It may be "free money" as you frame it. But a cash stream that provides n dollars per year forever can be valued in today's dollars, assuming a discount rate of d, at n / (1-d). So it's reasonable to prefer cash now to revenue forever, at that exchange rate, depending on your corporate interests.
You have the right idea, but you got the formula wrong. That's evidenced in the source you link, but you can also reason it from first principles: a higher discount rate should make the cash stream less valuable, not more. The correct formula is n / d.
The discount rate is doing a lot of work here. There is a discount rate such that we're not talking about shortsightedness. Getting it right is difficult. But as an example, how much would you buy an investment that pays a hundred dollars, guaranteed, next year for? Trivially, the discount rate includes at least the expected amount of inflation; it's not worth a dollar.
For assets line like IP you have to factor in how risky the returns are, how much investment you'd have to make to see them (e.g. making a movie), and overall strategy (do we want to be in that line of business).
All this to say - if you have IP that pays 10 million a year, you can value future returns on that IP in today's dollars. If someone offers you more than that to buy it, you should take the deal; you come out ahead.
Tons of comments already in the direction of 'eat less' and 'exercise more'. Of course. One thing I've found extremely helpful is to cook more! WFH is amazing for this. If you can find 15 minutes of meeting time throughout the day where you aren't an active speaker, you can chop vegetables or do dishes for lunch and dinner.
Back to the primary topic: cooking your own meals makes 'eating less' a matter of dialing in the portion sizes over time. I eat ~110g of pasta, or 0.8 cups of rice, or similar, and about 0.3-0.5 lb of protein, and about 1-1.5 vegetables per meal. Overall, I eat two meals like this, black coffee, a breakfast bar, and a snack, which puts me at 2000-2500 cal.
Now, on to a weird tip: meal planning! What a fun mix of intuition and ideas from operations research.
When you say tech industry pays more, I'm assuming the effective long-run hourly earnings of poker players is way less glamorous than it looks from an outsider seeing highlights? Doing a basic google search led me to https://pokerdb.thehendonmob.com/ranking/6737/, which seems to suggest that 81 players earned over 1mm USD in 2021. But presumably those players don't earn that YoY, and even if they did, plenty more software engineers earn > 1mm USD per annum, so this seems to support your conclusion. OTOH, I have no insight to the home games or whatever sources are not included in these rankings, so I'd love to hear a little more.
FanaHOVA mentioned the main thing you missed. One other thing I'd mention is that a lot of pros have staking arrangements (the article briefly mentions some versions of them). For example, I know one pro (who's won one of the more prestigious WSOP bracelets) who gets 30% of his tournament winnings in exchange for having all his entry fees paid by his backer. I suspect he's traded off more upside in favor of stability than most, but there's a lot of deals out there where the players aren't taking home anywhere close to 100% of their public winnings.
Those aren’t net of buy ins. Negreanu for example won $3.1M, but spent $2.56M in buy ins, so net profit was around $600k (doesn’t account for all his content ofc, but that’s not every player)
The other commenters said it for tourneys. I can’t stand tournaments.
For cash games, I was only playing on the casino. For the 1/3 (1$/3$ blinds) the earnings rate seems absolutely abysmal and you have to account for rake. There are grinders who play these games, along with 2/5. My assumption is any grinder playing here is down under or not making enough to compete with a professional career in the tech industry.
Next is 2/5. Here I could’ve sustained a decent salary. A pro at these stakes showed me his past year earnings which was 180k.
It’s hard to tell with the 10/25, 25/50 pros. There were definitely dudes sitting on 500k+, maybe way more, chips that they kept in the casino and I assume similar amounts of cash at home. There were guys leaving with big nights but still losing 90k in a night. I’d hear about some being down under 200K at any point. There were probably some money launders. There was the occasional private investment fund guy that came in, was short with slick back hair, very loud mouth, would brag and show off one of his 17M personal banking account. Then there were the very quiet, calculated guys. It seemed that they were operating on some formula I never cared to figure out. They’d avoid big hands for the most part. Some form of grind I didn’t have the discipline for. I also didn’t have the discipline to handle large swings at those stakes, nor the bankroll, or desire to play lower stakes to build a bankroll.
It's interesting to compare this to the physical-world analog. What if somebody shows up at the US-Canada border, start shooting rockets at Canada, and the US refuses to acknowledge this as a crime or extradite? What if the rockets are aimed at a critical piece of infrastructure near the border and can cause billions in damage? One could argue that if the US condones these attacks, they have effectively already declared war on Canada.
Well that's the main issue, currently the states do not consider such cyberattacks the equivalent of sending troops or rockets across the border that justify a "kinetic" response but rather the equivalent of earlier espionage activities which usually justifies only a diplomatic response. Of course, that might change in the future.
See Matt Levine's 10 laws of insider trading. In particular, "5. Don’t do it by planting bombs at a company and shorting its stock." Somebody nicely put a non-paywall link here: https://github.com/0xNF/lawsofinsidertrading.com
Buying puts in this fashion will generally be a remarkable, traceable event. There's a reason ransoms typically go through cryptocurrency.
I started looking at the map in the article and realized I had just booked a flight where there was an existing rail line ! So I checked out the Amtrak site, and what's 1h20m by plane is 14h40m by train -- and 8h by car. Maybe by getting a sleeper cabin I could have had an enjoyable trip by train, but as the trip scales things get dramatically worse.
Of course, maybe this is exactly what the future of transportation should look like: more localized travel on modes that can be powered by renewable sources or nuclear.
Compare the journey from my small village in New Mexico to Chicago, about 1200 miles. We just happen to have an Amtrak station 5 miles away. The drive time is about 18-20 hours without stops, which is long enough that an overnight stop is going to be likely. The flight time is only about 3 hours, but that requires first driving 40-70 minutes to the airport, spending time waiting in the airport, and then arriving at O'Hare, and then the 50 min metro journey back into the city.
The Southwest Chief, however, arrives here around lunch time, and arrives in Chicago about 24 hours later.
If you were optimizing for minimum travel time, you'd probably fly. If you were optimizing for cost, you'd probably drive. But if you want a nice journey, the train is fantastic and faster than driving if you're going to stop.
So, there are variations on the theme, and sometimes the train wins, sometimes the train loses.
> If you were optimizing for minimum travel time, you'd probably fly. If you were optimizing for cost, you'd probably drive. But if you want a nice journey, the train is fantastic and faster than driving if you're going to stop.
Train fans vastly overestimate the number of people who will optimize for "sitting in a train and staring out the window for days".
> optimize for "sitting in a train and staring out the window for days".
I'm acknowledging that it's not for everyone, but that's not really a fair depiction. I get a lot of work done on a train -- the atmosphere is similar to a coffee shop in some ways. Other people enjoy going to the common car and chatting and playing cards with strangers. Or reading. Or just watching Netflix on their devices, like they'd probably be doing at home anyways. We were also talking about an overnight trip, not "days".
I work in Ft. Lauderdale, Florida. The home office is in Seattle, Washington. I've made it clear at work that I'm not about to fly to the Seattle office [1] (not that I've ever needed to, but others in my office have). On a lark, I decided to see what it would cost to travel by train. On the plus side, there is train service between Ft. Lauderdale and Seattle. On the down side, it cost about $2,000 one-way. No way my company would spring for a ticket than costs more than first class air line ticket, but I wouldn't have mined the week travel one-way, as long as there was decent Internet connectivity (so I could continue to work).
[1] I don't have a fear of flying. I don't fly because of the security theater and the "presumed guilty" attitude. Also because of the declining comfort and service because people are prioritize the bottom line. It sucks.
Especially given that, if it's business travel, I expect a lot of companies aren't that big on you adding a couple days of travel time because you feel like taking the train.
In the case of the parent's scenario, it seems pretty reasonable in that you're really only talking about maybe an additional half day of travel. But that's probably about the upper limit.
This seems like a demonstration of routing issues, more than anything else. I have no idea how many changes that journey would involve, but there's no way that 55 hours to cover 1269 miles is representative of the time a train (even a clunky Amtrak train) would take to cover that distance. So this would seem to be an argument for increasing routes/services, rather than an argument that the train can't ever work.
OTOH, the flight is hard to compete with in that instance, so I suspect even with a better route, you'd likely still fly between those two places. Others might not, and the new service/route would benefit people making shorter journeys along the way.
I love the Lamy-to-Chicago train. I've taken it several times. You get to sleep through Kansas (where there's nothing to see anyway; no offense to Kansans!) and you wake up crossing the Mississippi River. There's 110v power and cell service for most of the trip.
(I always get a sleeper; without that it wouldn't be worthwhile.)
I travel form Chicago to Detroit pretty regularly. The train takes 4.5 hours, driving takes 4, and flying takes 1.5. Flying ends up being the slowest though because you have to spend 1.5 getting too and from the airport plus waiting at the airport. The train costs $25 which is cheaper than driving and I get to get work done. It's by far the best option imo.
As an employee at a megacorp, sure, you can see a little bit about how things are going from the inside. That said, you probably spend less time looking at reports than professional analysts. Secondly, sure, maybe you can see how great your company is, but do you have a frame of reference to a million other companies and their projects and cultures? There's a strong bias towards thinking you know more than the market that I would be wary of.
That depends a lot on your past background. A number of the employees at these megacorps have either worked in finance before, or at one of the big consulting firms, or they've done the Valley dance between the hot growth companies of the moment. So they do have that reference point between many different companies. There's sort of a revolving door among many of the elite institutions that run the world - Ivy League universities, large dominant corporations, big-name consulting, finance, and government.
If this doesn't apply to you, and a megacorp was your first job out of college that you stuck with for your whole career, you may want to be a bit more cautious with your capital.
As somebody with some ML background but no expertise in crypto, is the following ELI~20 summary correct?
We take an ML model trained on unencrypted data, use a 'homomorphic evaluation' technique (let's just leave that as magic here) to convert the model operation-by-operation to a model that runs on encrypted data, do a little more crypto magic, and we've solved the business problem described at the beginning of the article.
(In particular, if you train a model on encrypted data you get a really bad model, right?)
As someone who is not a cryptography expert, is there any hope of using similar logic to train on encrypted data? Naively it seems like you could perform the same operations on the back propagation steps (or any other update algorithm you're using for non NN models) to arrive at the encrypted version of the parameter updates, which you could then decrypt to get the updated model. Am I missing something here?
You're quite correct that evaluating the back propagation could be done exactly the same as the forward pass. However, if you want to train for more than a handful of steps you'll have to use an operation called bootstrapping to periodically "refresh" the ciphertexts that encode the model. Bootstrapping is essentially evaluating the decryption circuit of HE using HE itself (with an homomorphically encrypted version of the secret key). The problem is that bootstrapping is much more expensive than the other operations in HE.
People have done very effective training on encrypted data using simpler models, like linear or logistic regression. See for example this work [1] from my colleagues at Microsoft Research.
Training is a lot tougher. Just doing one gradient update step isn't all that bad (although you may have to play with the loss function a bit, e.g. logit cross entropy is probably tough to evaluate). However, then you need to go and actually do all the steps and gradient updates, so you probably need some form of bootstrapping to be able to evaluate computations of that depth. Also, the use case is slightly less compelling. For training, you can probably get all the parties who have data to coordinate and evaluate an MPC more cheaply than you could with HE alone. I think it'll require a very compelling use case for somebody to go and think through what the best way to do it is and it'll probably depend on the specifics of the application (who has what data, and what are we willing to leak as we go along - e.g. it's a lot easier if you don't care about keeping the weights secret).
There are definitely compelling use-cases and there are people working on it (though not me). Developing tools/systems to handle sensitive data in a secure way is extremely expensive and time consuming. If you can create data collection and model training pipelines that can operate effectively with just encrypted data then you greatly reduce the number of vulnerabilities (e.g. fewer employees need to actually see the sensitive data and fewer points of attack on the system itself).
There are certainly a number of factors to consider besides data security when evaluating the practicality of such an approach but I just wanted to confirm that it was technically possible before getting in to any of that. Thanks for your response and the post, I knew almost nothing about HE before today.
I'm piecing it together too. This is my understanding: the "model" has a set of rules (add, divide, bit shift) and that they do funky encrypted actions (add by making every third bit flip or something). If you run all of these actions in place of what you'd normally do (ie, just add like normal), then return the product, it can be decrypted as a result of the functions ran on it.
Any good references on the art part, or is it just an intuition you develop over time? In my experience, all the ML education will teach you a ton of theory and basics but none of the practical details you're referring to.
It takes time and a lot of hands-on experience. Many ML teams tend to work on one or just a few tightly coupled project for years. By contrast, we’ve worked on a lot of unique projects with real-world constraints so it gives us a different perspective. An important part has been developing a rigorous process; sort of a framework for applying the “art”. As you mention, this often isn’t covered in ML or data science education, which tends to focus on (important) fundamentals.
Sure, Wiles obviously understands the main thrust of his proof. But one could argue that Wiles' result depends on lots of other results, which in turn depend on other results, and so on through decades and decades of work, ultimately going back to the foundations of mathematics. Neither Wiles nor anybody else can claim to rigorously understand all of it. You can imagine this as a tree, with Wiles' work as root, and his dependencies as ancestors, and so on. An error at a lower level of the tree could, in theory, invalidate the root node.
I do agree with Buzzard that it's hard to be sure. I've definitely read papers where a critical argument isn't well written or what is written seems wrong. However, if there are low-level errors, I suspect that with some work things could be patched up.
Right, speaking as a lapsed mathematician, I definitely see errors or gaps in published work. Wiles’ original FLT proof had one. And yes these can generally be patched up. I’m not quite as alarmed as the author is, because generally a major false result would have all sorts of alarming ripple effects and implications which would be pretty easy to spot. FLT is an extreme example where literally anyone with a calculator could in theory disprove it. The fact that no one has suggests to me that it’s likely true.
You don t need to understand everything. People spend there all lives creating new demonstrations. Each one is a new prospective on a subject. And if the first one was made of milions of nodes an other one can be made of two nodes only. Take the index theorem there are proofs that have nothing to do with each othere some are very long some aren t.
And finally from the beginning of math people misunderstand their on theorems, doesn t mean that there students won t do better.
https://www.investopedia.com/terms/p/present-value-annuity.a...