What has artificial/machine intelligence (AI) to do with environmental destruction and global warming? What goes into the making of such systems?

The language of “cloud computing,” “virtual reality” and “cyberspace” has inured us into thinking that the web and AI systems are floating in an ethereal, other-worldly sphere that is divorced from physical bodies and their natural environments.

AI is not as artificial or intelligent as many imagine it to be. It is part of an extractive, late-modern capitalist economy that strip-mines the internet for all our personal data, just as strip-mining for coal and mineral resources were the foundation of early-modern capitalism.

This mass data is then fed into devices that manipulate and manage us in ways more powerful than all earlier methods of surveillance and social control. Data is the new capital.

Also, like 19th-century robber barons, AI development is largely in the hands of a few hi-tech giants in the US and China who wield concentrated, unaccountable power.

The giddy “hype” that attended the dawn of the “Information Age” has now been replaced by sober soul-searching by the more reflective practitioners in the field.

With regard to AI, the racist, sexist and other biases inherent in training data sets and many algorithms have been exposed.

Employees of Amazon, Google and Facebook have publicly complained about the dehumanizing nature of much of the work in these companies, and – given their scale of operations – the difficulty of regulating them.

They are asking the basic questions: Who is making these AI systems and why? What are the effects on the planet as well as on “ordinary” peoples’ lives?

One such prophetic voice is Kate Crawford’s new, deeply researched book Atlas of AI: Power, Politics and the Planetary Costs of Artificial Intelligence.

“Exploitative forms of work exist at all stages of the AI pipeline,” Crawford notes, “from the mining sector … to the software side, where distributed workforces are paid pennies per microtask. …. Workers do the repetitive tasks that backstop claims of AI magic-but they rarely receive credit for making the systems function.”

Her fascinating tour of the world of AI begins with a reminder that strip-mining is more than a metaphor for the plundering of our data on the internet: it literally is what supports the development of AI.

For instance, all our smart phones and laptop computers depend on lithium in their batteries, and lithium reserves will disappear within the next 20 years.

The “cloud” takes up a vast amount of land. The world’s largest data farm is in Langfang, China, and covers 6.3 million square feet, the equivalent of 110 football fields.

The obsessive drive to collect ever-larger data sets in order to “train” machine language algorithms means that the computing industry is carbon intensive and could make up 14% of all greenhouse emissions by 2040 – about half of the entire transportation sector worldwide.

Researchers from the University of Massachusetts Amherst calculated that the carbon emissions required to build and train a single natural language processing system was about five times the lifetime emissions of the average American car.

An interview with Crawford about her book is available here.

We tend to forget that, like everything else humans do, our “virtual” communications are not ethereal but embedded in physical objects: power stations, data centers, undersea cables, overhead satellites, batteries and cooling systems.

Inside every wind turbine, smart phone, medical scanner and electric car are “rare-earth minerals.” This small group of 17 elements is in extraordinary demand, but the supply is limited to China and Australia.

Extracting rare-earth minerals is a difficult and dirty business, typically involving the use of sulphuric and hydrofluoric acids and the production of vast amounts of highly toxic waste.

Gold is also an element common in smartphones, primarily to make connectors. And gold mining is a major cause of deforestation in the Peruvian Amazon.

Extraction of gold from the earth also generates waste rich in cyanide and mercury, two highly toxic substances that can contaminate drinking water and fish, with serious implications for human health.

It is not only the environmental costs of our device-dependence that we forget. Jaron Lanier, one of the pioneers of virtual reality, laments the fact that “people degrade themselves in order to make machines seem smart all the time.”

In his book You are Not a Gadget, he observes: “We have repeatedly demonstrated our species’ bottomless ability to lower our standards to make information technology look good. … The attribution of intelligence to machines, crowds of fragments, or other nerd deities obscures more than it illuminates. When people are told that a computer is intelligent, they become prone to changing themselves in order to make the computer appear to work better, instead of demanding that the computer be changed to become more useful.”

Lanier continues: “People already tend to defer to computers, blaming themselves when a digital device or online service is hard to use. Treating computers as intelligent, autonomous entities ends up standing the process of engineering on its head. We can’t afford to respect our own designs so much.”

There is nothing new in the way engineers take the most advanced machines of their day as models and analogies for human functioning. But there is a short (though calamitous) step from modelling to identification in which we imagine that machines have those functions themselves.

When we speak of “clocks telling the time”, what we mean is just that they enable us (conscious human persons) to tell the time.

The philosopher Raymond Tallis refers to the “fallacy of the displaced epithet”. Walking sticks don’t actually walk, and running shoes don’t run.

The same applies to “radar searching for aircraft,” “telescopes discovering black holes” or “smart phones remembering our appointments”: they do not literally search, discover or remember.

If there were no conscious human persons using these prosthetic tools, these activities would not happen.

The lesson: pay attention to language, and ask questions about technology – for every benefit to some, who bears the costs?

Editor’s note: A version of this article first appeared on Ramachandra’s blog. It is used with permission.

Share This