4 ms·
I read somewhere[1] that because we know far more now about human neurons than we did five years ago, modern CNN's are better compared to a _single_ neuron (but
by j2kun 10y ago
I read somewhere[1] that because we know far more now about human neurons than we did five years ago, modern CNN's are better compared to a _single_ neuron (but are probably still simpler than them). I'm far from an expert in brain biology, but what I took away from that is that it's specious to make these comparisons based on our current knowledge of either field.
[1]: http://timdettmers.com/2015/07/27/brain-vs-deep-learning-singularity/ http://timdettmers.com/2015/07/27/brain-vs-deep-learning-sin...
- gwern 10y ago> modern CNN's are better compared to a _single_ neuron You obviously cannot reach human-level performance on ImageNet or do anything at all with a single neuron, so such an absurd equation refutes any such calculation or attempt to use a comparison as an upper bound.
- j2kun 10y agoThe linked article explains how a single human brain neuron is a lot more complex than previously thought. Obvious is also a word we should avoid using in discussions about brain biology.