@davidfowl@Eirenarch@Dave_DotNet In practice it's only 3 years the first time. After the first 3 years your options are the new STS with 2 years or the 1 year old LTS, which also has only 2 years left.
@keyokku LLMs are based on statistics "Which text is likely to follow the previous text?". I think generally on the internet there are more negative texts than positive on AI, also with more details. So in your scenario the LLMs would be biased towards being negative.
@keyokku That is mostly fearmongering, often by people who want to sell you their AI stuff.
LLMs area trained on the whole internet and replicate its content including:
- Fiction about AI uprising
- Posts discussing AI uprising
But in the end LLMs still "just" generate text.
@tsoding That sounds awful, but also doesn't match what I know as tutorial hell. For me it's continously doing the same or similar tutorials around an idea instead of actually working on that idea. E.g "I want to create an FPS" and then endlessly watching "Unity FPS" tutorials.
@YuchiVT I remember @PirateSoftware saying, pre roll ads are hugely detrimental to getting new viewers. This matches my own experience, in the past I definitely left streams of new channels due to first having to sit through ads.
@giffmana Personally I find the clause 1 v even worse: "You will not use the Llama Materials or any output or results of the
Llama Materials to improve any other large language model (excluding Llama 2 or
derivative works thereof). "