No, they really aren’t and I’m not sure why I keep seeing this take. ML weights are binary and it’s painfully obvious.
They are the end result of a compilation process in which the training data and model code are compiled into the resulting weights. If you can’t even theoretically recreate the weights on your own hardware it isn’t open source.
> The six-layer cortex appears to be a distinguishing feature of mammals; it has been found in the brains of all mammals, but not in any other animals.
What it evolved out of is entirely irrelevant -- everything evolves out of something else in some fashion.
While the neocortex does have distinct layers, the neocortex itself is not “layered” around the rest of the brain - it’s deeply integrated all over the place with the rest of the brain and nervous system. There is no hierarchical relationship between the neocortex and different systems it integrates with.
The reptilian equivalent to the neocortex is the dorsal ventricular ridge which evolved separately and in parallel. This presents two problems to the hypothesis: first the much simpler DVR serves much of the same purpose as the neocortex which was completely unknown at the time and second the most interesting bird species (the smartest ones) often don’t have an equivalent structure at all. There isn’t even a clear relationship between intelligence, complexity, evolutionary age, etc. After 250 million years of evolution any similarities are accidents of random convergence.
Not sure what you’re talking about. They pretty clearly distinguished between base and total pay even for the highest earners…
> For instance, the highest reported base salary was for a level 7 software engineer, earning $718,000 in base salary and a total compensation package of nearly $800,000.
Yes, I think the bonus factor over base salary is 25% or more for L7, so I'd expect salary + bonus to be $900k. The total comp would likely well over $1M, since GSUs (Google's version of RSUs) were not even mentioned.
They are the end result of a compilation process in which the training data and model code are compiled into the resulting weights. If you can’t even theoretically recreate the weights on your own hardware it isn’t open source.