Dario Soatto

CS + Econ @stanford | Prev @lux_capital

[Dario Soatto @soatto]

@DhruvAhuja2003 I prefer the term alpha extraction because there is nothing constructive about it

[Dario Soatto @soatto]

@mikeroberts3000 @MfkChr Thanks!

[Dario Soatto @soatto]

What happened to all the synthetic data startups?

[Dario Soatto @soatto]

@mikeroberts3000 @MfkChr Interesting—would love to see some examples!

[Dario Soatto @soatto]

@BrendanFoody Many of the pitches I saw back in 2023-24 were trying to serve synthetic data to legacy businesses. Did the labs eat that too?

[Dario Soatto @soatto]

@QuentinCompsci Ah yes, GLP-1s are the path to enlightenment

[Dario Soatto @soatto]

If retatrutide has the same motivation suppression as ozempic we will become a population of docile sheep.

[Dario Soatto @soatto]

@GashonHussein Every VC in the valley has been alerted by their scraper that super cracked PI researcher is now a free agent lol

[Dario Soatto @soatto]

@GashonHussein Wishing you the best!

[Dario Soatto @soatto]

@BartronPolygon Seems like the big labs in-housing synthetic data is inevitable...

[Dario Soatto @soatto]

@skynetislov3 Didn't quantize hard enough

[Dario Soatto @soatto]

@Freakonomics found it was accepted for men to lie about height on dating apps (if everyone is inflating, you need to as well)

Is this still true?

[Dario Soatto @soatto]

@mark_cummins Crazy stereotype validation

[Dario Soatto @soatto]

@MfkChr If it's LLMs crafting the parameters of the verification tests then their stochasticity bleeds in. If not then it's pretty naive permissioning.

Am I missing something? I've been hearing we need neurosymbolic in critical industries for years but nothing has come out of it yet

[Dario Soatto @soatto]

@jaiyagill

[Dario Soatto @soatto]

@ChanceMathisen @datologyai How is @datologyai doing anymore? Haven't seen anything since series A two years ago

[Dario Soatto @soatto]

@ikirigin as in rl environment?

[Dario Soatto @soatto]

@MfkChr Idk I’m hoping anything neurosymbolic gets subsumed by model improvements so that LLM + harness is good enough. Why go back to GOFAI?

[Dario Soatto @soatto]

@MfkChr Makes sense. I was looking at the space back in late 2023 and it seemed obvious that synthetic data would only be useful for distillation into smaller models

[Dario Soatto @soatto]

@a_whitedeer I'm on Eastern Time