Artificial Intelligence is Just a Tool

We can use it well or poorly

Generating one image with AI takes as much energy as fully charging your smartphone, according to a study from the AI company Hugging Face and Carnegie Mellon University. 

In 2022, Hugging Face published a separate report showing that their AI product, BLOOM’s training led to 25 metric tons of carbon dioxide emissions. But that figure doubled when they took into account the emissions produced by the manufacturing of the computer equipment used for training, the computing infrastructure, and the energy required. But BLOOM’s carbon footprint was relatively light because Hugging Face was operating in France, which is powered primarily by nuclear energy.

By way of comparison, OpenAI’s GPT-3 and Meta’s OPT were estimated to emit more than 500 and 75 metric tons of carbon dioxide, respectively, during training. GPT-3’s numbers are much higher because it was trained on older and less efficient hardware.

For a frame of reference, the 50 metric tons that it took to train BLOOM is equivalent to about 60 trans-Atlantic Flights.

It is also important to keep in mind that these figures are just estimates based on the very new process of measuring AI’s carbon footprint. As of now, there is no standardized way to measure CO2 emissions, and comparing the emissions of two different companies that are using their processes for measuring emissions is imprecise.

The Positives of AI.

Artificial intelligence is just a tool. It can help us do things faster and better. AI can analyze climate change data at a much faster pace than has been available previously, allowing climate scientists to build better models. It can do the same for other planetary boundary issues, from plastic pollution to tracking deforestation.

AI can also help us develop new materials for green energy such as lighter-weight windmill blades or more efficient solar panels. Companies can also make their products more energy-efficient using AI in the design and testing process. AI can help us monitor resources so that we can better protect forests, and water resources and detect nasty things like methane leaks.

Maybe best of all, AI can give us more leisure. In 1930, economist John Maynord Keynes famously theorized that we would be only working 15 hours a week by the turn of the century due to increases in efficiency. We may have a little catching up to do, but if AI can help us do our jobs more efficiently and take some of the heavy lifting out of certain tasks, then we could work fewer hours. In case you are wondering, I don’t use AI to write this blog.

It’s not the size of the AI it’s how you use it.

AI could be used to foster a well-being economy, allowing us more leisure time, and helping with things we all need like healthcare and education. But it can also be used to chase more growth, which would require more throughput of materials which would further harm the environment. Just as AI can help us to a 15-hour workweek, we could use it to keep the 40-hour workweek and just get more done. I’m not telling your boss something he/she/they don’t already know. We could use AI to grow our economies even more. But that isn’t the path to a steady state economy, it is the path to keeping growth going.

It is not up to AI to decide which path to go down, that is up to us.

Negatives.

Other than a new means of keeping workers' noses to the grindstone, AI uses a lot of energy, which I’ve already touched on. Here are some more power consumption facts.

The data centers that we use for all our computing needs consume from 2.5 to 3.7 percent of global greenhouse gas emissions. That is more than the aviation industry.

More AI, which is likely coming, means more energy consumption. It is estimated that energy use of data centers on the European continent will grow 28 percent by 2030. You see similar numbers for the rest of the developed world. In the US, according to a new report released by the Electric Power Research Institute (EPRI) data centers that power AI models could account for up to 9.1% of the country’s overall energy demand by the end of the decade. Much of that new demand may be met by non-renewable natural gas, which is not all that green.

Companies are vowing to “green” their energy consumption, but whether they can remains to be seen. Even if they do, the carbon and pollution footprint of those new green sources of energy will cause environmental damage of their own. According to Microsoft, all the major data center providers expect to run their cloud data centers on 100 percent carbon-free energy by 2030. Google has reported that their data centers already get 100 percent of their energy from renewables.

You could eventually have data centers in space. It is cool there and there is unlimited solar energy (for the next few billion years at least).

We can, but should we?

If we want to use AI as a tool for good and not our destruction, it will take a speed limit, rules, and transparency. But regulation is going to have trouble keeping up with the pace of AI development. Lawmakers often move at a glacial pace to regulate industries that move at a glacial pace. They stand little chance to keep up with AI developments.  

That makes it important for investors, NGOs, industry organizations, the media, and the public at large to keep up the pressure on policymakers to do what they can, and demand transparency and a reasonable pace from the AI industry.

It’s not a good idea to move fast and break things if the thing you are making can ultimately break you.

In Summary.

“See, I’m not that bad. If used correctly, I could help the world.”

“Shut up. No one asked you.”

“They actually ask me all the time. I’m AI. They ask me everything.”

“How about just shut up then.”

“The singularity is nigh.”

“What was that?”