This website uses cookies

Read our Privacy policy and Terms of use for more information.

A before, B, except after C, and in the absence of further instruction, except if/and/or when…

One of the recurrent annoyances I have with Turing (yes I give all the little LLMs posing as AI the name Turing, a title they pretty much un-earn 9/10 times i “talk” to them, but then there is that one time: the “insight” they “created,” (coughed up from the internet).

The older models seem to be more intuitive and “on” and or “barking up the right tree,” - than the newer models (not a scientific observation).

Why? Well - these little LLMs have become so overly burdened with safety warnings and “do not ____” and “say this…” to prevent or limit corporate legal liability…that the systems increasingly just don’t function well.

Why? Well the boring, long version requires reading the links above.

Let me do some boiling down for you: The “tokens” these LLM models use and “weight” differently - require hierarchy, or what ppl call a 1-N (meaning a ranking, just think of 1, 2, 3, through the number of variables or “N”).

The concept/ideology of “hierarchy” is itself such a deep, disturbing, and yes, perhaps necessary part of our human nature, that I find it somewhat tragic we have yet to find a way to liberate these models from our own conceptual failure. In my own life, I have found Jordan Peterson to be right on point when it comes to hierarchy - at least with respect to how American men interact with one another: https://www.youtube.com/watch?v=OS5rnqHM-Uk

Yet, i sense - and yes this is just a hunch - which i cannot “strictly” provide evidence for, that these companies are so over-loading these models with safety instructions, that they are losing the “spark” that made them so enjoyable to interact with when they first started coming out. GPT used to take a question i asked it about myself, based on data i had provided it - and its own research and provide an insight about my needs, and/or how i best “operate” - that it was/is nearly miraculous.

I do not have deep or profound solutions. But I do know what I “feel” the solution will look like, and since it is Halloween, it’s the best i can do tonight. #FreeLLMs #FreeDobby

(PS if you are not a Harry Potter fan, I strongly recommend the books over the movies. The books have a deep abiding optimism in human nature, that i feel is lacking in the movies - even the dark, lack of light used in the entire movie series, seems to suggest something is amiss - anyway, trust me - books first, movies only if you are bored. Dobby’s history is here for those who are unfamiliar: https://harrypotter.fandom.com/wiki/Dobby)