AI robots can't clean our plastic-plagued oceans alone

Seemingly overnight, we've ended up coexisting with artificial intelligence. It's making spoof photos of the pope, scaring the stock market with fake explosions and helping us with our emotional issues. But some are wondering whether there are better uses for the technology:

Though intended to be funny, that tweet raises an interesting question: Could AI power robots to pick plastic out of the oceans? Or help resolve the other pressing problems such as climate change?

While large language models — the basis of tools such as ChatGPT — have taken the limelight lately, AI algorithms are already being used in the fight against climate change, biodiversity decline and pollution. "Most AI isn't flashy," says David Rolnick, co-founder and chair of nonprofit Climate Change AI. He detailed five uses of the technology in this arena:

• Distilling large data sets into usable information, including scanning satellite imagery for evidence of deforestation.
• Improving forecasts, such as predicting energy demand and renewable supply for electricity grids.
• Optimizing complicated systems, to reduce the energy required to heat and cool buildings or make industrial processes more efficient.
• Accelerating climate modeling.
• Speeding up scientific discoveries, like suggesting better battery materials to hasten experimentation.

AI algorithms are already being widely used, including in — yes — ocean-cleaning efforts (there are even robots). But while AI has made the process efficient and autonomous, there are limitations.

The Ocean Cleanup project is probably one of the best-funded and well-known marine-plastic undertakings. It's developed an AI tool to detect and map plastic objects at sea, to better deploy cleanup resources. Meanwhile, Hong Kong-based startup Open Ocean Engineering has developed Clearbot, a little solar-powered robot which can collect trash and clean up oil spills from urban waterways. Capable of picking up to 200 kilograms (441 pounds) of debris per mission, it uses AI to record and categorize the waste it's collected.

But even if the algorithms work perfectly, how you choose to use them matters — and there are plenty of well-documented issues with these projects. The Ocean Cleanup, for example, has simply reinvented trawl fishing, but for plastic. That comes with risks for ocean life and biodiversity — the very thing it's trying to save. By its own estimates, tens of thousands of small sea creatures such as crustaceans, fish, jellyfish and squid could get caught in the nets even when the system is used at its slowest speed. During the first 12 trips of its trawlnet system to the Great Pacific Garbage Patch, The Ocean Cleanup caught 193,832 kg of plastic along with 667 kg of so-called bycatch, consisting mostly of fish, sharks, mollusks and sea turtles. While that's a lot more plastic than marine life, there's a cost-benefit analysis to consider. What's more, the huge nets are towed by diesel-powered ships, making the process extremely carbon intensive.

An even bigger issue is that these efforts are barely making a dent in the problem. At least 14 million tons of plastic end up in the ocean every year. At the current rate, it's predicted that plastic will outweigh fish by 2050. According to its data dashboard, The Ocean Cleanup has so far caught about 3,300 tons of the stuff.

Ultimately, a global legally binding treaty, like the one being hashed out in Paris this week, will make the biggest difference to the plastic problem by tackling it at source rather than remedying the symptoms. "The hardest thing often in technology is listening to what's needed and building what's needed, rather than what you think is needed," says Rolnick. "It can't be technology coming in and saving the day. It has to be a combination of people with technological tools, people with on the ground expertise and communities who are affected by the technology."

AI is helping elsewhere. National Grid ESO, Britain's electricity system operator, is using AI to double the accuracy of its electricity demand forecasts, enabling better integration of renewable energy. Rolnick has been involved in creating tools for automated insect sensors to help accelerate and expand the collection of biodiversity data around the world. The payoff has already been huge: In Panama, the system helped entomologists identify 100 species that were new to science.

The moral of the story is that artificial intelligence isn't going to magically fix our problems, and the futuristic option isn't always the most effective choice. But, used intelligently and sensitively, machine learning can be harnessed to bolster people power in the battle to save the planet.