"There are many things one needs to live a rich and fulfilled life (according to AI researchers). A good initialization [Mishkin and Matas, 2015], attention-based neural networks [Vaswani et al., 2017], and a good title for your research paper [Myself, just now], to name a few.
In this post, we discuss another piece of eternal wisdom from AI researchers: “less is more.” Specifically, how foundation models can be fine-tuned for new capabilities with small data, in many cases less than one-thousand samples, and often outperform the same model fine-tuned on larger datasets. Meditate on that for a moment (suggested pose in figure above)."
First couple of paragraphs:
"There are many things one needs to live a rich and fulfilled life (according to AI researchers). A good initialization [Mishkin and Matas, 2015], attention-based neural networks [Vaswani et al., 2017], and a good title for your research paper [Myself, just now], to name a few.
In this post, we discuss another piece of eternal wisdom from AI researchers: “less is more.” Specifically, how foundation models can be fine-tuned for new capabilities with small data, in many cases less than one-thousand samples, and often outperform the same model fine-tuned on larger datasets. Meditate on that for a moment (suggested pose in figure above)."
Less (but higher quality) is more :)