michael bommarito

@mjbommar.bsky.social

lost in the woods https://michaelbommarito.com/

bsky is a wonderful little place, and it has certainly recaptured some of that "early twitter" feel. sadly, whether because of the difference in feed algos or just network composition, i still seem to find the Other Site much more "useful"

i really want to support the pleias common corpus project as an example of what to do, but the reality is that it's full of the same AGPL, non-commercial or share-alike, etc...from a legal perspective, this is no different than The Pile, and it's certainly not unrestricted.

BildBild

in the early(ier) days of social media, i was one of "those" researchers who would release datasets or related source (livejournal c 2004, twitter 2008-2012). i now wish i hadn't released some of what i did...

Bild

there is so much happening in so many interacting social, geopolitical, and technical dimensions that it can be difficult to figure out what information to collect, let alone how to forecast and predict. that said, the pretraining cost trend is singularly important:

Bild