Apropos of never ending discussions about whether ANNs are "good" models of the nervous system, here is a slide I present to masters students showing a network that is found in motor control circuits *across phyla* (that's pretty ubiquitous!) I ask them to guess what it does...
9
61
305
40,245
This question reliably elicits puzzled looks. I give this lecture at the end of a term-long course where they get taught about ANNs, RNNs, SNNs, DNNs and so on. To break the awkward silence I tell them that both synapses are inhibitory....
1
1
20
2,367
This often piques their interest: "Ahh, you didn't show the input" is a common response. I tell them there is no input. Again, quizzical faces. Something must be wrong. Who is this guy, anyway?....
Dec 17, 2024 · 2:42 PM UTC
6
1
45
2,409
The rest of the lecture is devoted to intrinsic dynamics: we demolish the notion of a threshold, we see short-term memory in hippocampus arising from single cell properties, and voluminous evidence that our beloved Pyr cells respond to *fluctuations* rather than steady input...
2
51
2,301
I readily accept that many of these students will write off this material as a bunch of curiosities and edge cases. That for every successful theory there are cranks questioning it to the grave, and that maybe I am one of them. I get this response from *some* colleagues, too...
1
26
2,230
But I never question the power of connectionism. I just try to excite people by showing there is more to learn. That animals with handfuls of units can perform complex behavior & that *this fact alone* tells us how far we are from engineering anything rivalling nervous systems...
1
1
59
2,503
If only 15% of students absorb this message, I'm happy. Why? Because I have learned, repeatedly, over nearly a decade, that students who are a few months and years into a subject are *incredibly* sensitive to intellectual ideology. If a vaunted tech guru speaks, they listen...
2
1
38
2,112
This isn't their fault. Curiosity hurts. In an education system that pushes the *economic* value and *opportunity cost* of learning, they need heuristics, shortcuts, TL;DRs. The thing is, most of those students aren't students any more. Some percent are faculty.. or tech gurus...
1
1
27
2,257
This is why banging on about the power of simplistic models is problematic. It kills curiosity. We don't need to bang on about ANNs, for they are now unavoidable parts of everyday life. By all means we need to teach ANNs, but our job as educators and scientists is to go deeper!
2
35
2,650
for those interested in a nice review that touches this topic, try this:
cell.com/AJHG/fulltext/S0960…
2
7
71
2,510
