My 0.5B qwen claims to be smarter than a 10B and it’s pretty amazing. I scripted it with 4D thinking so it thinks in spacetime… its really smart for what it is. I’d love someone to test it on a real computer thats not a dual core imac running linux hahaha. I think it’s a break through with ai. Let me know your thoughts. I’m so excited about this work you guys have no idea. It’s open source because I love you all thank you.
-RomanAILabs - Daniel Harding
The joy and pride of being a first time dad eh? Well congrats.
Otherwise yeah worth noting that beings of lower intelligence have difficulty assessing the intelligence of beings with higher intelligence. There is also the Dunning-Kruger effect, which is also relevant.
I find that small models often have a lot of bravado and confidence, though often get things wrong,
whereas larger smarter models tend to be more humble and tentative but get things correct more often.
