Ocho

Ocho

[Ocho @Ochotonidays]

What else is lost by low expectations? It "feels" like we're slower to mature, less social/caring, less attentive, less productive, etc. because it's not really expected of us.

[Visakan Veerasamy @visakanv]

just saw a running kid absolutely wipe out onto the floor, and his mom very smoothly said “it’s okay baby you’re good get up let’s go” and his dad picked him up cheerfully. the kid looks completely fine not a single cry or even any tension in the face. Amazing

[Ocho @Ochotonidays]

Listening to podcasts is similar to seeking "alpha" than it is buying the index.

Which is great if you recognize it as such.

[Ocho @Ochotonidays]

@Plinz Lost me at harmonic

[Ocho @Ochotonidays]

@diegocaleiro Child mortality was abysmal for millenia

[Ocho @Ochotonidays]

@sirbayes @kareem_carr This is the logical marketing successor to the “HTML 5 Supercomputer”

[Ocho @Ochotonidays]

@sirbayes @kareem_carr Honesty is the best policy

[Ocho @Ochotonidays]

@srush_nlp @BenevOrang Yeah the “Less Wrong” / Yudkowski stuff is a pretty charismatic sect of the EA religion

[Ocho @Ochotonidays]

@srush_nlp a common conclusion is to attempt promote logic and demote emotions when deciding how to contribute to the world.

[Ocho @Ochotonidays]

@srush_nlp The critique of it being cold and analytical is largely by design. E.g. arguments like Singer’s “A life you can save” try to highlight how our emotions don’t match the morals we’d like to induce, and …

[Ocho @Ochotonidays]

@rdesh26 Dan Povey is everywhere

[Ocho @Ochotonidays]

Making myself legible, but this somewhat mirrors how I interact with my inner critic at times

[shb @himbodhisattva]

@tszzl @nickcammarata ngl what helped me was seeing myself as a complex, but comprehensible, ML system. self-judgement comes from childhood, when you learn high recall strategies for avoiding danger like criticism (which the system perceives as high negative reward). learns low-precision heuristics

[Ocho @Ochotonidays]

@EarlenceF Why the obsession with epochs

[Ocho @Ochotonidays]

@isarth23 Would you rather train 10,000 “epochs” on one example or train on 10,000 examples?

[Ocho @Ochotonidays]

@isarth23 No

[Ocho @Ochotonidays]

@isarth23 If I tell you “I’ve trained a model for 2 epochs” - it conveys no interpretable information without knowing dataset size. But if I’ve told you “I’ve trained a model for 2000 gradient steps” it is meaningful

[Ocho @Ochotonidays]

@joshu Suggestive

[Ocho @Ochotonidays]

Crypto is going to raise a generation that hates capital gains taxes

[Ocho @Ochotonidays]

@AlexGDimakis This is a pretty generous definition of “modular”