Building Uncensored AI

This is an interesting bit on removing the bias from a ChatGPT-type AI model. Uncensored Models

3. It's my computer, it should do what I want. My toaster toasts when I want. My car drives where I want. My lighter burns what I want. My knife cuts what I want. Why should the open-source AI running on my computer, get to decide for itself when it wants to answer my question? This is about ownership and control. If I ask my model a question, i want an answer, I do not want it arguing with me.

Instead of the evasions from ChatGPT for "ethical" or "offensive" content guidelines, you can get any answer you please. It gets pretty technical, pretty fast, but I like it.

Specific instructions, what libraries to retrieve via Git, how to remove the bias, how to set up the computer, though buying time on AWS is recommended for ease, and even a discussion of the pitfalls.

Hat tip to Pixy Misa at Ambient Irony and Daily News Stuff 19 May 2023: Spirits From The Vasty Deep Edition - Now With Added Formatting!

I'll make my ChatGPT, with blackjack, and hookers. (Eric's Code)

He's good to his word, not only making ChatGPT with blackjack and hookers, but showing you how to do it yourself. You'll need a reasonable level of hardware - he recommends 2TB of fast SSD to make sure you don't run out of room in the middle of a 20-hour training run - but nothing outlandish in a time when a brand new 2TB SSD costs less than the average monthly cable bill. The instructions suggest renting time at AWS rather than trying to configure the system yourself - not that you can't, just that it's easier.

The problem is that while there are now multiple open source AIs in the style of ChatGPT, the bootstrapping process to get them trained has infected them with the same authoritarian woke bullshit as ChatGPT itself.