A recent study by environmentalists at the Chinese Academy of Sciences in collaboration with Reichman University in Israel raises serious concerns about the environmental impact of generative artificial intelligence (AI).

According to the research, demand for the specialized hardware required to power applications such as ChatGPT is rapidly increasing, causing older electronic devices to become obsolete.

“Our findings suggest that this e-waste flow may increase, potentially reaching a total accumulation of 1.2 to 5 million tonnes between 2020 and 2030,” the researchers said. South China Morning Post reported.

The explosion of AI will create an estimated 2,600 tonnes of E-waste in 2023 alone. However, according to the research, if proactive measures are not taken, this number may increase to almost 2.5 million tons by 2030.

To put it in perspective, the amount of e-waste generated would be equivalent to the estimated 8.5 billion people on the planet throwing away almost two smartphones.

The cost of the AI ​​boom on the planet

The study highlights how the rapid development of artificial intelligence is making old equipment obsolete and how constant change is increasing the problem of e-waste, which poses serious risks due to the presence of toxic substances such as lead and chromium.

Meanwhile, although many of these components contain recyclable metals such as gold and platinum, current e-waste management procedures often need to catch up.

According to the research, the majority of e-waste will originate from North America, particularly the United States and Canada, where many AI data centers are operated.

The study claims that geopolitical constraints, such as the US ban on the sale of advanced GPUs to China, could also worsen the situation. Chinese data centers may start using outdated technology, which could increase inefficiency and increase e-waste backlog.

Such a strategy “could lead to a loss of computing efficiency in countries that do not have access to such chips, resulting in increased demand for physical servers,” the researchers said.

The research also compares NvidiaIt’s a modified export version of the H800 that has less ability to meet export laws to China than the state-of-the-art H100 chip, which U.S. regulators have banned from being sold to Chinese consumers over national security concerns.

“For example, the Nvidia H800’s bandwidth efficiency is half that of Nvidia. H100the number needs to double to achieve equivalent performance,” the researchers said.

“Our analysis shows that a one-year delay in obtaining the latest chips could lead to a 14 percent increase in end-of-service e-waste, which could reach a cumulative total of 5.7 million tonnes from 2023 to 2030, accounting for global small is higher than the amount of waste, information and communication technology waste in 2022.”

Customized e-waste strategy can help

To address this impending disaster, researchers support a circular economy strategy that emphasizes recycling and reuse of electronic components.

These tactics can reduce e-waste production by up to 86% worldwide.

Lead author Wang Peng underlined that various regions need to implement customized approaches to deal with their own problems.

“The United States should focus on being more responsible in hardware production and algorithm development because it is at the forefront of artificial intelligence development. The global AI industry can be more sustainable if the resource is managed well,” Peng said. SCMP.

“As for China, the focus will be on strengthening regulations on operation and disposal processes. AI hardware is produced in large quantities, followed by frequent updates, eliminations and replacements. Regulations regarding waste handling also need to be tightened.”

The study was first published on: Nature Computational Science.