@penilejones@signulll +1 NYC is a pay-to-play city. It lacks ubiquitous pub/park culture of London, plaza culture of Barca, bistro/promenade culture of Paris etc.
Public socialisation, beyond concentrations in a few heavily-shared open spaces, spills over into food - easily handled by NYC incomes.
@MarkSerbu@Rainmaker1973 Energy doesn’t tell you about the time taken. It’s just a conserved quantity (assuming no friction). An object in freefall vs a gentle slope, same height, shows this.
The calculus of variations finds the optimal path given start/end points e.g the brachistochrone problem
@colin_fraser Solid thread. I think FPR/FNR, not Precision/Recall, should be "the same" (1-X) as Specificity/Sensitivity.
I love the interactive at the bottom of this to see the dependence of the FDR on prevalence.
https://t.co/ztmjICgO9g
With Casual Conversations v2, @MetaAI has adopted the Monk Skin Tone Scale for labeling skin tone in their data set.
Very nice to see @MetaAI take this step and join @GoogleAI in adopting more inclusive ways of handling skin tone in AI.
https://t.co/OaSF9xGDTY
30 Tracks. 70 mins. Largely disco House laced w/ 90s House and a dash of sth sth. Over-engineered in Ableton, as proof of concept.
Some things I learned in the process:
- Default warp mode is not what you want in Ableton. Use "Complex"
- Ableton is only good at identifying BPMs ~70% of the time
- There's this thing called Camelot numbers which aren't hard n fast but helpful
https://t.co/OAcC85TaC0 helps