bertil

36 ms·

391.
by bertil 3y ago
For instance, all commercial A/B testing services (as far as I can tell) compute the 99%-Winsorized stats for their reports. This is partially due to several common cases that can mess up an average computation, especially between two
392.
by bertil 3y ago
What do you mean by honey pot? I only know the meaning where it is a deliberate thing to attract bad actors. Is the entire company not meant to provide genetic information?
393.
by bertil 3y ago
The best option is to use a similar process to HB, but using “green” H2, so yet more inefficiencies from “excessive” renewable electricity generation. https://royalsociety.org/topics-policy/projects/low-carbon-e...
394.
by bertil 3y ago
I'd be very surprised of the government decided to subsidize hydrogen without Toyota's input.
395.
by bertil 3y ago
Your model needs to introduce a third piece of information: whether an email is a corporate communication—or a deliberate intervention.
396.
by bertil 3y ago
You seem to assume that they would update them as they sell them—at the same time. I’m not sure how long the update would take, but I’m assuming longer than many customers would like to wait.
397.
by bertil 3y ago
But there would be no clear way to tell which box was done, so you rely on proper inventory—or even when updating, to tell when it’s done, presumably.
398.
by bertil 3y ago
I don’t know if it’s deliberately confusing, but it certainly was reviewed by several lawyers.
399.
by bertil 3y ago
It seems like the solution would require employees to put the boxes on pads one by one… That seems to be a tremendous amount of work. I have not been impressed by the quality of the service in-store (waiting more than an hour for basic requ
400.
by bertil 3y ago
> Or do I need to provide that graph to this function? You need to do that, and the math can help you measure how much each arrow contributes. The idea that you need to provide your model of the world is strangely not a key part of most
401.
by bertil 3y ago
That’s generally because the metrics you are looking at do not represent what users care about. It’s different than the testing methodology, often overlooked, and a lot more important. I’ve argued that A/B testing training should focus
402.
by bertil 3y ago
All the alternative methods require the same sacrifice. More importantly, most suggested treatments fail to cure deadly conditions or have major side effects or risks that are just as unethical to thrust upon people untested. If you look at
403.
by bertil 3y ago
I am a big fan of what Sean and you are trying to do–I wrote up a chapter about it this weekend, actually. I’m worried that you both have worked for companies where a lot of work has been done to identify relevant dimensions (metrics and ca
404.
by bertil 3y ago
Causal inference is useful, but it's neither quicker nor cheaper.
405.
by bertil 3y ago
The main reason for not using causal inference is not because data scientists don’t know about the different approaches or can’t imagine something equivalent (a lot of reinvention); forecasting is one of the most common tasks, after all. Th
406.
by bertil 3y ago
I usually compare those to flights that arrive conveniently early for business travelers and cheaper, less business-friendly flights combined with a hotel night; night trains usually win when there’s a convenient option.
407.
by bertil 3y ago
The article gives very little detail about how the train is set up. There are some interesting ideas about how to preserve individual privacy within the same form factor. Those are probably better explained here: https://youtu.be
408.
by bertil 3y ago
Yeah, simple models work too.
409.
by bertil 3y ago
This is entirely due to people not testing the impact of too many marketing notifications on whether people keep their notifications on—partially because Apple doesn’t share that information with developers, so they have to guess based on w
410.
by bertil 3y ago
Apple would be gaining so much goodwill and transforming the industry if they asked for feedback on notifications, and trained an ML system to decide which one to show.
411.
by bertil 3y ago
If the emails addresses are registered in Apple Mail: > I only ever use those email addresses in Apple's Mail app then they are associated with the matching iCloud account.
412.
by bertil 3y ago
They have to “boil” but not get the water to 100ºC. Water evaporates in the air at any temperature; it’s faster when the water is warm, and the air is warm and dry. Technically, that’s boiling, even if it’s not exactly like how your kettle
413.
by bertil 3y ago
I’ve always wondered how hard it would be to have standard solar panels water-cooled and use that heat for hot water storage.
414.
by bertil 3y ago
Good idea, but it probably makes a very cheap system more expensive — kitchen-equipment expensive.
415.
by bertil 3y ago
That's obviously the way to go, but… Where does the 99% come from?
416.
by bertil 3y ago
I am tempted to agree: if separating the two is key (I’m not convinced that it is, but happy to assume) why not copy the interface and the infrastructure of the feature flag and offer it as a configuration tool. I feel like you could easily
417.
by bertil 3y ago
Yes! Thank you so much. I’ve been trying to remember where I read that for almost a year now.
418.
by bertil 3y ago
Does anyone here remember an article comparing the Industrial Revolution to the internet, and the Gin craze to social media, and, incidentally, Wikipedia–wasted resources because those had become suddenly worth less: adult workforce in the
419.
by bertil 3y ago
No, the risk of uneven bucketing of more than 1% is minimal, and even when it’s the case, the contamination is much smaller than other factors. It’s also trivial to monitor at small scales. False positives do happen (Twyman's law is th
420.
by bertil 3y ago
You are confusing: a. the Family-Wise Error Rate (FWER what xkcd 882 is about) and the many solutions of Multiple Comparison Correction (MCC: Bonferoni, Homes-Sidak, Benjamini-Hochberg, etc.) with b. Contamination or Interaction: your two v

More ›