Something Big Is Happening #1

An update on the environmental, cognitive and economic impacts of the GenAI fever, plus a bunch of recommended reads to go deeper.

Hello everyone and welcome to a new issue of the Something Big is Happening pseudo-regular column.

As a reminder, this is a recurring space trying to decrypt what is going on in the (Gen)AI space once you remove all the hype, vibe reporting and abundant marketing material we are bombarded with.

In each issue I share interesting findings from independent researchers and/or journalists and usually add my own perspective and reading on the facts presented.

All articles in the Something Big Is Happening series¹ are for supporting subscribers who have opted in for the paid tier. Thanks to all of them, I’m able to publish most of my content for free. If you like this content and would like to support my work, please consider upgrading to paid.

Now, let’s see what we have in stock for today.

The most enthusiastic reporters and commentators believe that GenAI will solve the main challenges that humanity is facing, and then some. Today we’re going to take a look at how we’re doing on that front

• GenAI's outstanding contribution to solving the climate-crisis

• GenAI's unparalleled contribution to making humans smarter

• Who cares about that? It was all about cutting costs… was it?

So, let’s look at the first promise. Despite Sam Altman's repeated statements on the subject², things don’t look too well for planet Earth.

Let’s take a look at a reputable source on the topic³.

Hello World, How Are You Doing?

It’s interesting that one of the first pieces I’ve ever written on the topic of AI, about two years ago, was focusing on the immediate and tangible environmental effect this new gold rush was having.

Relax: AI is not coming for your job, only for your energy and water

That was about two years ago, and data was still limited. Most of those concerns were being dismissed as only transitory, irrelevant or just necessary in order to reap the upcoming massive benefits. You know, just the same old magical-thinking BS that by investing enough money and resources into this technology (god), it will solve all of humanity’s problems (miracles).

Let’s see how things look almost two years later.

Greenpeace released a recent piece on the currently observable impacts of GenAI. The article covers a lot of ground and dimensions, but specifically on the environmental aspect, it cites a recent study from Greenpeace South East Asia focused on the environmental aspect of AI chipmaking (not even taking training and inference into account).

The findings are staggering:

• Global electricity consumption from AI chipmaking increased by more than 350% year on year, from 218 GWh in 2023 to nearly 984 GWh in 2024.

• Global emissions from electricity consumption related to AI chipmaking increased more than 4-fold in 2024, rising from 99,200 metric tons of CO₂ equivalent in 2023 to 453,600 metric tons of CO₂ equivalent in 2024. The emissions were driven in large part by heavy reliance on fossil fuels in East Asia’s power grids.

• Across East Asia, growth in electricity consumption from chipmaking is being used to justify false climate solutions such as gas and nuclear.

When we move past the production of chips to their actual deployment and usage, things don’t look any better.

A thorough article from last year’s MIT Technology Review reported an interesting finding: the carbon intensity of electricity used by data centres (in the US) is 48% higher than the US average.

In layman's terms, that means that the electricity used by data centres relies significantly more on fossil fuel than the average in the US. So much for contributing to the green transition.

Once you realise that, it’s not surprising that the big tech race to buy carbon offsetting credits has exploded in recent years, leaving some commentators to believe this is a good thing… because it will serve as an incentive for more carbon capturing and offsetting investments.

It’s a bit like saying that smoking cancer-causing tobacco is a good thing for healthcare, as it will create bigger incentives for cancer research and treatments!

The problem with this massive increase in electricity production and carbon emission is that it only makes things worse. By significantly increasing the overall amount of emissions, we’re making it harder to reach what are already close-to-impossible goals for ensuring survival for most people on this planet.

This is basically what most recent reports from big-tech sustainability departments are saying quietly. Decarbonation is a marathon, we’re pragmatically optimist (WTF does that mean?), and we believe that by 2030 everything will be green (without laying out a clear and solid path on how to get there). 2030 is less than 4 years away, and the trend is worrisome.

A recent article from the LA Times does a good job of summarising the situation:

Over roughly the first five years of their climate commitments, for example, Google’s emissions jumped nearly 50%. Amazon’s rose by 33%, Microsoft’s more than 23% and Meta’s more than 60%.

How can we believe any promise that things will suddenly get better in the following 4 years?

Offsetting is not net reduction. Offsetting is just a greenwashing way to make people believe you care about anything but greed.

But hey, humans are intelligent creatures. Our creativity and ingenuity got us this far (is that a good thing?). AI will surely help us continue on the same journey. Or won’t it?

Hello Brain, How Are You Doing?

OK, AI might be contributing to making this world a lot less livable, but at least it will make us humans a lot better. Right?

Not so fast. I’ve been positively surprised to observe a recent resurgence of another AI-related topic I’ve been writing in what looks like it was a long time ago (about a year):

Is GenAI Digital Cocaine?

Now, since that provocative but informed article came out, we’ve made a lot of progress in building a more nuanced understanding of the real impacts of AI usage on our cognitive ability. In rapid succession, a bunch of articles have come out recently on this topic.