>>we developed statistical models to predict bird behaviour. These models were based on long-term observations of bird interactions led by Philip Gibbons at the Australian National University. By simulating how birds might use artificial branches, we could refine our designs to better meet their needs.
It sounded like a great use of an ML classifier interpreting raw data (possibly video footage?) of bird observations... Until the line where they reveal their data source for the model was already classified data from a previous study. And the previous study also had similar conclusions about horizontal branches on tall old trees being the most preferred.
I'm sorry I know this HN and not reddit, but your formulation is so close to the meme "Thinking quickly, Dave constructs a homemade megaphone, using only some string, a squirrel, and a megaphone."
Ai is just a tool. Its not necessarily better than any other way of modeling data until you prove it is. That being said it is a newer tool, and whenever a newer tool comes out theres a lot of low hanging fruit for researchers to validate existing notions with the new tool. These won’t be seminal papers or anything but serve as proof of work so the author can get a relevant job.
I don't know... you might as well ask people who live near trees which ones are preferred by birds?? I know I have lots of trees around my house and I can clearly see which ones tend to attract birds, and which kinds of birds. And I am no expert... with a little training, people like me might be able to contribute floods of data.
But of course, that's not as cool as using lasers to scan trees and AI to analyse the data. And I admit that in the case of the Australian bush, there may not be enough people living in the affected areas to collect enough information, so perhaps this is in fact the way to go.
Citizen science approaches are a great way to collect data but the data can also be 'lumpy' (ie. focused on charismatic species, or easy-to-measure things). But the main idea of this approach was to produce quantitative descriptions of trees that could then be fed into a design system. This system can then output many different designs using those as constraints but also other constraints (ie. being lightweight, fitting around existing objects, etc).
This is just an initial test - the next step is to deploying prototypes in the field and observing bird responses (ie through some citizen science initiative). The idea is that a bird can 'read' a design (by using one structure over another), and that data can be fed into the system to then produce an updated generation of structures. Lots of work before we get to that stage!
I'm curious as to whether the study is considering whether birds will use an otherwise undesirable tree (or a part of it) if there is some other factor (e.g. adjacent food) that also drives their behaviour. For example, I know from observing my own (UK) garden birds that many of them will perch in trees near to my feeders for an initial recce or to eat collected food, but that they don't otherwise use these trees for perching.
Really neat work. I have a friend who designs green spaces in urban environments and it makes me wonder what he could do with these kinds of tools/approaches.
> you might as well ask people who live near trees which ones are preferred by birds
I live in the woods and I have zero idea which trees the birds prefer. They go in most of them, I think. Sadly being near the birds and the trees for years has not made me a bird and tree expert.
> There's one glitch that occasionally confuses people. When the software lobs a story, it displays a rolled-back timestamp—not the original submission time, but a resubmission time relative to other items on the front page
It was first posted 2024-07-01T09:39:05 - if you mouse over the "8 hours ago" you'll see the true time, or if you click past algolia will show you the age from original submission.
It sounded like a great use of an ML classifier interpreting raw data (possibly video footage?) of bird observations... Until the line where they reveal their data source for the model was already classified data from a previous study. And the previous study also had similar conclusions about horizontal branches on tall old trees being the most preferred.