Are Technologists Too Locked Into The Status Quo?

In 1903 the New York Times argued that human flight would take one to ten million years to develop.¹ Just 69 days later, the brothers Wright took it to the skies. Well, maybe not skies, since they only covered 36 meters, but the direction was clear. It took 86 years from the bicycle to the aeroplane, and it only took 66 years for humanity from reaching 3 meters of altitude to landing on the moon. Since the moon landing in 1969, long-distance communication became cheap and effortless, smallpox has been fully eradicated, and other diseases that would’ve been death sentences are merely inconveniences these days.

The non-stop media cycle and its profiteers want us to believe that life is terrible and doom is imminent, but sticking to the facts, life is generally more comfortable than ever before. Individual tragedies and temporary setbacks are no good reason for pessimism.

Change is happening at unprecedented speeds. But what would you expect? More people are collaborating than ever before, and have access to more education and resources than they ever had. This makes progress not only easier, but makes profiting from it more accessible. Nobody has to seriously wait for a patent application to go through. You can start building and selling in many parts of the world, and access to funding is broader than ever.

But I think that experts are missing out on progress that is happening elsewhere. Domain experts who work in niche areas are overtaken by people who didn’t even know that the field existed a year earlier, but have the hustle and business acumen to move forward. They do not know which obstacles lie ahead; they jump in and solve the problems as they go. Especially in Germany, there is a culture of tradition and heritage: Experts are experts because they have been doing it for a long time. Of course, that lasts only until somebody new comes along, does it better and cheaper and faster, and destroys the incumbent.

If you asked a robotics researcher in the 2000s how to improve self-driving cars, they would answer with a ton of reasonable ideas: More computer vision, specialised detection, custom hardware, more money, more hands on deck. Executing on it would yield respectable improvements on the technology. But it’s becoming increasingly clear that end-to-end neural networks will be much more adaptable and reliable than hand-coded rulesets. The ability of LLMs to translate between languages more precisely than ever before does not come from a thorough understanding of language by experts, but from throwing stuff against the wall and hoping a pattern will emerge.

Worse than Experts: Expert Regulators

While experts struggle to have a vision, people in charge of regulation are able to waste even more resources: European regulation introduced mandatory e-invoices, and decades of discussions and negotiations yielded the standard EN 16931. As if that wasn’t extensive and complicated enough, EU members started gold plating it and introduced national flavours of it. Thanks to this development, which took the finest bureaucrats more than a decade to overcomplicate the process of saying “you owe me a tenner, pal” the implementations are so broken that the most effective way of parsing a standardised machine-readable invoice is to actually send the PDF version through a vision LLM pipeline instead of trying to parse the XML. Your tax money at work.

Plainly spoken: Their expertise has found a seemingly optimal solution for a future that ceased to exist in the long span of the development cycle. The smart thing would be to bury that endeavor and move on.

I don’t wanna knock long-term incremental improvement and general expertise. An A380 has needed a ton of elbow grease after the Wright brothers “flew” their scaffolding. Personally, I love to optimise existing machinery: Fiddling with a system on a workbench to get it to run a little bit faster, a little bit smoother for a little bit less money is a fantastic feeling, if you’re anything like me. And sometimes efficiency unlocks new use cases that wouldn’t have been feasible before, but the big breakthroughs will happen if people are a little bit ignorant or contrarian.

Particularly the idea of having two teams competing inside a company does fascinate me. I have definitely seen it going wrong, but I still believe that one team should work on incremental improvements and polish the product and they will have to compete with a team that will try to build the next generation. Both get the resources that they need, but both need to ship. Shape up instead of scrum and sprints. And don’t give up too early.

People will self-select: The moonshotters will be more of the free-styling researching kind, the optimisers will be more structured and incremental. They will both discover new aspects of the technology and have their own breakthroughs, but if you organise an intellectual exchange through the power of the written word, both factions will profit. So much knowledge is currently being lost in the trenches of production: Obstacles and mistakes are seldom documented and worked on in the R&D departments, and are only retained in the heads of people who don’t have the time to fix them.

Alternatively, if you’re not in a position to have parallel teams, give people a time allotment for new developments. I think there is so much potential by just giving somebody six uninterrupted weeks of building a thing. Constraints often make one faster and better: A successful six-week experiment that yields a prototype and is rewarded by a six-week polishing time frame will be more successful than twelve weeks of tinkering. Even companies which are stuck in quarters are salvageable in my opinion (my optimism does not seem to have any boundaries after all).

Behemoth companies are using acquisitions of start-ups on an unprecedented scale to offset the lack of internal innovation (see innovator’s dilemma), which is logical, but expensive. Smaller companies get squeezed on paper, but the reality is that the advantage that frontier companies have is smaller than ever, and is only held together by regulatory capture. No matter if it’s software or manufacturing, the methods that are available to small innovative companies are unbelievably good. If you don’t have to go through IT approval, submit a public tender or route your order through SAP, you are already in the lead.

And it’s not just for VC-funded businesses: Even small, self-funded endeavours have access to all the same tools that the giants have. Not having to develop the state-of-the-art LLM, but being able to use it for pennies is a massive advantage.

Experts are of course still needed, but those who can think outside the lines and be creative and adaptable are the ones who design the future.