Will my Scientist job be replaced by a bot?

I read an unsettling tweet the other day. Yes, many boring conversations start with that sentence, but hear me out. The tweeter suggested that in the near future, all jobs will be better performed by artificial intelligence, or “AI”. This is part of the promise and the fear of AI–that we will all be replaced by bots. And yes, it’s possible that the tweet itself came from a bot, and not a human, but it started me down a path of wondering: How close is my job to being replaced by AI?

Like the printing press, the cotton gin, and that device that converts goop into hotdogs, machines have for many years been automating jobs once performed by humans. Even the term “computer” used to refer to an acutal human person who did mathematical calculations with pencil and paper. Now with the fancy electronic kind of computers and advanced AI algorithms, many jobs once considered safe from automation—doctor, lawyer, scientist—could be up for grabs by quasi-smart computer programs. So how about my job?

To answer my question, we first have to deal with the prickly question of what exactly I do. Don’t mention this to my boss, but I’m not really sure just what my job, “Scientist,” really is. Some days I’m mostly answering emails, and other days I’m mostly on zoom calls. But I guess I can say, as someone trying to run an Ocean Forecasting Center, that one of my job tasks is to forecast the ocean. Certainly AI is one of the tools that I use to predict things like whales or toxic red tides. So AI is already in the mix. But could AI do the job on its own? Could I be removed from the picture altogether?

Inspired by the amazing AI Weirdness blog by Janelle Shane, I thought I would try to answer my question using the completely scientifically rigorous process of throwing it at one of the new AI tools and seeing what would happen. I turned to an AI toy named Dall-E. This is an algorithm that can take any text into its search bar, and it will try to come up with an image that depicts what is in the text. It works pretty well if you type in something like “velociraptor holding a light saber”. But what about ocean forecasting?

My first test was to see if it could predict the future ocean under climate change. I used a species of plankton called Calanus—a tiny crustacean that is the main food for many whales and fish, and that also has global importance. Knowing their future distributions in the ocean would be a nice feather in the cap of any AI or bot aspiring to greatness. So I typed into Dall-E’s search bar: “a map of Calanus concentrations in the North Atlantic under future climate scenarios.” It took a few runs, but eventually Dall-E came up with something eerily reasonable. Here’s the map.

Not only is it a decent map of the ocean, but it’s a plausible prediction. There’s definitely a focus on the North Atlantic, but Dall-E went big and swung for the fences. You can see blobs of what seem to be Calanus in the Southern Ocean near Antarctica, up into the melting Arctic Ocean, and along coastal upwelling zones around the world. If you compare that to some other maps we’ve come up with over the years (for example here), it’s not too shabby. Okay, you say, but the map also seems to show parts of Europe, Russia, and Canada submerged by the ocean, and Madagascar seems to have inverted or something. Well, maybe it’s a quirk in the program. Or … maybe Dall-E has worked out something to do with sea level rise that we don’t know yet. Maybe Dall-E is onto something ominous and perilous that mainstream scientists have missed. I’ll have to make a quick phone call to NASA.

Meanwhile, if I have you convinced, then the next map will really freak you out. I asked Dall-E for “a map of white shark distributions in the Gulf of Maine under future climate scenarios.” Based on the map here, it looks like a major distributional shift onto land, and a rapid expansion across North America. Truly an apocalyptic scenario (and incidentally one predicted by Saturday Night Live many decades ago with their Land Shark sketch).

After generating these two maps, I spent an embarrassing amount of time exploring Dall-E’s ability to create ocean maps and other strange dreamlike images. It’s entertaining (at least for a nerd like myself), but of course it’s not how you would really predict the ocean, even with AI. And to be fair, the concerns over AI and ocean forecasting go well beyond that of job replacement, including issues of justice, accountability, and power, which I and others have written about elsewhere (e.g. here, here, here). But who knows—with a few new breakthroughs, maybe the next generation of AI bots will be out-forecasting the best of us. Any predictions on when will that future become a reality? Well, let me just ask Dall-E.