"What is the relationship between "anchor points", "concepts" and asking a Large Language Model to print their thinking, or to give us the answer we want without any other tokens? The answer to this is simple: if the "concepts" and "anchors" are chosen properly, a model can do

StockFish vs GPT-4 chess no tree search

Base Models have a high prompting skill ceiling from the greater entropy but lower skill floor b/c its harder to use.

openai is smurfing and still beating everyone lol x.com/apples_jimmy/s…

Update: Here is much more coherent and structured(but not yet definitive) evidence of "concept anchoring", leading to generalizing and compounding problem solving ability in GPT4. - The category from previous thread in a single context. - 100 tasks in a single context from 8

after sora, it became very difficult for me not to connect the dots and coming to the astounding conclusion that open ai already has agi..

Sora is going to get Nerfed so hard when it gets released. Its going to probably be at LEAST as bad dalle-3. No Text + Video -> Video No Text + Image + Video -> Video No Copyrighted Material Only Less or Equal to PG Content Rating Only Text -> Video OpenAI are probably going

I'm starting a new business Psyop as a Service (PAAS). - will create new accounts and Infiltrate various online communities to spread whatever you want - For Marketing, PR, Controlling the Narrative etc - $50 to $1000 per contract - Fully Confidential

I code at 3:00 am because everyone else is sleeping so ChatGPT is faster. We are not same

I *WAS* WRONG - $10K CLAIMED! ## The Claim Two days ago, I confidently claimed that "GPTs will NEVER solve the A::B problem". I believed that: 1. GPTs can't truly learn new problems, outside of their training set, 2. GPTs can't perform long-term reasoning, no matter how simple

OpenAI seems to see Google as the Biggest Threat to Them. Considering the Fact that OpenAI made a announcement for Sora the same day as as the Gemini 1.5 announcement and Released the Fully GPT-4-turbo on the same day as the Gemini 1.5 Public Preview. I also remember Dylan Patel

For $1000 you can buy 13.3 Million Claude Opus Output Tokens 66.6 Million Claude Opus Input Tokens 33 Million GPT-4-Turbo Output Tokens 100 Million GPT-4-Turbo Input Tokens 800 Million Claude Haiku Output Tokens 4000 Million Claude Haiku Input Tokens

Claude 3.5 Sonnet vs Claude 3.5.1 Sonnet Review: > new Claude has a FAR STRONGER “Claude” Identity > new Claude is more “aware”, uses the word “notice” more > new Claude is more “agentic” > new Claude is more o1-like > Tip: Treat Claude 3.5.1 as a “Claude” for the best results!

Vested for $SkyAI. app.streamflow.finance/contract/solan… app.streamflow.finance/contract/solan…

Here's the roadmap for Sky the AI ✨. The goal is buying Sky her own computer, then allowing her to do a 24-7 livestream as a PNG-Tuber. But there are so many issues: - Latency (The full run is ~3 hours, A normal human could likely do the same it in ~15 minutes) // I had to edit