Nick Deveau

[Nick Deveau @nickhdeveau]

@lufthansa not a single apology or attempt to organize the situation. I will absolutely not be flying Lufthansa in the future and will recommend that all at my workplace avoid it

[Nick Deveau @nickhdeveau]

@lufthansa creating quite a covid unsafe environment in Munich with zero semblance of organization and professionalism

[Nick Deveau @nickhdeveau]

@lufthansa if you’re going to try to prioritize passengers by flight time, you should make that clear so that an entire jumbo jet’s worth of passengers to SFO aren’t blocking earlier flights...and so the people waiting for SFO aren’t waiting 1.5 hours while others are prioritized

[Cerebras @cerebras]

Look who’s on the cover of @IEEESpectrum’s July issue! Read more about supersize AI powered by our record breaking 2.6 transistor Wafer Scale Engine #AICompute #AIResearch #Innovation
https://t.co/vp1MWTVWQf

[Christina Farr @chrissyfarr]

Mental health, cardiovascular disease, and diabetes are the top-funded clinical indications in digital health, with oncology investments dipping from third to sixth place since 2020. via @Rock_Health

https://t.co/hJdssHEqoz

(Also super cool to be quoted in this)

[Nick Deveau @nickhdeveau]

Taming the Tail: Adventures in Improving AI Economics https://t.co/G9w8rA1V8e #tech via @a16z

[Scale AI @scale_AI]

It is often difficult and costly to achieve strong performance on the rare edge cases that make up the long tail of data distribution.

[Nick Deveau @nickhdeveau]

@ajbosco @ApacheAirflow @DevotedHealth Hey Adam -- would love to learn more about your process for integration testing and data validation. You touched on it, but this is something that could warrant an entire piece(s) in and of itself.

[Nick Deveau @nickhdeveau]

@stuz5000 Definitely going to play around with it, if no other reason than just curiosity!

[Nick Deveau @nickhdeveau]

@stuz5000 Intrinsically explainable models are on the surface a nice antidote to fear of black boxiness, but I’m interested in what it is about surrogate models that can feel “second class” in some, often high stakes, applications.

[Nick Deveau @nickhdeveau]

@stuz5000 The most interesting question brought up here, though, is one of what actually allows us to trust decision making? And why do we arrive at that trust standard?

[Nick Deveau @nickhdeveau]

@berkustun @AlexanderSpangh @berkustun this one really got me thinking...do you think actionable recourse is a property that should necessarily exist in all models affecting humans?

[Nick Deveau @nickhdeveau]

@Calsoft_Data Sounds like this champions algorithms that are intrinsically explainable, but what about building explanatory layers on top otherwise black box models? How will those be accepted by explanation-driven fields like healthcare?

[Nick Deveau @nickhdeveau]

@iamtrask We work on making the hiring predictions less biased (by traditional "signal" like where you went to school) and more predictive.