How do you handle progressive overload? Do you have a bunch of different kettlebells of various weights? Seems like that would defeat the space savings. Or do you cycle out the lower weights or put them in storage once you've progressed past them? Curious to see if I could make this work for me, as I also have time/space constraints.
Yes, that seems like an important problem, but one separate to what they're describing in OP's article. (Again, assuming I'm understanding this right.) Their constrained RL approach is still relying on our ability to enumerate and assign costs to the undesirable behaviors, right? From reading the article, I get the impression that they are focused on addressing that scenario, and leaving the problem of how to enumerate all undesirable behaviors to separate research.
I'm not super familiar with AI/ML/RL at all, so I'm sure this is a naive question, but isn't it obvious that the answer is to just build in costs to the utility function for behaviors you want to avoid (what they seem to refer to as constrained RL in the article)? That seems both the simplest way to handle it, and most elegant in terms of mapping to the real world domain. Like are there alternate solutions that are even remotely competitive with this? I'm sure I must be oversimplifying and I assume that there's some nuance I'm missing. E.g. is this more about how you design those constraints to minimize the overall loss in learning efficiency, or something like that?
But Dropbox's mission statement from their S-1, which is what this article is referring to, is an 8/10: "Our mission is to unleash the world’s creative energy by designing a more enlightened way of working."
I suppose, but how many late-stage private companies (i.e. late enough that you're getting stocks instead of some kind of option or other you-aren't-technically-gaining-anything vehicle) are having wild downswings in stock price? I'm not aware of any, and while I could imagine there are some I would assume it's a relatively small number, but I could just not be in the loop.
I'm not an expert on small business loans, but I would guess that incentives take care of most of this problem. The higher the loan you have, the more cash your business must have been generating in the first place, therefore the higher incentive you have to figure out how to keep it running. There probably aren't too many scenarios where you are both eligible for a material loan from Stripe and yet not highly enough incentivized to figure out how to make the business work to the best of your abilities.
Energy in/out, at least in the sense most people and your grandparent refer to, can be wrong without violating conservation of energy. E.g. if you poop out food from which not all calories have been used.
I'm not sure I even understand your argument. It's possible we're saying different things? What I'm saying is that if X people try a sport the records they set won't be nearly as good as if 100X people are trying it. Winners are by definition outliers, and the larger your population, the more (and more extreme) outliers you will see. So if there are five times fewer women than men () competing in a sport, that will impact comparisons between women and men, even at the top levels, at least if you want to compare innate ability. Almost certainly, the women's records and top female performances would look better if there were five times as many women trying the sport as there currently are. Do you disagree with that?
I pulled up the most recent Ironman race; there were 5 men for every woman in the competition. So I'm using that as a rule of the thumb. But the same logic applies to any population imbalance at the top of the funnel.
I don't think it's quite that close -- there may be sample size/participation factors involved here. For example, the top US women seems to be competitive with the top ~10 US men, which in more testosterone-dominant sports would never happen.
To be fair, the differences here seem to be smaller than in other sports, suggesting that while testosterone is a factor it may be less than it is in other sports. If 90% of the competitors in these races are male, that would further impact this discrepancy.
You have to explicitly write your code to prevent this from happening. You typically have an engine that is executing all of your logic (rendering, game logic, AI, etc.) in cycles. By default, it will execute these cycles as quickly as possible. So for anything with real-time components, faster CPU = enemies move faster.
I forget which game first introduced a fix for this, but the solution is to have the engine run on time-based ticks -- you wait to execute the next cycle until the requisite time has passed.
I don't read his comment as intended to be demeaning, but more a statement of fact. And it reads true to me. I have no skin in the game, but I follow a lot of the bootcamp-related discussions online, and what he said sounds exactly right to me. There are a ton of people who hear the idea of an ISA and say "so we're bringing back slavery?? thanks SV geniuses!!", and they seem to disproportionately come from the late stage capitalism crowd.
I feel like you could make 6 minutes of air last an hour or more if you breathed really slowly (like one breath every 30-60+ seconds). Maybe I'm misunderstanding the device he had, or maybe it already has that assumption built in, or maybe he just wasn't quite that efficient but that is part of why he was blacked out but not dead when his friends pulled him up.
I do the same but I use DND 24/7 but with calls from contacts as the only exception - those ring as normal. I get phone calls from people I know so rarely that it's not a significant hindrance.
Grinding leetcode or similar things would probably work if you have the right work ethic. You could also consider a coding bootcamp like Lambda School. Way cheaper/less time than a 4-year degree, and they have much better job placement programs. (And will probably teach you the skills needed to pass an interview better than most 4-year undergrad programs would.)
Without knowing your skills, there's a good chance you've just been working for the wrong companies. There are lots of companies in Canada that pay absolute garbage to software engineers. And if you don't have the skills that higher-paying companies look for (mainly coding+algorithms+data structures in interviews), you can get that without going to college. Also look into working remotely for US companies.
Furthermore, they are limiting the reaction time to 200ms, to match good humans (I suspect some pros are actually faster than that) and remove any advantage there. So it doesn't have a meaningful advantage over pros in any mechanical/reaction time sense; it's truly just trying to play the game more intelligently from what I can tell.
As someone who works/hires in the field, the salary reports are all wrong. I have not seen a single one yet that was close to accurate. In software a 90th percentile engineer can easily make 2-3x what a 50th percentile engineer makes. (And it can get even crazier as you move higher up the tail.) This is not true in virtually any other industry, which seems to cause problems for the salary reports.
Dumb question: Why would YC get diluted so much more than DH? I've started to read up on venture financing and this doesn't match my mental model for any of the mechanisms I've learned about. (Like if anything I'd expect YC to have less dilution due to follow-on/pro rata investments, if they did any.)
Sexual harassment and power structures are intertwined. That's part of what this whole story is about. If they weren't so powerful, it would be much harder for them to get away with this, and they would probably be less likely to do it in the first place.
Their whiteness is a part of their power. It's not the only, and maybe not even the primary reason for their power, but it's certainly relevant.