Post

It hugely reduces the friction, provides direct access to Google’s massive suite of satellite images, and I’m sure the model that Google is using is optimized to produce realistic looking satellite imagery.

People will use this to create fabricated imagery that they can reuse in other places/documents/social media accounts/fabricated news stories.

Oddly enough, I noticed yesterday that Google Maps has the entire city of Minsk and its surroundings completely bleached out. Not blurred, like some special military installation, but bleached of colour, and quite useless as a map.

Google has also decided to effectively kill the desktop version of Google Earth Pro - which has been a genuinely useful mapping tool for many years. They’ve apparently decided those resources are better spent allowing users to burn AI credits on creating satellite imagery disinformation.

theverge.comGet Google Earth Pro while you still can.In an update on Tuesday, Google says it will block users from downloading its Google Earth Pro desktop app starting June 25th, 2027. People can continue to use the app after that date, which offers ac...

Google Earth Pro has been guilty of gross misinformation even without generative AI. It always tries to render a 3D view even when it doesn't have the actual 3D data. Thus, this first image shows the actual appearance of the Valle de Cuelgamuros, the Spanish Civil War memorial built by Franco,

featuring a colossal cross; and here is how it is rendered by Google Earth Pro, which does a good job at guessing the 3D shape of the hill, but not such a good job with the cross. (Fallen and squashed down onto the ground).

The Mac version hasn't been updated for "Apple silicon" processes--a trivial recompile, I think?--so it will quit working in fall of 2027 when Apple stops including the built-in Rosetta 2 emulation layer. I will be pleasantly surprised if Google updates it for "Apple silicon."

Sad to see a useful tool gutted I used it for work to figure out the number of cemeteries in the US and the distribution of their sizes. I joke that I’m the foremost leading cemetery market sizing expert. Rip to a real one

There’s so many GIS-adjacent folks that are gonna be fucking livid when they can’t install google earth on a new machine. The web-based stuff shits the bed with big/complex datasets. Sounds like GIS folks are gonna have to go back to making a lot of pdfs. Joy.

I'm trying to figure out what the imagined non-evil utility of this is supposed to be. Why would people want the ability to add random shit to Google Earth images? Is this just a case of "we need to figure out how to shoehorn generative AI into every product?"

There's likely an element of that - like when they went all in on Google+ a generation ago, internally they've been told that all bets must now be Gemini bets...

Try showing what your area looks like with 3m sea level rise. Or show the effect of a dam failing. Or what your city looks like with twice as many trees. Or what your city looks like if everyone takes mass transit and the number of cars is reduced by 90% and parking lots are repurposed.

Yeah, I was just trying to imagine a business model. Get people hooked on this, then put it behind a paywall and charge $20 a month for 40 images, 50¢ a pop? But that doesn't really make sense. It's underpants gnomes all the way down.

Just posted about this. The use cases are here blog.google/products-and... bsky.app/profile/wend...

blog.googleTransform any place with Nano Banana in Google EarthCustom images — generated from Google Earth satellite, aerial, and 3D imagery — help you redesign any location.
wendyspirit.bsky.social@wendyspirit.bsky.social

Yes any tech can be abused. Not sure of the utility in seeing a historical place turned into “a futuristic sci-fi utopia” by #AI Other use cases Google is suggesting for education and actual real estate projects are pretty cool. Full disclosure I’m a techie and have Google Earth as my watch face 🌎

Another post. Good news. The EU is putting some overdue regulation on AI bsky.app/profile/wend...

wendyspirit.bsky.social@wendyspirit.bsky.social

📌 Starting Sunday 2-Aug, an EU Commission will have an arsenal of regulatory powers to fine or ban #AI models from the market if they don’t ensure safety. This is the first major global effort to implement a standing rulebook for the world's most capable AI models to address the most serious risks.

As someone who wants to do that sort of stuff, I wouldn’t want genAI to do it. I want the ability to tweak and edit not draw random finalish products. Its unlike apple’s magic select or code generation where you get an intermediate input that can be tweaked to make something useful.

Here are a couple of examples. Each image took me ~3 hours while watching TV. The left maps use real roads and parks with the zoom in to the town of New Thrace being made of a rotated Tulare. Using GIS let me manipulate details to make a plausible fake place. 1/

I love it when AI is applied to "solve" a problem that was already solved decades ago in my Latin textbook (i.e. get an actual expert archaeologist to draw exactly what it actually looked like 🫠)

It looks like a mix of "a hammer in search of a nail" and "my supervisors will be impressed" to me. No actual social value. Obvious social problems. I doubt one person who studies social impacts of technology for a living was consulted.

I mean, they recently put out a statement on the lines of “we know we’re miserably failing on our climate targets, but we like AI more than we like keeping CO2 emissions down so we’re not sorry.” They haven’t given a shit in a while.

Google should have to go through an IRB review process every time it makes a product change. The slowness and cumbersomeness of the process should be exactly proportional to the number of people who would be affected by the product.

I've done something marginally similar ("here's what it will look like if we put up the building in this project proposal") and even for that this doesn't seem like it would work

One use case: people have started to use GenAI to push back against redevelopment proposals, but I'm not sure that "We can all make heaps and heaps of shitty thoughtless renderings!" is really going to improve an already fractious public process. bsky.app/profile/imag...

David Ramos@imagineterrain.bsky.social

Over on the Inner Harbor Forum FB group, Baltimoreans have been using GenAI tools to visualize their ideas for how the rebuilt Inner Harbor should look, in opposition to the redevelopment proposals. I should be cheering the democratization of tools, but cheap ideas are also corrosive.

The commoditization of truth. Destroy credibility in government, privatize access, control the information domain and rule the world. This is the definition of evil.

But suppose you're Marco Rubio, and you need scary photos of Cuba's vast WMD arsenal, and mainstream woke satellites can't seem to find any. It's a real problem, but now Google can solve your problem. You'd pay unlimited taxpayer dollars for this kind of evidence.

And the resulting war is really going to drive revenue over in Google's defense business unit, and it's all just exponential synergies from here on out, no possible downside to any of this.

Yes, this is obviously potentially catastrophic and put an awful lot of people in real physical jeopardy, but have you ever considered how much Alphabet has wagered on AI and how much money they stand to lose if people don’t wholeheartedly adopt Google’s AI for basically everything?

Sure, people may die as a result of this, but are you prepared to countenance the horror of Sergey Brin wanting to take a yacht trip with some friends, his main yacht needing repairs, and the backup yacht doesn’t have a cool chandelier in the billiards room? Can you live with that scenario?

I used to work on specialized machine learning research that generated models of historical trends in satellite imagery to then try to predict things like droughts, such that aid could be shipped to places before a drought became so severe as to cause a famine. Google's is the opposite use case. /1

If the case I worked on was one where it could theoretically dramatically enhance global stability by heading off water wars & food wars, Google's creation here could easily create enough misinformation as to incite severe destabilization. To call it negligent is, yes, a dramatic undersell. /2

Sure, anyone could do same by grabbing the satellite imagery & then pulling that into a regular image generator, or even before that in Photoshop. However, making it so easy for the end user enables the worst behavior *as a service*, like how half-baked autopilot creates distracted driving. /End

People that pooh-pooh'd all of us who talked about the malevolence of Project 2025 back in 2023 are now doing the same when we keep on stating that AI is a dangerous tool of tech titans: "You sound like a conspiracy nut" or "Omg. Give it a rest".

Add another entry to data that cannot be trusted if created after the introduction of AI. "asbestos of the internet" theory strikes again.

What with this and yesterday’s AI-eats-books thing, it’s like they want to remove any access to knowledge & truth, any means to independently verify anything.