QCon London kicked off this week, opening with a sombre keynote from Leslie Miley, Microsoft technical advisor to the CTO on matters including analytics and sustainability. The intense compute requirements of AI means bigger datacenters and more CO2 emissions and environmental damage, he said, while bias may be amplified by the current technology race towards increased use of AI.
The impact of ChatGPT has been huge, Miley said, likening it to the ascendance of the world wide web in the late nineties. He is not the first to make this analogy, while others have compared its impact to that of the iPhone. “I know something was different when I first saw ChatGPT,” said Miley, describing it as technology “meeting people where they were at.”
Fewer people though are thinking about the environmental and societal impact of AI, said Miley. It is well-known that training AI models is compute-intensive; in fact OpenAI itself (the company behind ChatGPT) remarked in 2018 that the amount of “compute per model” was increasing by “roughly a factor of 10 year each year. It’s been partly driven by custom hardware that allows more operations to be performed per second for a given price (GPUs and TPUs), but it’s been primarily propelled by researchers repeatedly finding ways to use more chips in parallel and being willing to pay the economic cost of doing so,” OpenAI predicted that the trend will continue.
Miley, who told us he was speaking in a personal capacity and not on behalf of Microsoft, referenced the fact that Meta, the company being Facebook, halted work on European datacenters to remodel them for AI work, as mentioned in The Register’s report last year, because “AI workloads at scale require a different type of datacenter than those built to support our regular online services.”
According to Miley, “the newer datacenters take anywhere from 10 to 19 million litres of water to cool them, per day,” at a time when some communities are suffering devastating water shortages. “We’re doing this so you can generate pictures, so you can have a conversation with an AI, so you can play around,” he said.
Miley referenced the huge US interstate roadbuilding programme in the fifties, under President Eisenhower. Although this transformed the country, it came at a cost to communities, especially poorer communities, some of whom were deprived of property or saw their neighborhoods divided.
The AI revolution could be similar, he said. “We will degrade environments, we will destroy communities, we will impact children.” Miley also showed a headline, “Historic black cemetery moved for Microsoft datacenter in Virginia,” acknowledging that “this one was really difficult for me.”
These sustainability considerations, important though they are, probably have less impact than AI bias. “This technology has the potential to not just incorporate our biases, but amplify them,” said Miley. He referenced reports that fashion brand Levi will use AI-generated models to simulate diversity as an example of “cultural appropriation by prompt engineering,” as designers look to AI to get the images they need to present an impression of diversity that is entirely superficial.
Another target of Miley’s critique was cryptocurrency. “Crypto just took off, people threw money at it, all of a sudden you had coal plants coming online that hadn’t been online for years. It had billions of dollars thrown at it, most of it was a fraud … China was right to ban [crypto] mining, there’s no benefit to it,” he said.
Miley’s conclusion was that “we have an obligation to make sure that we are meeting [technology needs] with compassion … if we do it wrong, I don’t know what the world is going to look like.”
He did have a few tips for mitigating the issues. “Small models are better,” he said, adding that “you can’t manage what you don’t measure,” suggesting that developers take advantage of things like the Emissions Impact Dashboard for Azure and the AWS Customer Carbon Footprint tool. He also recommended the work of Dr Timnit Gebru at the DAIR Institute and Dr Joy Buolamwini who founded the Algorithmic Justice League.
QCon is a vendor-neutral developer event and Miley is not the first speaker to use it to champion ethical development. Previously, ThoughtWorks Chief Scientist Martin Fowler spoke at QCon 2010 on ethics in software development, addressing such matters as dark patterns in UI design or coding for vendors with questionable business practices. It is true that developers are in a position of influence in this respect; the difficulty though is that the demands of commerce generally outweigh individual choices. There is always someone else who will do the work.