a short story on learning without context
Long ago, there was a snail. When it was hungry, it would look for food. In service of this simple need, it crawled around pebbles and other obstacles, resulting in intricate paths of slime — evidence of the “snail-ness” left behind.
In the present, only the slime trails remain. Snail scientists studying these trails made a remarkable discovery: one could accurately predict where the slime would turn next by looking at the slimes that came before it. Soon, predicting the paths of these slimes becomes trendy research, with competitions on who can best predict the trajectories of these slimes.
“Clearly this is a data and compute problem!” the scientists declared. So they got to work. An abundance of slimes were codified , consolidated into the colossal clean crawled corpus. The computers burned with enough energy to feed every snail family that ever lived, reasoned about inscrutable slime paths that no snail could fathom. Yet, the slime trails remained unpredictable — Sometimes the slime would inexplicably form little arcs, and other times it would abruptly change its direction. The scientists have dubbed this the “base entropy of the slimes”.
If only the scientists would be able to see the snails and the world it lived in. The little arcs was when it had to crawl around pebbles, and the sharp turn was when it suddenly smelled food. By modeling the evidence of intelligence rather than intelligence itself, the scientists have failed to catch up to the humble snail.
Thanks for reading!
— evan
