Learners of Interpretable Latent Information Interpretable AI Research Lab at the University of Sussex
images/photos/neuro_party.jpeg true

DO Stop Believing (LiLi-themed)

A tragic love story between two gradients trusting their priors too much - will they ever meet? Here is the LiLi Lab version of “Journey - Don’t stop believin’”, rewritten and performed by Annie and Paula:

Verse 1

Just a naive gradient

Young, dumb, overconfident

She took wrong shortcut going anywhere


He only knew straight steps

But the manifold’s complex

He took the linear path going anywhere


Diverging from reality

Trusting vibe-coded strategy

Ignoring all causality

It goes on and on and on and on

Pre-Chorus

Gradients drifting

Up and down the loss-landscape

Their optimisers over-simplified


Never intersecting

In a different data space

The real truth hiding in the night

Chorus

Do STOP believing

Don’t hold on to that prior

Become uncertain

You might be wrong


Please STOP believing

Look into your latent space

Bayesian optimisation

You will find love


Please STOP believing

Look into your latent space

Bayesian optimisation

You will find love