Gerry McGovernJun 5

Imagine something that in 2020 didn't really exist but has so exploded onto our environment that in 2030 it will demand the water of 1.3 billion people and the electricity of 2 billion.

In a time when we have a global historic severe drought and when our civilization is about to collapse due to over-consumption of energy.

This is AI.

This is Big Tech

These are the tech bros. Here to burn it all down for greed and ambition.

https://news.un.org/en/story/2026/06/1167658

AI’s environmental costs threaten water, land and climate

Artificial Intelligence is not only responsible for worrying amounts of earth-warming greenhouse gases: the technology's environmental footprint is also expanding at a pace that could strain the planet’s natural resources.

UN News
Show threadKumar McMillanJun 6
@gerrymcgovern @patrickhadfield ironically, the Chinese are working hard to reverse engineer and distill all the frontier models into open weight models that everyone can run *locally* on high powered laptops, no data centers needed aside from initial distillation. It’s a weird thing to cheer for but I am.
Show threadDr Susi ArnottJun 6
@kumarvibe @gerrymcgovern @patrickhadfield
Don't fully understand this toot, but find it encouraging nonetheless. (Is this a sign of personal cognitive decline in my late sixties, out of interest?) #hellishdatacentres
Show threadK

@SusiArnott @kumarvibe @gerrymcgovern @patrickhadfield

https://www.mindstudio.ai/blog/ai-model-distillation-attacks-explained

AI Model Distillation Attacks: What They Are and Why They Matter

Anthropic, Google, and OpenAI all reported distillation attacks from Chinese AI labs. Learn what model distillation is and why it's a security concern.

MindStudio
Jun 6 at 10:46amWeb
Show threadKumar McMillanJun 6
@6502ALU @SusiArnott @gerrymcgovern @patrickhadfield Yep. The Qwen models are pretty good for things like coding and you can run them locally on a fast laptop. https://qwen.ai/home
Qwen Studio

Qwen Studio offers comprehensive functionality spanning chatbot, image and video understanding, image generation, document processing, web search integration, tool utilization, and artifacts.

Show threadDr Susi ArnottJun 7
@kumarvibe @6502ALU @gerrymcgovern @patrickhadfield
So - even existing #datacentres and levels of #datatraffic are unnecessarily bloated? That had been an unconscious assumption of mine, but seriously under-informed and superficial - thanks!
Show threadKJun 7
@SusiArnott @kumarvibe @gerrymcgovern @patrickhadfield it's really a question of where, in the long-tail of diminishing returns, the potential for value isn't worth the cost. It's still in the speculative bubble phase, so probably yes!
Show threadGerry McGovernJun 7

@SusiArnott @kumarvibe
I spent almost 30 years in the data / content industry and at least 90% of data is crap and should be deleted.

@6502ALU @patrickhadfield

Show threadKumar McMillanJun 7
@SusiArnott @6502ALU @gerrymcgovern @patrickhadfield For AI specifically, the demand on data centers doesn’t seem sustainable. I’m just saying there is an alternative which is running AI models on local laptops, i.e. not in data centers. It doesn’t perform as well yet but I’m hoping there is a future where it does. Many people (not just in China) *are* working toward this future.