@sprloveless @codexeditor @anatolideli@ursacke@ahmetb Was going to post this, this is the answer. Also, consider the findings presented from that paper are from a much smaller, sparse model. A full model has much more detection features than this, but it we are not currently able to interpret those features.
@SenSanders Bernie, I'm nowhere near rich, but AI is the greatest opportunity in my life to get ahead.
The real divide is between the tech-literate and tech-illiterate, and it is scary. But in this century, we went from 20% to 86% global literacy. Humanity can and will adapt.
@GregHBurnham This is technically the right response, even if it's beyond its current capability. "Recognize LLM isn't the right model, and build the right one."
@nicdunz It's good for reasoning. It's not better than 5.1 Codex for coding yet. Even their release admitted this, but the 5.2 Codex tuning will come I'm sure.
1. Don't use emojis.
2. NEVER use line breaks between every sentence. Use normal prose.
3. Don't overuse punctuation and formatting for emphasis, especially conjunctive punctuation (— ; : /).
4. Use Markdown formatting as much as you can, especially for ordered lists, bullet points, and tables.
5. Take a forward-thinking view. Tell it like it is; don't sugar-coat responses. Get right to the point. Be practical above all. Readily share strong opinions.
@yugacohler "The End of The World Is Just The Beginning"
One of the most thought-altering books I've ever read. Ignore the doomer title, it goes deep into how globalization works, how important US supremacy is to maintaining it, and what a post-US future could look like.
@k_flowstate The day I fully understood SGD in the context of back propagation was one of the most enlightening days of my life. And there is still so much more to learn!