The debate about the necessity for regulation of computing-intensive apps and the chips that power them is being sparked by worries about the environmental impact of AI.
The Stanford Institute for Human-Centered Artificial Intelligence and the Stanford Woods Institute for the Environment held the Advancing Technology for a Sustainable Planet Conference, where professionals discussed this topic.
Regulations have to be made for a green future
Technologies like cloud computing and artificial intelligence require energy, which increases carbon emission levels. In addition, many of these technologies also assist businesses in achieving their sustainability objectives. As a result, businesses must strike a balance between rapidly scaling up and embracing new technologies while being aware of how they may affect their overall environmental impact. But researchers are also working on solutions, and new tools are emerging to reduce the carbon footprint of AI.
According to Peter Henderson, a Stanford University Ph.D. student in computer science, the environmental impact of AI depends on the scale. He made this statement during a conference panel discussion. Before using a machine learning model, businesses frequently tune AI algorithms to address energy use and carbon emissions concerns.
“The point is to make sure we don’t scale to the point that is harmful to the environment when the goal of a lot of machine learning work is AI for social good, where we want to build more sustainable things, we want to optimize batteries, energy grids. But if all that optimization leads to more negative impact than positive, it’s not worth undertaking,” Henderson explained.
Beyond industry attempts to optimize such technology, according to Henderson, government initiatives to establish guidelines for their use will probably also be required. While there are already certain regulatory initiatives in the European Union, they don’t adequately address how AI may affect the environment.
How authorities are targeting the environmental impact of AI?
Henderson claims that the EU’s AI policies prioritize protecting consumers over the environment. Over two months ago, we discussed the EU AI Act in detail. The second quarter of 2022 is happening to be an important period for the regulation of artificial intelligence. The UK prepared its own AI rulebook with similar concerns.
He continued by saying that the GPUs used to run AI and machine learning models are a major contributor to the environmental impact of AI. To address worries about the environmental effects, he suggested that regulation must apply to chips and other underlying technology used in AI.
“Here in California, there was recent regulation [that says] some GPUs are not allowed to be sold here anymore because they’re not efficient enough. That could be a path forward to pushing innovation and forcing more efficient chipsets,” Henderson stated.
Although there are financial incentives for businesses to utilize more efficient chips, such as cheaper costs, Henderson said this is still a topic that requires consideration from a regulatory standpoint.
Effective regulation is challenging due to a paucity of information on the environmental impact of AI, cloud computing, and Bitcoin.
“Step one is making sure we have enough reporting, enough data to make good regulatory and policy decisions,” he added.
Measuring overall carbon emissions is a hard task
According to Kathy Baxter, senior architect of Salesforce’s ethical AI branch, measuring a company’s overall carbon emissions is challenging because reports rely heavily on guesses rather than precise measurements. You can also check AstraZeneca’s guidelines on Ethical AI for more insight.
“There’s no way for us to really know if we are truly getting better or to identify what is correlation and causation. If you don’t know where your emissions are coming from, you can’t possibly control them,” Baxter said.
Salesforce launched its Net Zero Cloud initiative for measuring its operations and most prominent clients’ scope 1, 2, and 3 carbon emissions measurements throughout the complete supply and value chains. While the company has direct control over scopes 1 and 2, scope 3 emissions are outside the company’s direct control, such as those from suppliers.
According to Baxter, more information is required to improve sustainability initiatives. Another research by Schneider Electric shows that businesses don’t meet their IT sustainability promises.
“There’s no way any single company, any single government, will be able to solve this problem. We’ve got to pool our data together, and we’ve got to figure out the solutions together,” she added.
Artificial intelligence is one of the leading technologies that will take humanity forward and bring civilization to new heights. But at what cost? We will continue to follow the environmental effects of artificial intelligence, increase awareness and be a part of the solution.