You know the story about how we killed science? Scratch that, killed isn’t the right word. We didn’t kill science, we lost it.
You see, I was there when it happened. I used to be a scientist, back when people still were.
It began with gorillas for me.
Maybe you’ve heard of them? We lost gorillas around the time we lost science. My research group was studying them in Virunga National Park. Or we were trying to. Gorillas lived in dense jungle, they were big, but quiet as ghosts when they wanted to be. They were a nightmare to track. We’d spend hours, sometimes days, locating them. We’d get maybe a week of observations, then they’d get spooked and vanish again.
It was frustrating, to put it mildly. We needed to learn as much as we could if we were going to save them, but we weren’t collecting data fast enough. There was a less noble concern too, we needed data to publish papers. If we didn’t publish we’d lose our funding, the research would end, and our careers would be as extinct as the gorillas.
One of my new postdocs suggested we try using autonomous drones to track them. I was sceptical, I didn’t think the technology was good enough, but she was so enthusiastic, and I was desperate, so I agreed.
I couldn’t have been more wrong about the drones, they were a revelation. We released clouds of them, small as your fingernail. They flew off into the forest without a sound. Within 24 hours they’d found and were observing every gorilla in Virunga.
They recorded everything. Every gesture, every interaction, every behaviour, what the gorillas ate, how long they slept, who mated with who, absolutely everything. We even adapted drones to take DNA samples from the gorillas as they slept.
Those drones helped us create the perfect dataset.
They gave us insights into gorilla behaviour and ecology that would have been impossible otherwise, all without us having to leave our camp. I was so happy and relieved.
The drones might have been the perfect field assistants, but they had one big problem. Before we were struggling to collect enough data, but now we were drowning in it. In one field season we recorded more than four million hours of video, it would have taken my group the rest of our lifetimes to analyse it.
I was lucky though, or so I thought. I managed to get more funding and I hired another postdoc. She wasn’t an ecologist though, she was an expert in machine learning, and she built us an algorithm to analyse our data.
I’d been so focussed on my research till this point I hadn’t appreciated what was happening, but now I realised something big had changed. The proverbial frog finally noticed the change in water temperature.
The next field season we went to basecamp but we had nothing to do.
The drones took care of themselves, if I needed to adjust their parameters I could do it remotely from my laptop. The analysis algorithm was running on a high-performance computer cluster in Norway. We didn’t need to be in Virunga. I sent my group home. I stayed on to fuss and oversee, but I knew I wasn’t needed there either.
We weren’t the first ecological research group to adopt automation, other colleagues leapt aboard that bandwagon, you had to. Science was a cutthroat business, you produced high-impact papers or your funding was diverted to those who could. You stayed ahead of your competition or you died. You had to use every advantage.
It wasn’t just ecologists either. Collecting and analysing data was tedious and routine in most areas of science. Humans got tired and made mistakes, robots and algorithms didn’t. Whether you were tracking animals, sequencing DNA, or counting supernovas in the night sky, automation was the only way to go.
Next came statistics.
If a scientist ever woke in the middle of the night in a cold sweat, chances were it was due to worry over statistics. You see statistics was at the heart of science, but few scientists were genuine statisticians. Few of us understood the stats techniques we used quite as well as we should.
It was a constant fear that gnawed at you. Were you using the best technique, were you making mistakes, would a rival group analyse your data differently and overturn years of your work?
But the next leap forwards in machine learning gave us another gift. We got algorithms to do the stats for us. My data flowed in through the drones, was analysed and prepared by one algorithm, then passed to the new stats wonder algorithm for it to work its magic.
It decided which statistical approaches were most relevant and applied them. None of us understood quite what it was doing, just that it was fast and trustworthy. It was a hell of a relief.
Updates were released and it quickly advanced to the point where we didn’t even need to define our research questions anymore. The algorithm digested our data, then told us what would be the most suitable questions. It was spooky.
So much had changed by this point, but none of us could prevent it.
The snowball was careening down the hill, getter bigger and faster and harder to stop.
I’d stopped employing new PhDs. I’d become an ecologist because I loved the fieldwork, but I hadn’t done any in years, I was the custodian of a team of AIs instead. I was producing incredible science, the kind of work I’d dreamed of, but I didn’t feel like it was mine anymore. I was still writing the papers myself at least, but that didn’t last long either.
That part of the jigsaw began with the reference software, a handful of programs most scientists had been using for years to manage papers.
First recommendation algorithms were introduced to suggest suitable new papers. But there was too much to keep on top of. Hundreds of papers were published in my field each month, trying to keep up-to-date was exhausting. There was another gap in the market for the reference software to fill.
It was hard to put your finger on quite when it happened, but these once simple programs eventually evolved into literature AIs. The software began to synthesise and summarise papers for us. They scanned and digested all the literature. They made connections in disparate areas of research that no human could make.
Eventually they began to propose new hypotheses and fruitful areas of research. It wasn’t long before they were writing the first drafts of the papers for us. By this this point I realised I was only a scientist in name.
The pace of change kept accelerating though.
In 2038 the first paper conceived and written solely by an AI was published. The press made a huge joke about it, called it AI PI. Then in 2041 Nature published its first entirely AI-written issue. That was the coup de grâce, the research reports, the opinion pieces, even the Futures short story, nothing had been touched by a human hand.
Different fields and journals changed at different rates of course, but the inevitable was inevitable.
For the younger generations it must seem odd that something as important as science was ever trusted to human hands, like medical care and music I guess. But it’s still hard for me to get my head around.
For centuries scientists used their talent and their passion to push the frontiers of human understanding, the Darwins and the Newtons and the Einsteins and all the others who have been forgotten now. All that genius and creativity, replaced by unthinking networks of silicon and code.
Think of that. All the big new ideas that shape the future of our civilisation, all now had without any form of conscious thought.
It breaks my heart, but it was the right thing to do.
You know what astrology is, right? Not astronomy, not the old science, astrology, star signs and predicting the future and all that bullshit? Well I used to think astrology and astronomy were so different, but now I’m not so sure.
Astrology was an attempt to try and make sense of the universe, to make it simple enough for the human mind to grasp. It was way off the mark, of course, reality proved too complex, if you wanted to understand the universe you needed science.
But maybe the human version of science was little better than astrology? It worked, it was predictive, it gave us insights astrology never could, but maybe it was still just an attempt to oversimplify the world, to bring the universe down to our level? Maybe the machine’s version of astronomy makes ours look like astrology?
I guess the AIs are the rightful inheritors of science. We still reap the benefits of the discoveries they hand down to us. We still get to read about science in the papers and the news stories they write for us. But I still miss those days out in the field with the gorillas.”