shutting the minds bg running tabs off and just focusing on a hackathon project is a practice i am liking now.
you are in the moment and don't care about anything about the outside world for this 36 hours.
kinda peaceful
i hate this feeling when i am haflway through a snack and the perfectly complimenting coffee that i got ready is about to finish, and there is no way to manage coffee to last till the end of the snack.
prayush retweeted
my feed at this point is basically a car flipping upside down and gujjus spinning round and round
the act of helping someone can sometimes mean more to the helper than the person being helped.
if the recipient is made to feel the burden of owing a favour, something about the spirit of helping is lost.
to be able to help someone is an opportunity in itself. an opportunity to be of use to another person, to contribute something to the world around us.
and that's something the helper should feel grateful for, rather than taking pride in having done someone a favour.
"my current address is the rms titanic and after that i am on gods good humour"
never not intrigued by this scene from titanic
was just thinking about this.
how we were concerned for the last decade on how meta, google, etc are tracking people at all the external surfaces. almost in a sense intruding the privacy by listening, reading activity at all possible instances.
and with muse, we are inviting them even more. in all the praise for how good a product muse is, i rarely saw any discourse on tl surrounding the privacy concerns
privacy is no longer something two ppl can mutually assume in a conversation, physical or digital. ai meeting note takers changed the norms first, & now your emails, texts, maybe even photos are being ingested into who knows which ai model & which product. if you delete it on your end, it does not mean anything & it could be on anyone’s machine. & it’s perhaps likely training at least some models.
i got to know that the compute being spent on pretraining is getting way less compared to post, and the analogy here is that pretraining is the preliminary schooling and rlhf is the skill building phase.
i wonder will the pretraining compute ever go below a kid's school years.









