Oi, misleading statistic.
Older papers are going to be cited more by virtue of both being older, but also more basic.
It's still cited. It's still the basic model of ANN learning. We've gotten better at more of the same.
Yes, this is both to be expected and very cool
It's very cool. Not expected. What was expected was something closer to human intelligence. Instead, we get slug learning.
It's a logical consequence of humans not being the only possible intelligence.
Mice are vastly smarter than our best computers.
Psssst. FPGAs.
Not remotely similar.
Whether or not hardware re-writes itself depends on what you mean by "hardware."
No, it doesn't.
On the most basic, mechanical level, the whole machine works by the hardware causing mechanical processes that leave the hardware in another state.
Part of it, yes. But these states are built in. They are not flexible and necessarily so.
If we step up the abstraction a bit to include the notion of software, then the brain is strictly less flexible than a modern computer.
This is so far from true it would be funny if you meant it as a joke. A neuron is more flexible than a modern computer. Computers are designed from the ground up to be rigid, neat, and compartmentalized. That's why they're such great calculators.
The brain implements its ability to adapt and change its programming by hardware
It doesn't and to think of it this way is incredibly misleading. The brain has no programming it "implements" and "it" can't change "it's hardware" because there is nothing in the brain but "it".
however, I'm fairly sure that that adaptation is limited, leaving certain aspects of the mind invariant. (e.g. how we learn.)
We learn through adaption. That's what ANNs try to imitate. Adaption.
A modern computer, OTOH, can rewrite any aspect of its software into any configuration - all computers are universal, unlike the brain.
All computers are universal because they are universally limited. They
cannot implement any configuration because they are limited by a universal hardware set-up: a physical implementation of Boolean algebra. All software must be
very strictly designed because computers are so incredibly limited. Mice learn faster and better.
There's no distinction "central to computer science" here
There is. Computers are finite state machines. They must have some permanent implementation that is compatible with a basic turing machine. Making the hardware "re-configurable" changes the entirety of computer science.
the brain's software is implemented a level of abstraction down compared to a normal computer's
It isn't. At all.
given that the brain is built of von Neumann replicators.
It isn't.
How "permanent" do you mean?
Memory is stored separately and compartmentalized until it is changed according to specific routines. The software/hardware separation is absolutely fundamental to data integrity. The brain doesn't have some special "memory" storage that remains until it is accessed. Memory and processing necessarily go hand-in-hand. This makes human memory less reliable but is essential for conceptual processing. It means memories are literally connected to constant thought processes as well as other memories (they are not ever completely distinct from either). That connectivity is vital to understanding. It's terrible for exact recall. There's a trade-off. When computers can't recall data it's because something went wrong.
You can probably build hardware that'll store it as something "constantly active,"
You can build it to delete data too. The point isn't that it is stored as active connections
per se, but
how it is. It is so vastly different than computers the comparison between the too is utterly flawed and continues, alas, to mislead.
Have you looked at agent-based programming?
Yes. It's not particularly impressive. I work a great deal with computational intelligence paradigms, soft computing, machine learning, etc. Some to understand the brain but often because these are key for high dimensional data analysis quite frequently as well as programming interactive surveys, tasks, etc. for studies.