Resurfaced 2020 GPT-3 paper details how scaling to 175 billion parameters unlocked in-context few-shot learning
Task accuracy improved with more in-context examples without fine-tuning.
Many users expressed hindsight regret over missing Nvidia investments after the GPT-3 paper showed large models' potential, rather than reacting directly to the research itself.
No Digg Deeper questions have been answered for this story yet.
Most Activity

@dumbfook idk why I didn't buy Nvidia back then 😭

@dumbfook

@dumbfook i should have read this instead of rolling snake eye on the nicotine post

@dumbfook should have paid attention to this instead of playing dota

@dumbfook havent seen hardmaru on my tl in a while, weird algo

@dumbfook @DanielleFong pure sauce

@DavidSHolz @dumbfook i kept thinking that nvidia was overvalued, that other chips would come out. it always feels like a risky buy but it was actually a generational buying opportunity several times over. the custom chips are getting there but so far underwhelming in terms of their impact on $NVDA

@DavidSHolz @dumbfook 🪦

@yeti_0x @dumbfook Sakana is doing some great stuff. Marlin (https://sakana.ai/marlin-release/#English) looks incredible, but sadly I'm not rich enough to use it.

@DavidSHolz @dumbfook You’re probably early to SpaceX though

@DavidSHolz @dumbfook you can still buy $SPCX

@abyssalblue_ @dumbfook We do not speak its name.

@DavidSHolz @dumbfook I sold Nvidia after the Bitcoin crash! Oops

@DavidSHolz @dumbfook you and me both 😢

@vvkgopalan @dumbfook Well they did advertise themselves to the dota folks back in 2017…