OP-ED: AI Usage and Resources by Jake Ryan

EDITOR’S NOTE: TXAN 24 News is publishing the following op-ed as a contribution from an guest writer. The views and opinions expressed are solely those of the author and do not necessarily reflect the views, positions, or editorial stance of TXAN 24 News, its staff, or its affiliates.

TXAN 24 News is committed to providing a platform for thoughtful discussion and diverse perspectives on issues impacting our communities.

————

TEXAS (TXAN 24) — I keep seeing people post “AI uses 2 liters of water per image” like it’s a clean, proven fact.

Here’s my issue with that… we’re addicted to Facebook. And Facebook isn’t “non-AI.” Every swipe is powered by machine-learning systems deciding:

• What shows up in your feed (ranking)

• What gets recommended next (especially video)

• What ads you see

• What gets flagged, filtered, boosted or buried

Meta even explains that “Feed” is ranked by AI predictions about what you’ll engage with.

So if we’re going to be serious about “AI’s impact,” we can’t only point fingers at someone making a stupid image.

Doom scrolling is an AI workload too — it’s just spread out into thousands of tiny predictions plus a ton of video data moving through data centers and networks.

And the International Energy Association has pointed out that with streaming, the data transmission can be a major part of the energy footprint.

Also: that “2L per image” number is not a universal measurement. Researchers usually have to do it in steps:

1. Measure electricity used to generate an output.

2. Translate electricity into water using data-center water metrics that change by location and cooling design.

Meaning: water use varies. Climate, power grid, cooling type and what “water use” even counts (cooling only vs electricity generation, withdrawal vs consumption) all change the result.

So my point isn’t “AI doesn’t use resources.”

It does.

My point is: if we’re going to talk about AI ethics or resources, we should also talk about our own screen habits. Because a lot of the AI load is coming from the endless scroll we normalize every day.

If we’re worried about impact, maybe the first step isn’t yelling “stop making AI art” but maybe it’s asking: How much time are we giving the machine that’s training us to keep scrolling?