🧵 Mark. V Shaney

Anonymous at Tue, 18 Jun 2024 11:57:14 UTC No. 16240896

>"Wash the blackboard. Watch it dry. The water goes into the air. When
water goes into the air it evaporates. Tie a damp cloth to one end of a solid
or liquid. Look around. What are the solid things? What are the only ones
that take part in the clouds themselves. As clouds move air tosses the tiny
droplets of water. Sometimes the water which leaves the body in the leaves of
green plants. It is quite porous, almost like a sponge. But the walls of the
hollow tubes and cells are very strong. Chemical changes take place when
something burns...."

> The trick is to apply Shannon's algorithm for Markov chains but with entire
words instead of characters as the concatenated symbols. As MARK V. SHANEY
scans a text, it builds a frequency table for all words that follow all the word
pairs in the text. The program then proceeds to babble probabilistically on the
basis of the word frequencies.

https://www.clear.rice.edu/comp200/09fall/textriff/sci_am_paper.htm
https://en.wikipedia.org/wiki/Mark_V._Shaney

With all the schizo posts on "AI", here's a serious question: How are today's LLMs not just roided up version of the stuff programmers were doing in the 90's? If you throw enough data and compute at the problem and use sophisticated models like neural networks you can simulate intelligence...but it's still just a simulation. It's not actual intelligence. Sure it's surprising what these probabilistic models can generate with images and video et.al but in the end aren't they just more sophisticated versions of Mark. V Shaney?