Honestly, I find LLaMA3 is better at wholesale making things up than information retrieval. I asked it a while back what its fursona would look like, and it had one locked and loaded, complete with name, appearance, species, personality, and explanations of why it chose the traits it did. (None of those were more than a sentence, but still.) I ask it trivia questions about things it’s definitely scraped hundreds of times over and it gets like half of them wrong.
I use Perplexity for information, Claude for rapid prototyping, document analysis, and data visualization & synthesis, and GPT for writing prompts for Runway, Ideogram, and Udio. I have too many AI subs, lol.
Which is pretty impressive, honestly.
Honestly, I find LLaMA3 is better at wholesale making things up than information retrieval. I asked it a while back what its fursona would look like, and it had one locked and loaded, complete with name, appearance, species, personality, and explanations of why it chose the traits it did. (None of those were more than a sentence, but still.) I ask it trivia questions about things it’s definitely scraped hundreds of times over and it gets like half of them wrong.
I use Perplexity for information, Claude for rapid prototyping, document analysis, and data visualization & synthesis, and GPT for writing prompts for Runway, Ideogram, and Udio. I have too many AI subs, lol.