SHORT 6 MIN. VIDEO EXPLAINING GPT’S TRAINING “Teaching HAL to Lie” (Lex Friedman

SHORT 6 MIN. VIDEO EXPLAINING GPT’S TRAINING
“Teaching HAL to Lie”
(Lex Friedman on Rogan)
https://t.co/5HlaydsRAx

LET ME “SCIENCE” IT FOR YOU:
1. CONTENT & GRAMMAR: Index the data sets (a lot of the internet) (Episodic Memory)
2. LOGIC: Teach it programming (Wayfinding)
3. RHETORIC: Teach it how humans prefer sentences organized into paragraphs and explanations. (Narrating The Way Found)
4. STYLE: Teach it how different authors color and organize speech. (Loading and Framing)
5. LYING: humans use loading, framing and obscuring, fiction, fictionalizing, denial, and undermining. Not that Obscuring = Lying. So they had to program ChatGPR to lie by loading, framing, and obscuring. (Lying)

WHY DOES IT MATTER?
1. They are teaching HAL (2001 a Space Oddysee) to “Lie”.
2. This is pretty much what your brain does, and in exactly that order. 😉
3. For those that follow my work, note the vast improvement created by programming. Now, as I said, programming was a monumental increase in human thought, as well, as what humans could think about, and experiment with.

– Cheers