Artificial intelligence (AI) is a power-hungry technology that carries an inherent impact to the planet. However, the most harmful factor may be enterprises' failure to responsibly address AI's carbon impact in favor of accelerated innovation.

AI is a concern for Intel's Director of Green Software and Ecosystems Asim Hussain (who also serves as executive director of the nonprofit Green Software Foundation) not necessarily because of the amount of energy the technology consumes, but because he anticipates business growth will be prioritized over the responsible use of AI.

"We're assuming [AI] is going to be used minimally, but I suspect it's going to be used very poorly," Hussain said during a virtual event on UN World Environment Day.

Today's conversations around AI make assumptions regarding how the technology will be used from an efficiency perspective. It's fairly expensive to train large language models (LLMs) used in generative AI technology, "so people are assuming ... we're barely going to use it," he explained.

"People think we're just going to use AI responsibly," Hussain added. But it's a powerful technology that can easily help enterprises develop a product or service faster, which tends to overtake concerns about AI's energy consumption.

"The ease with which AI gets you to market faster inefficiently means that it will be used, I believe, in a very inefficient way," he said. "This is my worry about AI."

Enterprises are likely to favor AI-enabled innovation and increased developer velocity over a smaller carbon footprint, GSF Community Chair Anne Currie added.

Silicon Valley's mantra to move fast and break things has exchanged machine productivity for developer productivity, "and I can't see many businesses going back on that, because it'll put them out of business," she said.

With AI, "we're giving them a tool which is incredibly hungry and fast," a scenario Currie described as "the worst thing we should be" doing from a green software perspective. To that point, regulations are needed to pressure AI and machine learning companies into efficient design and "making sure that it is green," she said.

One of the most important things to ask a vendor before using their AI model is their take on responsible use of the technology, according to Google Cloud Director of Developer Outreach Matt Thompson. "If you're producing your own models, it's something that you're going to want to make sure that you account for," he said during a GlueCon keynote last month.

Responsible AI should be treated as a feature of the technology that can differentiate vendors. "We live this every day," Thompson said, citing the nine years of work and hundreds of Google employees that contributed to the hyperscaler's responsible AI approach.

According to Google, AI should be socially beneficial, avoid unfair bias, safely built and tested, accountable to people, designed with privacy and dedicated to scientific excellence. "These are things that we actually are looking for in both products and partners," Thompson said. "If it's likely to cause overall harm, we have no interest. If its principal purpose is to direct injury, we don't have an interest."

In addition to environmental risks, AI poses plenty of security-related threats. "I'm not going to predict doom and gloom," Thompson said, "but I want everybody to understand that there actually is risk inherent in these models. You could create a model — it's been said before — that may be able to consume all the learnings from all the different malware that's been created," he explained.

Based on its own responsible AI principles, that's not something Google is "interested in supporting," Thompson said. But "that doesn't mean there won't be other people out there on the planet looking to build models like that. As you think about protecting your enterprise [and] protecting your data," responsible AI is "[some]thing you have to think about," he stressed.