Progress is not inevitable

26 July 2025

It is no secret that the current hype machine for LLMs has caused a number of previously “sustainable” tech companies to abandon their commitments to a sustainable energy transition, in favour of burning as much methane as physically possible. While talking to an otherwise serious person about how devastating this envrionmental crime would be, they said casually “It’s alright of course, since fusion is just around the corner”. Given that nuclear fusion has been around the corner for several decades, I was completely taken aback by this. Didn’t he know? “Well,” he answered, unphased, “it’s just an engineering problem now. Look at Moore’s Law, or the vaccine for COVID. We’ll solve it, just give it a decade or so.”. Pressing him further, this optimism extended to all the major problems we face – climate change, antibiotic resistance, space travel, what have you.

On some level, it’s hard to deny this position. Human progress, especially in the last 200 years or so, really has been staggering, and often in ways that are hard to predict. The person I’m arguing with has worked with computers for their entire adult life, and has really bought into the idea that exponential growth is forever. Even when Moore’s Law actually ended we found other ways to “simulate” it, adding caches and increasing parallelism, and to a superficial observer this looks the same. GDP grows every year, the pool of humans we can feed constantly grows, and life expectancy always increases.

But I do think this is an error in thinking, and actually a very dangerous one. If we can’t distinguish a technology that’s 5 years away from one that’s 500 years away, we run the risk of making dangerous gambles with our future. I blame Elon Musk for this particular combination of simultaneously delusional optimism and casual pessimism – things that make him money (think self-driving cars, commodified space travel) are just around the corner – things that don’t (making Earth habitable, building trains) are probably impossible. The effect is that popular media treats terraforming Mars (probably impossible, certainly not happening within 1000 years) on a level footing with carbon-neutral high-speed rail (which could probably be done today if the political will existed).

It is not even obvious that progress, broadly defined, is inevitable. We could quite easily slow down, or even reverse course. To take an example at random from the news, life expectancy has been gradually growing for the past 150 years, but is now falling. Antibiotic resistance is growing, and we haven’t invented any reall new ones since the 1980s. We could easily run out within my lifetime. And many social progressions, like the reduction of death from war and the decline in violent crime, are at risk as the US backslides into fascism.

Even if we don’t fall backwards, uniform progression is the exception, not the norm. There are not many things in the world we can “just do better” or “just make bigger”. Most of the serious problems hit fundamental limitations, and genuine innovation is required in order to get past them. There is no guarantee that a particular technical barrier will be broken if we just assign enough people to it. We’ve known the escape velocity of the Earth for nearly 400 years, but we’ve only been able to reach it for 60. It took us only a few decades to go from seeing the atom to splitting it, but it took us thousands of years to see the atom in the first place.

I am not saying this to imply progress is inherently unmeasurable and we can never predict anything. I am only really saying that nature abhors progress, and the fire underneath it needs to be constantly stoked in order for it to continue. Prediction is hard, and there is no way around it than looking at each claim individually. Luckily, at least for now, we have an entire sector of the economy devoted to looking at facts and predicting the outcomes, and they generally don’t fall for this shit. In some cases, stable abstractions are available to the general public which make bullshit easy to detect.