ahh

The developer interviewed claimed that there was no domain-specific knowledge and implied (though didn't explicitly state) there wasn't any training against non-OpenAI bots or players. (I'd love to know the reward function they used for whatever Q-learning variation they ran with.)

If this is accurate, one of the things I'm most impressed with is that the bot figured out creep-blocking. (I can't find a good GIF, but this is walking in a wiggly path in front of the first wave of neutrals on your side, delaying their progress and pushing the lane towards you, which is good for ~reasons.)

Creep blocking isn't all that hard in dexterity--I am a terrible dota player and I can more or less do it. And it's one of the most common pieces of dota knowledge; every pro player does it and since it's relatively easy compared to a lot of pro micro, everyone else rapidly learns they should.

But nevertheless--the bot had enough games that it could randomly jump in front of the wave enough times that it noticed a win rate improvement for that slight wave push, and begin to do it intentionally? (And then get good at it?) Damn.

One thing I don't really know about Q-learning and the typical nets used for it: I am guessing it is likely that internally to the bot's evaluation functions, there is some learned feature whose activation correlates well to the location of the wave equilibrium (since that's a feature that correlates well with winning!) At that point, is it likely that the bot can learn in smaller increments--that is, it knows that pushing equilibrium towards itself is good, and thus randomly creep blocking a little becomes reinforced (rather than having to notice the creep block's effect on game wins?)

I think you might not know what a meta-analysis is: it's actually a summary of replications and duplicate studies on some topic. Generally (I have not read this particular one!) they are specifically designed to detect potential publication bias and other effects that would imply non-reproducibility.

It's a hard problem. I think there's a simple observation that helps explain why it's going to be so hard (the article hints at this without saying it): a healthy adult man is, more or less, fertile. At all times. If a adult man is infertile--if him having sex can't result in pregnancy--there's something at least slightly wrong with him.

By comparison, an adult woman is often infertile--three weeks (ish) out of every month, or whenever she's pregnant, she can't get pregnant.

So it stands to reason that it's harder to produce male hormonal contraceptives: we can (at least sort of) fake healthy-but-infertile states of women, but there are no such states for men; we have to construct one from whole cloth, so to speak, and that's much harder.

Are these remarks of Steve (not a person I know, but I wouldn't expect to...) public somewhere?

How do you feel about in-browser password managers--Chrome in particular?

I think you can easily separate understanding risk from being disappointed that the risky outcome occurred. It's the difference between knowing (or not knowing) that as a poker player, getting it in preflop with AA vs KK is a great idea, and being sad that your opponent spiked a K on the flop for a $5,000 pot.

Now, good poker players do talk a lot about not being "results-oriented" (even if your AA is rivered by KK, you played the hand well if you got the money in before that), and many players will tell you that the way to do this is to cultivate a sense of detachment about money coming in and out. I certainly try to avoid being sad when I lose a big hand; such is life. But being sad isn't the end of the world, so long as you got the money in good and continue to do so.

How did Jefferson do that while not actually being there?

Seriously. Jefferson wasn't a delegate at the constitutional convention. He was in France.

Einstein discovered a counterexample to Betteridge's law?

How, given arguments in GP, does this actually do anything of value? Note that I do not consider signaling how much we care as valuable.

You are just echoing the politician's syllogism: we must do something, this is something, we must do this.

Your point that we could not get anything better is not an excuse for doing something bad.

(I work at Google but don't speak for them; I have access to internal retention data but I'm not looking at it or using it here.)

I think 90% retention is (probably?) an underestimate. People tend to look at measurements of tenure, see a short number, and assume Google et al burn people out/have revolving doors. Instead, that's really a function of the still explosive growth we're going through (for better or worse; I say worse but who listens to me?)

(Which still serves Patrick's larger point.)

How much do you work out?

I like IF a lot too, but when I'm in a heavily active phase of my life--say lifting 3x week, climbing 1x, jiujitsu 3x, and full on dance rehearsals 3x a week--I find I start to crumble unless I eat more to keep fueled.

You can actually do adjustable weight sampling in logarithmic time using a modification of the binary search approach, which is likely superior in a lot of cases.

The trick is to keep a dense binary tree (arbitrary order--you can use the standard array mapping), where each leaf has one of your w_i and each internal node is the sum of its two children (and therefore equals the sum of all leaves in its subtree.) To update one weight, you just walk up the tree. To sample, you walk down the tree making a weighted decision at each point.

This allows log n time update and sampling. It can be quite fast in practice, though the alias method is faster if you don't need adjustable weights.

There are all sorts of other issues with RNG generation of doubles that affect these cases, though - I need to write a longer post about what I've done here.

I have had the same bug for about a year.

Before we ask about methods...has any American city ever fixed a traffic problem?

No, I'm serious: has any city in the US ever had traffic at some (bad) level, implemented some set of new policies, then had traffic drop substantially while maintaining the same population density (i.e. I won't count the abandonment of Detroit.) I certainly don't know of any, which makes me doubtful about anyone's proposals to fix traffic.

Anyone else a bit disappointed?

Normally these letters have some new brilliant insight or dive into a business I know nothing about. This one feels shorter and more peremptory. I see the financials for the major sectors and the same boilerplate explanation of insurance and railroads that's in every letter.

What's up? It's not like nothing happened with Berkshire Hathaway this year.

The land in question used to be anonymous office buildings owned by...HP or someone like that, if I remember.

Apple has taken away precisely what from the public by replacing it with this?

Note that he's not giving a figure for a chemical engine, but a fusion engine (sometimes termed a "torchship" due to the absurd power involved.) That would be strictly more efficient than VASIMR or similar. The figure would be even more ludicrous with a LOX/LH2 engine.

(I did not check his math, however.)

Wait a minute, wait a minute: if I just decide I'm not going to work anymore, they can't fire me if I say it's because I'm protesting for higher wages? And they have to keep paying me?

  But in less than 30 months following the Apollo 1 accident, 
  NASA flew five Apollo missions. During the final one of
  those five, Neil Armstrong and Buzz Aldrin stepped onto the
  Moon. Much later, in the last decade of his life, Armstrong
  would reflect on what it took to reach the Moon. “The rate
  of progress is proportional to the risk encountered,” he
  said. “The public at large may well be more risk averse 
  than the individuals in our business, but to limit the
  progress in the name of eliminating risk is no virtue.”

Sadly, Armstrong and his ilk no longer makes the calls at NASA (and haven't for quite some time.) And so we see no progress.

Aha, thanks; I didn't realize the first generational section was for an older version.

However, the riptide barrier description still isn't 100% clear how it avoids terminating a sweep (I believe "drain" in your termination) in the middle of a barrier execution; in particular I'm not concerned about the data race between the field write and the mark bit, but the GC reading mark bits, believing the drain has terminated, and freeing memory while a mutator thread has written a field but not written the mark bit.

Are you just relying on the stop-the-world behavior relative to roots that you mention briefly?

  Some accounting only needs to happen when visiting the
  object for the first time. The complete barrier is simply:

     object->field = newValue;
     if (object->cellState == Old)
       remember(object);

This surely elides some synchronization? One would think that either this has to communicate with the collector (to prevent a racing concurrent collection from missing the referent), or happen in the opposite order.

Or perhaps it is relying on the calling mutator thread necessarily holding a reference to the young object in question? Even then there is some non-trivial ordering work to be done here (in particular on weakly ordered architectures.)

I'm sure this is well handled but as this is one of the main difficulties in implementing good write barriers I am curious what they did.

I agree that unions don't work unless the government dictates that employers bow to their will.

Whether that's good for companies or the public is a whole other question.

I don't know for certain with this law in particular, but typically the answer is that the description is a lie: they are not being granted permission from the council to organize. The council is banning Uber from hiring anyone not represented by the union (should they vote for it, but the fix is in, as noted by the carefully gerrymandered voting rules.)

Anyone can organize at any time, and the employer is free to tell you to pound sand and hire someone else. Seattle's trying to change that. Expect rates to go up and availability to go down.

> I've had an accountant do my taxes since I was 19, and > feel like I've probably saved a pretty substantial amount > of money...simply by ensuring I'm doing them correctly > and, more importantly, on time.

Can you give one specific example of something you wouldn't have been able to do correctly and on time yourself, that has at least $1000 in value to you? I've been doing my own taxes (with a six figure tax bill and often 15+ 1099s) for ten years; I'm pretty confident I've not made any meaningful errors other than typos or transposing labels on the odd stock entry, a mistake that would cost me $0 even if audited.

I--at least for a W2 worker who is a US citizen living in the US--have a uniquely bad tax situation, and I'm pretty confident that I'm not missing any major tax breaks, short of doing absurd accountancy.

I don't hate this (and it's why I don't pay quarterlies--also, I tend to generate enough deductions to be okay.)

But it's absolutely not correct that paying people with bonuses lowers their effective tax rate.

This only affects the withholding of bonuses (the amount the IRS has you pre-pay as an estimate of taxes.) It does not affect the actual income tax on bonuses.

(It's mostly a disadvantage. I get about 60-70% of my income in bonus and stock which is treated the same; it just means that, if I don't plan ahead for it, I end up with a surprise tax bill on April 15th. I do plan ahead, but the whole process is obnoxious.)

Is there an easy way to remove a clock from Android? How about OSX?

Think about what you're saying: "It's a red herring! We know that schools spend money they get on dumb things!" Yes, that's my point. :) If we could magically fix that, it might--in principle--become a good idea to give them more money. Giving them money does not actually magically fix that.

It's worth noting that Maciej forgot that the tech barons he hates actually tried this: https://www.washingtonpost.com/opinions/how-newark-schools-p...

First off, that's a blatant ad hominem.

Moreover, if your best argument against the guy you're claiming is defrauding everyone is "I can't be bothered to read his work"...