5 ms·
Salaries will go down not because of less collective action (there was not any really to speak of with high salary people) but because now the (remote) applican
by ilaksh 6y ago
Salaries will go down not because of less collective action (there was not any really to speak of with high salary people) but because now the (remote) applicant pool has many people living in relatively low-cost areas, many even in countries/localities where the cost of living my be 70 or 80% less than the big US cities.
In such a situation, it is very easy to question the idea of hiring _anyone_ who lives in a big city and therefore requires a large salary.
As far as software eating software, it seems obvious to me that it is occurring, but may not be obvious to everyone because as that has been happening, we have also seen a rapid increase in software use in business and industry. And we have seen software projects become more mainstream, where in some places, almost every random person seems to think they need to build their own mobile app or web-based service, for example.
I think we are not yet close to the end of the level of automation of software development. Here is one thing that may be on the horizon: chat, voice, and UI-mockup - enabled software development bots that take advantage of large knowledge-bases of application templates and components, and use deep learning to smartly combine and configure them according to natural language instructions and intuitive UI interactions. This could also involve generating design assets with AI based on combining large libraries of starting points and styles.
I also personally believe that more advanced AI for programming is going to sneak up on everyone, including programmers, faster than most expect. Especially when you think of tackling specific types of software at a time. What the deep learning people are working on is making neural networks learn better-factored and more accurate models of the world. I believe this is probably feasible to do using mostly existing algorithms, by building up models (and networks) gradually with a progressive curriculum approach. Using that type of capability and very large datasets of specifications, design interactions, sample assets, templates, and final configurations, it will be possible to train AI to build many different types of software.