A friend did a fun demo for Claude 3.7 vs GPT 40 standup comedy! Humor is pretty good eval for how good the AIs are. Maybe this is be part of the official evals - who brings out the most laughs per token:D
@FreyaHolmer There is a way to name quaternions so they are easier to handle: "a_From_b"
When multiplying, their "connectors" must match:
a_From_c = a_From_b * b_From_c;
So the "b"s cancel each other out, similar to a/c = a/b * b/c
This is explained in the YT video with ID QI8U67CT42w.
@ESYudkowsky I doubt even a super intelligent AI will ever be able to predict the specific words we will say. Wolfram's concept of computational irreducibility applies here. Even though we have vastly more brain cells than ants, we cannot predict precisely what an ant will do.
How about having an AI-Wingman!
This is gorgeous experiment about how a LLM can be a party assistants for nerdy introverts like some of us 😎
People are still underestimating the impact #GenAI will have when combined with (upcoming #Apple) #VR headsets.
@YuriAOli On GitHub, check srcnalt's OpenAI-Unity for OpenAI, and SethRobinson's GPTAvatar for ElevenLabs and the general setup of this kind of avatar.
This is really cool demo of how the AI is steadily integrating into our life; LLM based intelligent NPCs will be our best friends, tutors, colleagues and maybe even lovers!
@sethicusprime Yes, I plan to write a post eventually about these experiments. I made a video about one with a Christian and a Buddhist monk: https://t.co/jz4Y7P8fJw