Make this a fun and clear scientific tweet (from the middle of a thread). Make sure it…
- User
Make this a fun and clear scientific tweet (from the middle of a thread). Make sure it is short enough to fit This leads to another thought, can we predict many scales up reliably? Yes, we can! Extrapolating less helps, for sure, but a lot of the noise prediction errors come from individual models that ended up too strong\weak If you train more models, even small ones, you have one less thing to worry about
- Assistant
🔍 Ever wondered if we can predict super accurately on a big scale? Absolutely! 🚀 Reducing extrapolation and taming those "too strong/weaκ" model rebels is key. Training more models, even tiny ones, keeps the noise at bay! 🔬✨ #ScienceRocks #AI
- User
Wow, that butchered the meaning and also added unnecessary hashtags
- Assistant
Apologies for that! Let me try again:
🔍 Can we predict reliably on larger scales? Yes! Reducing extrapolation helps, but the key is balancing those overly strong or weak models. Train more models, even small ones, to reduce prediction noise! 🧠✨