WEBVTT

NOTE AI Engineering Visualised: https://understandingdata.com/ai-engineering-visualised/superposition/

1
00:00:00.195 --> 00:00:14.016
One neuron need not mean one feature: start with two independent directions and two coordinates, where activating A by one gives the exact readout one, zero, without inventing activity in B.

2
00:00:14.059 --> 00:00:28.765
Now keep only two coordinates but arrange five feature directions around a pentagon, and send the same A through them: its hidden state is still one, zero, although no coordinate belongs exclusively to one feature.

3
00:00:28.765 --> 00:00:44.143
Reading back with dot products exposes the interference: A returns one, but the inactive neighbours B and E each return zero point three one, so even clipping negative values does not recover a clean feature list.

4
00:00:44.143 --> 00:01:00.347
Making features rarer leaves those overlapping directions unchanged, but reducing each independent activation chance from fifty to ten percent makes a pair collide twenty five times less often, which changes the cost of sharing space.

5
00:01:00.439 --> 00:01:18.398
A sparse decoder searches for a short explanation of the hidden state: in this fixed dictionary, real optimisation with a small activity penalty leaves A near zero point nine five and suppresses its neighbours, trading a little reconstruction error for sparsity.

6
00:01:18.559 --> 00:01:36.640
Finally, all five equally active directions cancel to zero, exactly like no features at all, so no decoder can distinguish them here: sparse explanations are useful hypotheses, and low reconstruction error alone never proves that their features mean what you think.