gbrown

39 ms·

361.
by gbrown 7y ago
You vastly underestimate the complacency of average consumers.
362.
by gbrown 7y ago
I thought there was little reason to expect a big rip, because it requires assumptions about an increasing cosmological constant, which there’s no evidence for.
363.
by gbrown 7y ago
People are way too uptight about lawns
364.
by gbrown 7y ago
A colleague of mine had one and is a real serious mountain biker. Science is awesome.
365.
by gbrown 7y ago
Any operating system that plans to stop supporting the hardware on devices worth hundreds of dollars (sometimes more) after a mere two to three years is deplorable in my book. I really dislike the iOS experience and its lack of customizatio
366.
by gbrown 7y ago
Part of the problem is that inadequate training data may not be a barrier to successfully marketing and implementing these systems - the tech is sexy and people are rightly excited. > Doctors make mistakes all the time and people in the
367.
by gbrown 7y ago
I think there's a huge potential for patient benefit if this is done correctly, though I worry that we'll just create mostly adequate classifiers based on existing data. I also think we need to watch out for the human-attention is
368.
by gbrown 7y ago
It’s the estimated (via simulation) sampling distribution of the number of heads observed out of three hundred coin tosses.
369.
by gbrown 7y ago
Well put - that's exactly how I feel.
370.
by gbrown 7y ago
> If you're in or adjacent to academe, you'll have to travel to conferences. I really think we should be working to change this - conferences are huge wastes of resources IMHO. I usually see a handful of inspiring talks, which
371.
by gbrown 7y ago
If the deliverable is a team report, then failures are super useful and still count. "I tried implementing X as a test after reading about method Y, but found that for our domain there are serious drawbacks."
372.
by gbrown 7y ago
Tie it to deliverables, like reporting back to the team about the content or giving seminar style talks.
373.
by gbrown 7y ago
You don't - you compare it to "standard of care" (antidepressants, CBT, etc.). If we can compare an inferior treatment to placebo, and we can compare a breakthrough treatment to this established standard, we can evaluate the
374.
by gbrown 7y ago
crosses fingers in OpenCL
375.
by gbrown 7y ago
Maybe it's my particular setup, but I've had nothing but problems using a Windows 10 workstation which I routinely remote into. Graphical scaling is unstable and screwy, I constantly have to select "reset view" in outloo
376.
by gbrown 7y ago
Interesting. I've never had trouble with the bog-standard MS keyboards that happen to come with most office workstations. I do like the feel of hybrid mechanical-rubber keys though, the click is nice. My wife says my typing looks reall
377.
by gbrown 7y ago
I never have any problems with wrist fatigue, and I've sometimes wondered if it's because I type "wrong". Holding your fingers at home row and stretching always felt weird to me, and I learned to type naturally. I primar
378.
by gbrown 7y ago
> Thought experiments are relevant because they can tease out our moral intuitions. I understand this, but I don't agree that your thought experiment usefully does so. You're essentially begging the question: "Well, what i
379.
by gbrown 7y ago
It's an important concern, but I think mitigatable - especially since we're not neccessarily talking about actual self contained replicating machines, and instead more likely automated factories producing drones - a multi step pro
380.
by gbrown 7y ago
With appropriate organization and accounting, there's no reason malfunctioning units couldn't be repaired or reprocessed. It seems like that would be a sensible regulation for companies building self replicating machines.
381.
by gbrown 7y ago
I disagree with your reading entirely: "It may well be the case that the gender split is 80% / 20%" That's a preposterous hypothetical. You're also incorrect in your implication - the existence of minor (and as yet
382.
by gbrown 7y ago
1. I feel like your flip side is a straw man. 2. Your description of algorithms and how they can/should behave is completely beside the point. Thats the problem with vague marketing language like "AI". The article isn't
383.
by gbrown 7y ago
Meh, as long as they're not designed to improve themselves, and there are strict protocols for ensuring all active units are controllable, it's not necessarily a recipe for solar system paperclipization.
384.
by gbrown 7y ago
"Assume a perfect and impossible thing, which is not precisely defined enough to formally reason about , but which nevertheless by construction supports my point, and you'll see that I'm clearly correct." "controls
385.
by gbrown 7y ago
> When would you want to hack the AI? This is not a meaningful phrase - it adds literally nothing to the conversation but confusion. To bend over backwards to give you a reasonable answer: when you're interested in conditional effec
386.
by gbrown 7y ago
Blame the "left leaning zeitgeist" all you want, I've seen no such evidence. What I have seen is a Hacker News thread full of people misunderstanding statistics and machine learning, advocating for bad models, and obsessing a
387.
by gbrown 7y ago
As always, it depends entirely on what you're doing with the (predictions/estimates/inferences). This is an entirely false dilemma.
388.
by gbrown 7y ago
> Say you have an AI that predicts probability of re-offending for prospective parolees. Already you've made one of the most common mistakes in this area: confusing the label for the latent truth. We don't measure "re-offe
389.
by gbrown 7y ago
I totally understand, and maybe I shouldn't have included that, but I do see phrases like "It's unpopular to talk about" get used as dog whistles all the time in other forums (and in real life). There's a whole subc
390.
by gbrown 7y ago
> The goal is to build a model that accurately reflects the real world. This is incredibly reductive - we use models for MANY purposes - prediction, inference, description, evaluation. Until you outline a specific use case for the model,

More ›