RE: @jameshowell@fediscience.org
This is your periodic reminder that we already know how to write large swathes of unhackable software. We have known how to do it for more than a decade, and we learn how to cover more kinds of software year by year. This includes, eg, control software for smart cars and autonomous drones.
The small print?
1. It is even more expensive than ordinary software.
2. Writing a new piece of unhackable software may take arbitrarily long time, even more than a new R&D piece of software.
3. Even if we wanted to, we don't have anywhere near the number of people globally to do it. The skills needed are highly specialised. A decade and a half ago, you needed a PhD in formal software verification. Today, some BA graduates of some universities have the right skills, but these are few and far between and it's not clear to me whether the trend is going up or down.
4. Even if we most universities did buy into teaching this skill set, we don't have enough postgraduates graduates to do so en mass.
5. Even if you did invest the huge amount of resources to produce unhackable software, the vast majority of the market does not value the added benefit at all. You've wasted your resources.
Imagine where we would be if the market did value unhackable software, universities did train graduates with the right skills en mass, and funding agencies did invest resources to overturn facts 3, 4 and 5. (Some do, but even for those, contrast the investment with, eg, generative AI and other new shiny capabilities.)
