Various applications of so-called artificial intelligence generation could create almost 1,000 times more e-waste by the beginning of the next decade than it is today, a global study published on October 28 in the magazine Nature Computational Science.
In terms of total weight, this would correspond to a situation where every person in the world (with an estimated global population of about 8.5 billion people in 2030) would throw away almost two iPhones.
Generative artificial intelligence is a technology that creates new content based on existing data and information. Unlike traditional systems that only analyze and classify data, AI can generate text, images, audio, video, or other forms of content that appear to be human-made. Examples include advanced language models such as GPT (Generative Pre-trained Transformer) or image generators that can create realistic pictures or images. |
However, e-waste, associated with the disposal of “end-of-life” hardware and chip technology products, typically contains toxic metals including lead, mercury, cadmium and chromium, which is dangerous to health and the environment, as well as expensive. metals such as gold, silver, palladium and platinum, or rare earth metals that could and should be recycled.
The bet on AI is starting to pay off for tech giants
AI
“Our findings indicate that this waste stream may increase, potentially reaching a total accumulation of 1.2 to 5 million tons between 2020 and 2030,” said researchers from the Academy Chinese Sciences and Reichman University in Israel in the paper.
“However, this number could increase even more, for example if there will be geopolitical restrictions on the introduction of semi-communication products or due to a rapid turnover of servers motivated by operational cost savings,” they continued.
📢 Peng Wang and colleagues develop a framework to measure the impact of generational AI on e-waste, suggesting that nearly 1,000 times more e-waste could be generated by 2030 and stress on the need for circular economy practices. https://t.co/K49tvu64eM
— Nature Computational Science (@NatComputSci) October 28, 2024
“A very material intensive sector”
In March 2024, the CEO of the American company Nvidia, which leads the market for artificial intelligence chips, saidthat the weights of the company’s graphics processing unit (GPU)-based AI platforms have increased more than 40 times in two years.
“Two years ago (for systems with 35,000 parts) it was, for example, 32 kg. But our current AI systems have 600,000 components and weigh 1,360 kg. That’s about the weight of a carbon fiber version of a Ferrari,” CEO Jensen Huang said of these special systems that are critical to artificial intelligence and can perform complex calculations quickly and efficiently. They are actually a kind of “miniature” supercomputers.
In the paper, the researchers also mentioned that the weight of Nvidia’s latest AI platform, designed to run large language models, learning and data processing, shows that artificial intelligence generation is a “very material intensive sector”.
“This rapid growth in the weight of hardware devices, driven by rapid advances in chip technology, could lead to a significant increase in e-waste and its environmental and health impacts at the end of the product’s life cycle -hard,” they said.
They found that more than half of the total flow of relevant AI waste would be located in North America – that is, the United States and Canada – about 58 percent, because there are large clusters of data centers running AI applications there. A quarter of the waste stream would be located in East Asia, especially China, South Korea and Japan, and 14 percent of the AI waste stream would be produced within the European Union and Great Britain.
The US is considering new export restrictions on chips from Nvidia and AMD. The target is not just China
Economic
US restrictions on the sale of advanced GPUs to China and other technical barriers could also harm the environment, as data centers in China “are now forced to use older server models, ” said experts based in China and Israel.
“This can lead to a loss of computing efficiency in countries that do not have access to chips like the US, leading to higher demand for physical servers. “
The researchers also compared Nvidia’s advanced H100 chip, which US regulators have banned from selling to Chinese customers over national security concerns, with its modified export version, the H800, which has about half of the transfer efficiency.
“Our analysis suggests that a one-year delay in getting the latest chips could lead to a 14 percent increase in end-of-life e-waste, totaling 5.7 million tons from 2023 to 2030, which is more than the global amount of small waste coming. from the field of information and communication technology in 2022,” the researchers say.
In 2022, the record number was officially realized worldwide 62 million tonnes of e-waste therefore. Only 22.3 percent of that was collected and recycled correctly – according to a global report from the UN. However, collection and recycling rates could drop as low as 20 percent by 2030 due to significant growth in global e-waste production.
Geospatial artificial intelligence will improve weather forecasting, nature conservation and spatial planning
Science and schools
However, many people in developing countries are also recycling e-waste unofficially and without meeting legal standards, according to the UN. These methods often release harmful substances such as acids, dioxins and furans. These self-help methods are not effective, they lead to waste of resources and high pollution of the environment. They also pose a health risk to workers and local communities.
To reduce e-waste production, the research team proposed a circular economy strategy (within the entire production chain) that could reduce its volume by around 16 to 86 percent throughout the world
This strategy also includes extending the life of AI-related hardware, reusing outdated components such as GPUs, CPUs and batteries, all in low-end applications . The hope is also to develop more efficient calculation algorithms and thus also (software) increase the calculation efficiency of the chips.
The battle between the chip giants threatens the future of mobile phones and electronics
Hardware
The study’s lead author Wang Peng, who studies the distribution and sustainability of material use at the Institute of Urban Environment of the Chinese Academy of Sciences in Xiamen, said that although the strategy would be adopted by all countries that ‘ development of artificial intelligence, that they could be different. areas of focus according to the level of progress achieved in this technology.
“For example, the US should focus on responsibility in the production of hardware and the development of algorithms, since it is at the forefront of the development of artificial intelligence. The global AI industry could be more sustainable if its resources were better managed,” he said.
“In the case of China, the emphasis here should be on strengthening the management of hardware operations and disposal processes. AI hardware is produced in large quantities, followed by frequent updates, removals and replacements. And all these activities related to waste management should be tightened,” he concluded.
The biggest chip maker increased profit by half thanks to AI
Economic
A new engine for artificial intelligence, Apple developed the heart of computers
Hardware
2024-11-08 07:51:00
#Millions #tonnes #ewaste #dark #side #artificial #intelligence #News