nmca

7 ms·

301.
by nmca 6y ago
No it isn't, in this case. I think Taleb has the best summary of this, for all his flaws: https://www.youtube.com/watch?v=9dKiLclupUM In short, risk due to coconuts vs sharks is a valid comparison because we have good
302.
by nmca 6y ago
I don't think this paper is very good. Because of the huge age and co-morbidity influence in covid mortality, the very large number of excess deaths correspond to a comparatively (eg compared to median death) small number of QUALYs. Bu
303.
by nmca 6y ago
Irrc we were using a Bayesian classification model on top of of fixed pretrained features from transfer, something along the lines of refitting a GP every time the number of classes changed. This was images as opposed to text, and after an
304.
by nmca 6y ago
Hey Mike - I did some work on an industry active learning system a few years ago. The high level finding was that transfer & nonparametric methods were huge wins, but online uncertainty and "real active learning" hardly work
305.
by nmca 6y ago
An entire section of society so precariously clings to financial stability that they'll smile at at any customer's joke...
306.
by nmca 6y ago
Robinhoof at least seem to agree lol: https://blog.robinhood.com/news/2021/2/2/its-time-for-real-t...
307.
by nmca 6y ago
Because the "no hospitalisations or deaths" claim from the first Oxford paper was comparing n=2 to n=0 and thus was miles from significance. If the new paper has updated data on this happy to be corrected.
308.
by nmca 6y ago
Somewhat offtopic I guess, but a huge part of this seems to be two-day settlement. Why on earth does the US still have two day settlement? Surely we're at a point where we ought to be able to send money nearly instantly for ~free. In t
309.
by nmca 6y ago
Eh, I doubt it's an ad - I left hsbc many of the same reasons. Final straw was trying to get my data from them, which they are obligated to provide in some sensible format and simply wouldn't. Indeed after a couple if escalations
310.
by nmca 6y ago
This is known as a "treacherous turn", a phenomenon I'm aware of. But I don't really see how that's relevant, my point is that a lack of physical grounding or pragmatism can lead to spurious conclusions about the su
311.
by nmca 6y ago
I haven't read this properly yet, but a skim leaves me skeptical. For example: > "Another lesson from computability theory is the following: we may not even know when superintelligent machines have arrived, as deciding whether
312.
by nmca 6y ago
This looks very good from my skim! It includes state-of-the-art approaches like Actor-Critic with MCTS and loads of recent and interesting things like GAIL and SMILE. The areas where I have expertise superficially seem like good treatments.
313.
by nmca 6y ago
Imperial study was in the context of a tiered UK system that keeps schools, so increased infection from young people is my personal and ill-informed hypothesis of how the new variant achieved 50-70% increased r0 under restrictions.
314.
by nmca 6y ago
But London is rather un-american in its practicalities right? It's a total expensive pain to drive, as roads are narrow and there's a 20usd a day fee to do so. But public transport is great and wherever you live there are walkable
315.
by nmca 6y ago
I think this analysis is a bit oversimplified. If your goal is to build houses, then local labour is clearly better. But if it's to recruit lifetime donors, then perhaps shipping out incompetent (but fabulously wealthy) Americans is th
316.
by nmca 6y ago
"Enabling the behaviour that's doing the most harm" - I'm somewhat skeptical of environmental viewpoints that hinge on individual consumer guilt. As one random reference point, a single cruise-operating corporation was r
317.
by nmca 6y ago
Thanks - I re-read this. There was one death in the control arm, which was one of two severe cases (table 5). No statistical analysis is presented of those results (if they were significant under any straightforward model I suspect it would
318.
by nmca 6y ago
Pfizer is storable at dry ice temp for ~months and regular vaccine fridge temp for 5 days [0] so whilst "large teaching hospitals" is currently accurate I don't think we're married to it. [0] https://www.pfize
319.
by nmca 6y ago
Ignoring herd immunity, there is also a reduction in mortality rate as older people are vaccinated. Do you have any sense which of the two effects dominates, and any citation for the 70% number? Thanks for the response :)
320.
by nmca 6y ago
Can you provide citations for this (first sentence)? My understanding is that there were no deaths in the treatment arm of the trial. But deaths are much rarer than infections so naively our uncertainty on this effect should be much larger
321.
by nmca 6y ago
Can anyone point to an explanation of how much the 60% Oxford vaccine will help things, both in the old case and in light of the new variant? I have approximately 0% intuition for how much a 60% vaccine helps.
322.
by nmca 6y ago
Sure, but in a democratic system with perfect information you should expect to lose the vote on your hypothetical "me vs 1000" trolley problem right? And in the absence of perfect information you'd I guess you'd mount a
323.
by nmca 6y ago
Most models have the population capping at ~10B, so people don't tend to worry about it any more? (10B number is recalled from "Factfulness", which I'd recommend)
324.
by nmca 6y ago
I think parent means for goods sold in the EU - and they correctly point out that we have lost direct influence over the regulations in a collosal trading block. (The effort was to ensure that we didn't have to follow the eu regs in th
325.
by nmca 6y ago
Pet peeve of mine when people call these types anything other than f32, f64 etc. Surely we should have learned by now!
326.
by nmca 6y ago
It's hard to see what you're getting at here; the sets of people that pay are not the same.
327.
by nmca 6y ago
How great would the benefit to humanity be, pricing in uncertainty of a practical legalisation scheme, before you'd be happy for the cartel leaders to get off scott-free? I haven't done the numbers in this case, but surely there a
328.
by nmca 6y ago
Short selling is good for price discovery, and an inability to short contributes to bubble dynamics. So the speculation has value!
329.
by nmca 6y ago
My understanding is that that trial condition (half->full) was administered accidentally (thus not pre-registered) and nobody expected it to work better, and the sample size is small and on the wrong age groups.
330.
by nmca 6y ago
https://deepmind.com/blog/article/alphafold-a-solution-to-a-...

More ›