In the 80s, probably influenced by the film Back to the future, flying cars were considered the epitome of technology and a certain future for us post-2000s. Technology breakthroughs did come and some were truly impressive. Most were improving existing technologies, but we also had a fair number of zero-to-ones.
The Internet was one of those zero-to-ones and one of the most promising technologies of the 90s. We started building on it, and what amazing advancements followed: sophisticated databases, lightning fast retrieval, web frameworks, a wealth of protocols, most of which were harmoniously orchestrated to deliver results. Back then, late 90s and most of the 00s, the Internet was honest, awkward, romantic, and obviously incomplete.
Today, about 20 years since then, we can imagine countless could-have-been futures for this project, but in this universe and reality track, we ended up with what we all know today as the Internet: ads, more ads, systems where you generate data in exchange for dopamine which are then showing you ads based on your data. Most of the internet is monetized with ads. In fact, some companies started with a clear useful product, only to realize that they had no idea how to make money out of it and finally resorted to the ads solution. We basically stopped watching TV, a major “I give you pleasure and you watch my ads” box and started “Scrolling and watching ads”. To create these systems, some of the world’s greatest minds worked hard. It goes without saying that there are good parts too but where the majority of users spends their time, is not on those.
An improvement of an existing technology, the Neural Nets, were the Large Language Models. What everyone today calls “LLM“, “GenAI“ or simply “AI”. A few years ago, they took most people by surprise and today they are being marketed as a lot of things including “company keepers“, “code writers” and “singularity bringers“. As an engineer I was both excited and concerned with their engineering skills which are remarkable. I’ve used them to get assistance with jumping into vast codebases very quickly, generating unit tests and reviewing code among others. We’ve heard that this technology will bring the next industrial revolution, which if true, would be amazing to experience it first hand! Along with the next industrial revolution we’ve heard that unemployment follows for specific professionals, such as Software Engineers and all their sub-professions. We've even heard that Software Engineering is going to cease to exist, or shrink significantly. The marketing element of this statement is quite clear: we give your AI → you fire your employees and AI does everything → happy days. I strongly believe that this is a politician-style logical fallacy covered as a reasonable fact. However I am mostly concerned with what will actually happen.
“LLMs”, or “GenAI”, or simply “AI”, in the scope of Software Engineering can make an engineer either faster by doing one task more quickly or pseudo-parallelizing multiple tasks, or make the engineer go deeper and therefore innovate or go to places where they wouldn’t go. From my experience they do the former better but can also help with the latter.
Has anyone seen any flying cars yet? Because I haven’t. Where is the tech effort going and why do we need better, faster, stronger engineers and engineering principles if it’s getting into new ways of doing ads? How are LLMs going to help if there is a throughput ceiling? Why aren’t we discussing about raising the ceiling and doing more things? Is that because raising the ceiling means innovation and LLMs aren’t good at it? Does that mean that mainly what is being sold is velocity with the existing SaaS / Ads / CRUD apps paradigm? That's not a dismissal of those products, but it is what it is, and honestly it's not that sexy.
We all need to understand what is possible, put the hype into perspective and reclaim our value as engineers. The people selling overfed NNs are doing their job, we should focus on doing ours. We are complete, diverse professionals, capable of innovating, going places and not waiting at any monopoly clique to lead the way. Similarly to how we have used Search Engines, tech Q&A websites and other technologies, we may or may not use LLMs, as a tool. Whether we are going to use them as an Iron Man suit or a background assistant, and what we are going to build with them, is entirely up to us. Their arrival hasn’t given us flying cars yet and honestly the world still looks the same. Let’s use our tools to scale up and out. Let’s raise the ceiling. Let’s build those flying cars.