Similarly, I recall a talk by Daphne Koller where she talks about a heart attack predictor using X-rays as input. They found that stents were the biggest factor for positive prediction. Which is useless because the doctor would know if the patient had a stent.
In addition to their excel errors, their analysis required excluding the Australia, Canada, and New Zealand. Once these countries were included in their analysis, their argument falls apart. To me, it's clear that this was a case of very selective researcher degrees of freedom to support austerity. Why anyone would take Reinhart or Rogoff seriously after this farce is beyond my comprehension.
I think a lot of things in the past ~10 years were a zero interest rate phenomenon. VC driven start-up culture is definitely included in that. Growth for the sake of growth, rather than business fundamentals.
The SEC determined that since ICOs trivially satisfy the Howey Test, ICOs are securities and must follow security regulations. Of course, if anyone were to follow security regulations and have a legitimate offering they would just do an IPO.
I was with you until the rant about taxes and regulations. I was under the impression that child labor laws are a form of regulation and that loosening these laws would be a form of deregulation.
A big issue is that the customer base is highly correlated/undiversified, in the sense that they're all funded by a small set of VC's telling them to withdraw their money. SVB was uniquely exposed to the whims of a small set of individuals in this manner.
Even a healthy bank had a run they would not survive, because fundamentally banks borrow short (deposits) to lend long (loans, bonds, etc). The old Bagehot quote comes to mind: "Every banker knows that if he has to prove that he is worthy of credit, however good may be his arguments, in fact his credit is gone..." Anything that would cause depositors to worry and demand their money back quicker than the banks' investments mature would run into this problem.
Reminds me of my number theory professor. Open book take home exam, 20 problems and you choose 4 of them to answer. He graded us on how much we were able to do, and not just the answer itself. 6 of the problems were equivalent to the Riemann Hypothesis (unbeknownst to the class).
First year is general classes (e.g. in economics that would be Micro, Macro, Econometrics, etc), 2nd year is classes more specific to your area of study.
What’s the trade-off though? If an Apple Watch sends an emergency call every time a skier or snowboarder falls, that’s several calls per day. I would think that would be too much and we should hold Apple liable for that level of false alerts.
It’s an extreme example, but I want to make the argument that there is a balance to consider.
As a hiring manager in the field, my opinion is yes. Companies are realizing reliably obtaining and processing data is harder and more critical than model built. Especially for early stage companies.
That’s the sense I got too, especially after reading the fellatious piece on SBF from sequoia capital. Trying too hard to justify getting rich, because that would theoretically help the world, in the end, regardless of the means to get there.
I think the equivalent is the test set. A test set is used to evaluate the performance of the model, after all the tinkering. As such, you need to take great care to make sure that you don't have train-test leakage (e.g. taking a random sample for stock market prediction when you need to do a time-based split), or that you don't use the test set more than once (e.g. choosing the best model based on the test set).