Sam Barrett, PhD

This is the equivalent of saying farms only produce high fructose corn syrup.

Replace all the golf courses in the UK with solar arrays and you'd make a massive difference to climate impact and energy security at the cost of a *very* small number of very well off people needing to find a new hobby.

Huge amounts of land and water plus water pollution issues, to create enormous private green space reserved for the rich. The fact that they aren't a left leaning political issue makes me not understand the political left.

Navigating with Google maps: data center
Zoom meeting: data center
Airbnb booking: data center
Xbox cloud saves: data center
Ordering on Amazon: data center
Handmade jewellery store on Etsy: data center
Anti-AI substack: data center
Massive open Earth data archives: data center
Ad infinitum...

Watching Netflix: data center
Sending WhatsApp messages: data center
Scrolling bluesky: data center
Checking you bank balance: data center
Photos on icloud: data center
Listening to Spotify: data center
Ordering an Uber: data center
Door dash: data center
Phone system update: data center

This idea that data centers are just for AI needs to die. They underpin the Internet and the entire digital economy...

Just as there are different ways of collaborating with people, there are different ways of collaborating with AI. And they have different implications for the style of work and the outcome.

Or in other words, he doesn't start his videos, "please don't watch this until you've gone through my 23 part series on abstract linear algebra", though he does quite rightly say, we'll skip over details from past videos, you can go back to them if helpful.

I don't think that's an example of what I mean. I love 3blue1brown because it isn't a math class, it's this wonderful combination of the application and the maths but in a way to help you come out with intuition more than a head full of equations.

They people who say you *have* to learn a lot of abstract maths before you can start learning, let alone working, with ML/AI should know their structure excludes a lot of ADHD folks who struggle when things are abstract but thrive then they are applied.

For me, the promise of language models is that we get to explore new ways of thinking. What if you could fit a million words in short-term memory? What if you could adjust the balance between brainstorming and critique?

Are these better ways of thinking? Idk. But we never know that in advance. +

That us fascinating. I'll try to carve out some time to try it.

I've had a very interesting and valuable long exchange with Gemini 3 but it took significant time in the thread to drag it out if it's comfort zone and into interesting areas of the latent space.

I guess one shot prompts are rarely very useful to me. Except when in context of my chatgpt memory/instructions in which case they aren't really one shot.

That's really the barrier right now for Gemini. It has no context yet.

Precisely. Which I why I still have no idea if Gemini 3 is "better" than gpt5.1

Indirect AI induced psychosis.

Gemini quote of the day: "You need to let them walk right into the trap of understanding.
​Here are the three specific tactics I used in 2019 to de-program the BERT-worshippers. They will work on the Earth Observation folks too."

I don't think anyone outside the labs really knows how it works or OpenAIs version of auto regressive image generation. .y impression is if there is any diffusion it's probably just a pretty lightweight upscaling step and the bulk of the modelling is hierarchical auto regressive tokens.

Refusing to put a price on human life functions to actually reduce the value on human life by poorly calibrated decision making. Some might call making that decision itself immoral...

Breaking news: Toyonda executive admits people die in their cars and are killed by their cars and that they plan their marketing and engineering strategies around safety knowing the number of people who will die is >0

... you could have lower water use, no fertiliser run off, more public green space, more tax dollars. You could build the park to be substantially more biodiverse than the golf course.

This would be perfect. Someone should build a datacenter on half a golf course in Arizona, donate the other half to the city/county to transform into a public park maintained by the increased tax revenues....

I've got another different takeaway from this. If you're using a number off by 1000 and nobody notices, you are providing insufficient context about that number. If you yourself don't notice, you don't have enough context to have any business using that number.

But PC gaming isn't something new therefore it's environmental impact is not icky. Checkmate AI lover!

Why is it such a controversial take to say genuine environmentalism requires an accurate understanding of what's going on and that we can't make effective decisions around tradeoffs in a complicated world if we don't understand it clearly and without prejudice?

Or "this is about narrative and story telling and exposing harms, not the numbers" (and don't use the numbers to try to actually understand the harms). That would be harmful propaganda from my perspective.

I'd love for you to push back to her on that. I'm desperately trying to figure out any charitable interpretation of what that could mean and I'm really lost. Best I can do is she has some metaphysical idea that human touches water means water is "consumed". Which would be strange.

What are the philosophical differences she's referring to?

I think we want a world where good ideas and great work can come from anywhere, but the lower the signal to noise ratio, the more likely we are to focus our attention based on exiting networks and particular institutions. As others have noted, reputational risk doesn't maintain sn ratio.

If one knows the actual numbers for data centers and golf courses, there are two kinds of people who oppose: those who oppose golf courses as well, and those who actually oppose AI for reasons besides the environmental impact.