3 ms·
Not a lot of humans do
by MrPatan 3y ago
Not a lot of humans do
- amelius 3y agoWhen a human has done the same thing many times they tend to try to generalize and take shortcuts. And make tools. Perhaps I missed something but I haven't seen a neural net do that.
- mitthrowaway2 3y agoIs that very different than the distillation and amplification process that happens during training? Where the neural net learns to predict in one step what initially required several steps of iterated execution.
- WalterSear 3y agoIMHO, yes. It's not an (internal) invention occurring as a result of insight into the process - it's an (external) human training one model on the output of another model.
- littlestymaar 3y agoTrue, but we're talking about Olympiad level math skills here.
- skepticATX 3y agoEvery single human has abstractions that are unique to them. Your world model isn’t the same as mine. It’s just that usually these abstractions are fuzzy and hard to formalize, so they aren’t shared. It doesn’t mean that they don’t exist.
- sdenton4 3y agoThis is laughable... Neural networks also have basically unknowable abstractions encoded in their weights. There was some work not long ago which taught an ANN to do modular arithmetic, and found that it was doing Fourier transforms with the learned weights....