I've started writing a thing, for anybody interested in #ttrpg stories: https://ryandlewis.dev/posts/ttrpg/introducingdmslog/
Was struggling to figure out why, even after many minutes and a couple thousand episodes, this RL algorithm I'm trying to implement didn't seem to be improving. Kept checking to see if my code was borked somewhere.
Finally went back to the paper and noticed the time axis of their training graph went from 0 to 200 million episodes lol. Gonna have to burn a lot of watts on this bad boy's education.
RT @sophywong@twitter.com
Y’all… I FINISHED IT
#3Dprinted dress made w/ @Fillamentum@twitter.com PLA in Wizard’s Voodoo
🐦🔗: https://twitter.com/sophywong/status/1446157675717283844
In this optimization problem "spending money to acquire happiness" doesn't necessarily equate to literal spending (could also be saving, investing, donating, anything one could do with money that might directly or indirectly bring more happiness that unhappiness).
Of course, the trouble with this kind of optimization problem is that its hard to quantify and belabored by tons of second and third and fourth order terms and all that.
Realized today that the more precise form of "money can't buy happiness" is actually something like "money's purchasing power with regards to happiness decreases marginally as spend increases".
Through that lens, the optimal amount of money to acquire is equal to the point where the marginal unhappiness gained acquiring money equals the marginal happiness gained from spending that money.
Love talking with my former-Microsoft-employee dad about tech, because he's got some of the most arcane "grumpy old man" tech opinions.
My favorite from today is "wikis are just a stupid branding gimmick". Whatever that means.
There's a very particular kind of person who beefs with the concept of wikis, and boy is he exactly that kind of person.
Hmm...I'm a bit conflicted on the Copilot "training ML on GPL code is a license violation if Copilot isn't also GPL'd!" discourse.
It kinda seems like everyone's confidently asserting that Copilot MUST be a derivative work, which...doesn't seem as cut and dried to me.
If I, a human being, learned to code using a bunch of GPL code samples, would every piece of code I wrote from then on need to be GPL licensed as a derivative work? If not, why is it different if the computer does it?
Very cool 5-min video about a robot that builds structures using fibers: https://vimeo.com/507950769 Would love to see more theses and academic papers with these kind of summary videos/demos
Ryan D. Lewis | I'm a master's student/developer/researcher/roboticist who writes free and open source code in his (admittedly limited) free time.
I try to make the internet a friendlier place!