#SpatialCognition question: any studies that show that if you learn a response (e.g. "turn left at choice point) in one environment, you are more likely to repeat this response in a new environment?
So, "response #generalisation" ?
I'm pretty sure this is a thing but can't find a good study to cite about it...
If we survive this era it will stlll be necessary to conclude that in general the boomer generation learned little about raising the next generation well and doing a good handover.
#o3 , la nouvelle #IA #AI d' #OpenAI , obtient des #résultats comparables à ceux d'un #humain sur un #test d' #intelligence évaluant l' #intelligence #generale #agi ( 85% au #test #ARC - #AGI ) , et a montré une #capacité de #généralisation en #résolvant des #problèmes avec #peu de #données .
Un nouveau modèle d’intelligence artificielle (IA) vient d’obtenir des résultats comparables à ceux d’un humain lors d’un test conçu pour mesurer l’« intelligence générale » — des résultats bien meilleurs que les IA précédentes. Que sait-on précisément de cette avancée, et que signifie-t-elle ?
This is the fascinating observation that it’s not straightforward to say how long a coastline is. If you were to measure the coastline of a country by using a ruler on a globe, you would come out with a vastly different number than if you were to pace around the edge. The closer you look, the more wiggles and squiggliness you come across and instead of converging on a more accurate length, the coastline just keeps getting longer. The smaller your ruler, the longer it gets. This was originally spotted, incredibly, in the 1950s, by an Englishman, Lewis Richardson, when trying to check a theory he had that the likelihood of war between countries depended on the length of their shared borders. Remarkably, he found that the quoted lengths of borders varied significantly. While measuring on maps at different scales he saw that the smaller scale map he used, or the smaller the width of his calipers he was measuring with, the length systematically increased. When looking at coastlines, instead of borders, some countries had wigglier coasts and so the length increased at a faster rate with the scale — for instance, Norway’s coastline, with it’s crinkly fjords, increases faster than Britain’s, which in turn increases faster than South Africa’s, as he zoomed in. The rate of this increase later became known as its fractal dimension. Long after Richardson’s research, Benoit Mandelbrot published a paper How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension that discussed how the wiggliness of something like a coastline at one scale can be repeated at smaller and smaller scales. The work led to the later term fractals. Many other things exhibit fractal-like behaviour such as river networks, borders, brains, frequencies, lightning or even the stock market. More mindbending mapping, and the difference between Great Britain, the UK and the British Isles. There’s a super section on this in Scale, by Geoffrey West.
OK, I apologise for my #generalisation. I was being a little bit #provocative.
Controlling Confusion via Generalisation Bounds
Why does everyone always make sweeping generalisations?
El Duvelle Neuro
eicker.news ᳇ tech news
Craig Duncan
Lammergeier Ferekt
Rod2ik 🇪🇺 🇨🇵 🇪🇸 🇺🇦 🇨🇦 🇩🇰 🇬🇱☮🕊️
💧🌏 Greg Cocks
CaGIS
Bob Williams
New Submissions to TMLR
RufuScott