Sven for short

This paper is foundational for thinking about where the economy is headed as AI capabilities advance.

Def was thinking about the brown note.

Google set out to tackle the hedgerow problem. The hedgerows, stone walls and little copses doing much of England's ecological work are too small for any national forest map to see, so on paper they barely exist. Google decided to fix this. National forest inventories only count woodland above a certain size, so the thin ribbons of trees along field edges, the shelterbelts, the lone stands in a pasture all fall through the cracks, even though together they store carbon, shelter wildlife and stitch fragmented habitats back together. Google has now mapped every one of these features across the whole of England and turned them into data you can actually work with. Start with the scale of it. England covers more than 130,000 square kilometres, and the system draws the fine woody features across all of it. The surprising part is how little hand-labelled data that took: roughly 247 square kilometres of carefully annotated samples, a rounding error against the national area. The model could stretch that far because it sits on a foundation trained on more than 300 million satellite images from around the world, so it arrived already knowing what trees, edges and fields tend to look like from above, and needed only a little local teaching to specialise. To separate features that overlap and sit at different heights, it reads two layers at once: submetre optical imagery for the fine shape, and one-metre LiDAR, which measures height from the air, to tell a low hedge from the tree standing behind it. Then it sorts each feature by what it does in the landscape, using a simple measure of shape. Round, chunky patches above thirty metres across count as woodland. Long, thin, winding shapes get flagged as linear features like hedgerows. Shape becomes a stand-in for ecological function, worked out automatically across thousands of map tiles. The whole national sweep ran in parallel on Google Earth Engine, chewing through thousands of satellite cells at once rather than grinding through the country one region at a time. The new release, Vectorized Farmscapes 2020, turns that into clean geometric shapes, each hedgerow or copse its own object with a length, an area and a position. You can measure it, add it up, and track whether it grew or vanished between surveys. link to full article: research.google/blog/from-pi…

Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory). AI starting to help us improve our understanding of the universe! openai.com/index/building-sc…

My friends were all talking about prepping so I made a survival LLM. A free app that runs an AI mdoel plus 1.5 million offline articles (medical guides, FEMA, survival manuals, wikipedia) entirely on your computer. No internet needed. brangus.io/survival-llm

Thank god we can find and intercept lost hikers as they mistakenly run away from rescue in dense forest!

In other words, people hate brutalism.

despite the unfortunate timing, we take disclosures seriously at @typesafeai at approximately 6:49PM this Friday evening, we became aware of a massive unauthorized leak of ALL customer noul outputs in effort to maximize transparency, here is the link: gist.github.com/diogo149/810…

For the next few months, we are looking at frequent discoveries of long-standing bugs - competent code auditing has gotten cheap and abundant. This carries a few consequences: