The coming change in society

I've been using LLMs to help write code since before ChatGPT existed, and otherwise have 30 years of traditional experience as a lead engineer. My AI driven development processes have been improving regularly over the past few thousand hours, and most of my production work is now entirely automated. It would be much harder to make my systems work if I had to coordinate with a team of human devs, so my work has basically been reduced to a solo position.

It's not a popular thing to say, but the problem most shops have is that people already aren't as capable as LLMs at engineering solutions to well known tasks (AI systems aren't as good yet at novel science, but that's changing quickly, and most development work is iterative - it builds on well know pieces which most often simply need to be properly orchestrated). Trying to get human involvement coordinated properly in the mix of organizational software development workflows, is what complicates the process.

My position won't be needed much longer for any engineering effort, although I expect I'll still get work because many humans will prefer to interact with a human project manager. But within a few years, the entire industry will not employ people at all like it did.

The first wave of layoffs was much too early - managers didn't understand that the LLMs were not yet good enough and that all the surrounding infrastructure was not yet mature enough to replace human work, but that situation is changing fast. The software development industry is inevitably going the way of the horse-drawn carriage industry in 1908.

But that's just the beginning. Understand the bigger perspective: within a few years, the world economy will no longer function as it does now, because it will be possible to entirely replace humans with physical robots which can do everything humans can, only, much more effectively, and for much less money. It will take years to manufacture that workforce, but that inflection point is coming. As soon as it becomes cheaper to replace humans with machines, people will no longer be employed in the same way.

When that turning point arrives, robots will quickly replace humans in most work roles. That transition will be fairly easy to implement. It won't require re-forming all of Earth's infrastructure, because the robots will be manufactured in the shape of humans, so they'll be a drop-in replacement wherever humans are needed. Of course, humans will want human interfaces to the machines, but most of the hard work will be accomplished by metal workers. That change is coming, just as surely as cars replaced horse-drawn carriages in the early 20th century.

We are in the midst of the biggest reshaping of human society ever in history. Of course the software industry is a mess right now, and of course it's confusing - it's just one of the first industries to be reformed. Most humans still think they're more intelligent and capable than AI systems, but that is no longer true, for many tasks that matter. AI isn't the same as humans, and it's probably not yet sentient, but I expect it will be, at least in every way that matters for productive work, very soon.

Hold on to your seats, this whole situation is going to move much more quickly, and the changes will be far more profound. I expect war and suffering during that transition, because people will need to adjust to a change of function, power, and purpose, but if humanity survives, a utopia will be possible. I'm not sure we'll survive...

I've been writing about software development and tech tools for more than 30 years. One of my latest posts in a thread about AI which I started in 2023, on a forum I started 20 years ago (about the Rebol programming language), to a developer who has a different point of view than me, about effective use of technology, was:

I should clarify a bit: it's not that I think your point of view about optimization is at all wrong. My position is that the world is changing dramatically. The most effective solution to achieving any engineering effort is to complete as much work as possible using AI tools.

There is no reason to write 100,000 lines of code, which can be written better and faster by an LLM. That doesn't mean I think human ingenuity is useless. In fact, it means the opposite. The tools which enable our ingenuity are improving fantastically. Refusing to acknowledge the usefulness and capability of those tools is fruitless. The goal of building optimized foundational (low level) tools is certainly a tremendously worthwhile effort, because the AIs can work even more efficiently with better foundational tools.

I do think it's ridiculous to avoid using LLMs to build more efficient foundational tools. And I think avoiding tools along the way that are currently available, and which can enable that process, is pointless. If Node.js needs to be installed on a server, where it takes up a fraction of a percent of the resources available to that machine, and that entire machine costs much less than a cup of coffee to run for a month, and using Node.js to run an AI agent which enables the creation of a foundationally useful piece of tech, it would just be stupid to not use that Node.js agent to create the new foundational technology. Avoiding the use of that tool on the principle that it's a fail because it requires some inefficient use of memory along the way, as a stepping stone to better tools, is just counter-productive. The goal is to create the better foundational tool in a way which is most efficient overall. Wasting your own human effort to achieve a higher goal is far worse than using a less than perfect tool such as Node.js + some LLM and agent along the way.

My work has never been about building better low level tools. I've always been interested in the application of technology, in real-world operations, in businesses and my own life activities. I've always been interested in leveraging my own time and effort to improve quality of life.

I've always searched for and chosen to use tech tools which enable productivity. Rebol was one of the best of those tools at the beginning of this century, but we're 26 years into the 2000s, directly in the midst of the biggest transformation to human society and to the role humans play in the surrounding universe. That transformation is inevitable, and we're just experiencing the very first few inches of a million mile journey.

To me that journey is the most exciting experience I could ever imagine being a part of. Others refuse to accept or understand that it's even happening. It doesn't matter what I say about what's happening. The reality is, it's happening, and it's not going to stop.

If you like horse drawn carriages, by all means, continue to make them - they're beautiful - but please don't expect the rest of the world to not build cars, trucks, airplanes, and everything else which is enabled by the creation of those machines - including, eventually, much more efficient, effective, and less destructive cars, trucks, airplanes and other downstream tech.

So, if you're interested in making a positive difference, use the LLMs, and whatever better tools come after, to improve the foundational things you care about. Your perspective about creating optimized tools is incredibly valuable. Marry it with a mastery of the most powerful AI systems we have available, and your productivity in achieving those goals which are most meaningful to you, can be improved 1000x.

https://rebolforum.com/topic/819