the equation, namely that of how to actually produce the thing that it can be sold reasonably cheaply, which is to say at a competitive price.
I remember a friend of mine once dabbled in product development in the aquaponics space. After many months of work, the selling price they thought they could achieve for a small box was a couple of hundred Euros per piece, clearly more than the average interested person would be willing to pay. What I want to get at there is merely that it is difficult, and that there are two sides to the coin.
The same now goes for AI, the big hype du jour. One of the prevalent projections out there is that it will render many jobs obsolete in the near future, amongst them many or most programming jobs. Now that’s very interesting, but being on the other side of the equation—of having to produce software as an engineer—you now should expect me or some of my colleagues to be in possession of the magic tools to create software with the snap of a finger.
There recently has been a study, which has put the effectiveness of the use of AI in programming to a test. It found out that programmers who used AI thought they’d be 20% more productive, whereas it apparently turned out they were 20% less productive. Take from that what you will. I personally think learning to use the newly available tools takes a little time and requires one to develop new skills.
Overall my impression is that there are certain things of the job of an engineer you cannot easily automate away (right now), and other things, and I’m mostly thinking about programming tasks here, which can be sped up significantly. While I believe it is good that somebody put out that study—to ground everybody a bit—from personal experience I find it unwarranted to believe that the use of AI in programming is anything but a game-changer.
Even if from here there will be no acceleration, by AIs now creating better AIs, or any other relevant theoretical breakthrough (which is hard to believe given the tens or hundreds of thousands of people working on it worldwide), I consider just having these new tools available an unrealized breakthrough in the sense that it will still take considerable time for them to spread fully across the entire industry. At some point, again assuming stagnation for a moment, AI will be used as a default tool, one tool amongst many, for some problems, the ones for which it will haven proven useful for.
Actually this is what matches my experience. In the same way that many people now favour ChatGPT searches over Google searches, there are now certain programming tasks which I’m beginning to feel reluctant about doing by hand. And I am getting better to identify which precisely those are, while another part of the industry is working on improving the tooling from their side; and I’m calling it ‘tooling’ to emphasize that even without the models improving, re-arranging the bits we do have now into better packages will do the trick. So, better tools and more experience with those tools it is. I think we’ve hit a transformative moment.