There are zero studies that disprove zinc + early has a strong effect. You are wrong. Also, the lancet study was retracted for flawed methodology, other studies show slight benefit of hydroxychloroquine. There are zero studies that show it has no effect.
You can’t show that it has zero effect without indefinite numbers of patients enrolled in the study. It hasn’t shown any dramatic improvements in the studies that are there. The rest is a tradeoff between enrolling even more patients into studies to prove small improvements (or not) or to move on and spend resources on finding things that do have a major impact.
There's no studies yet. But there aren't any preliminary results that look promising, and the only people who think it is promising also thought hydroxychloroquine was.
So it's pretty unlikely it's going to do anything.
The data we have on hydroxychloroquine is now convincingly negative-there's just a ton of misinformation and it's become a political issue. But when some institutions fail you get widespread mistrust of both institutions and experts, which maybe is part of the compounding problem.
The hydroxychloroquine study I'm referring to has less to do with the data and more to do with how the FDA behaves regarding the drug, where they gave the drug an emergency use authorization based off of what seemed like pretty blatant interference from the executive branch. If the drug had ended up working, maybe their reputation would have taken less of a hit, but that wouldn't have changed how inappropriate that authorization was, and how it demonstrated that you could get a drug approved if you had enough political power from the president behind you. That's not a message that's making me optimistic when we look at vaccine development right now - that core trust that the agency operates on science/public health instead of politics has been shaken.
There was some early evidence it could be effective. I’m actually supportive of the FDA’s move to do both the emergency approval and the retraction. Without the ability to take calculated risks in a situation where either approving or unapproving can cause harm, the agency could be a hindrance more than a help. And that means some emergency measures will prove wrong. If none prove wrong, the agency is being too cautious and allowing people to die because of it.
EDIT: I hate the politicization of HCQ by Trump, of course.
HCQ was authorized to be used as a treatment, not merely authorized to be used in trials or for compassionate use (as remdesivir was). That’s the scandal in the face of lack of evidence at that time.
It’s another in a very long line of mistakes over the past three years pointing towards one thing: process matters. It’s boring, it’s not dynamic, but it’s also the thing that prevents organizations from making some extremely bone headed mistakes.
Agree with the point, but the disregard for process has been going on for longer than 3 years. If we want our principles to be taken seriously we can't just call out the other side. Remember "I have a pen and a phone"? How well did that work out?
...and it wasn't our side that called in a drone strike on American citizens overseas, without even a nod to due process.
And who pushed for this drug? I don’t believe the CDC even cared to consider it that much. But the current administration and Trump were pushing way too hard on this. So they felt compelled to investigate. Look at all the damage Trump has corn to all of the organizations in the USA. He’s trying to take away any independent power and concentrate it into his office.
This just seems like something that will catastrophically fail. If you can build a good enough generator you can just build the ML model internally. And if you can't the statistics of what you provide are going to be off enough that any strong model is going to be wrong in strange ways.
This is an incredibly important point: in order for your synthetic data to be useful your simulator must have already solved the problem at hand. In theory there is no need to even fool around with generating the synthetic data and going through the charade of training a model on it; simply exact the outcome model from your simulator directly as that's implicitly what you are doing. For example, if you have a generative model that provides densities, you can simply compute P(Y | X) = P(X, Y) / P(X).
But this is not how generators work. They generally produce samples in the from
G: Q -> (X,Y)
where Q is some prior from which you are sampling. If they are not invertible then you straight up cannot get P(X,Y) out of the generator. Even if it is invertible getting P(X) requires integrating out the Y which might be infeasible (since the model is not integrable and is sufficiently fast changing that you need very, very many samples).
Very true. If you've solved the labeling/extraction problem using a means other than ML, you can use that means to generate synthetic data. The situation at my company is exactly this.
Say you use regular expressions to extract sensitive data from standardized, but numerously varied, form documents. The pieces of information extracted are very common classes of data: first name, last name, dates, physical locations.
During the extraction process you can save the complement of the extraction (the "leftovers") and insert generated data at the extraction points. Also, because you've extracted the actual sensitive data, you can exclude that from the set of values used for generation, if it's practical.
Sometimes people get caught up in the math and theory that they fail to see the practical solutions.
I agree that this is very tricky. I think the most interesting synthetic healthcare data generation I saw was using causal inference (where SMEs can bake in a bunch of expert knowledge during skeleton construction) and then generated data by getting the weights on the edges from a smaller dataset. At the same time, it is very hard to ensure that you synthetic dataset actually reflects real world. On one hand SME knowledge might give extra oomph to synthetic data generation (as this knowledge is equivalent to some highly abstracted training) but also if the "expert knowledge" is wrong then it's a recipe for disaster.
You can create much higher margins for products and services at the Apple price point. For Autos, people are very price sensitive. Even if Tesla slowly takes over the industry, not sure they can justify this valuation.
Why do you think college is 10x more expensive now? Seems like maybe some of those expenses are not materially impacting education quality, unless you think that the generation that got us to the moon was working with absolutely substandard educations.
I have a friend who is an advisor at a Big 10 school. While going through the old papers in her office she found the budget of the school from the late 90s. The school's expenses have doubled in that time. Part of it is certainly the amenities arm race. However, the amount of money they get from the state has only gone up by 50%.
> While going through the old papers in her office she found the budget of the school from the late 90s. The school's expenses have doubled in that time. Part of it is certainly the amenities arm race. However, the amount of money they get from the state has only gone up by 50%.
The implication here is usually that the amount of money they get from the state is insufficient to keep up with costs, therefore universities are forced to raise tuition. I don't think this is true. I suspect universities would raise tuition anyway, and if they got more money from the state, they would just increase their costs even more.
I'd almost be more interested in the change in variance across time. If education costs are the main factor, and the top 10-20% (? estimate) have parent's who pay for college, you'd think this might be yet another factor exacerbating inequality.
I'd also be much more interested in a more detailed net worth distribution over time. Reminds me of the quote Nassim Taleb is fond of, "Never cross a river that is on average 4 feet deep".
VC has historically done slightly better than the stock market, but returns over any time period are inversely correlated with how much money is flowing to VC (ie when everyone wants a piece of the action more bad deals are done, terms are better for companies etc) - so you'd expect the current period to produce okay returns (but probably less than the S&P). People imagine VCs return 100x or something, but if that really happened it would reveal an absurd amount of under-investment or really bad negotiating by founders.
And given that these reasonable but not stellar returns put them in the top quartile (and VC has high variance even relative to hedge funds), you can guess how well things have gone for other folks.
It seems like reasons you might invest in a VC fund could include portfolio diversification (if returns are counter cyclical or have low correlation with other asset classes), risk profile or expected performance. It would seem a little surprising if tech focused VC funds had low correlation overall with the NASDAQ but maybe it's the case. If it's mostly about risk profile and VC is riskier then at least according to traditional theory it ought to outperform the NASDAQ over a good period for tech like the last decade but would underperform in a down cycle.
It's somewhat hard to calculate the correlation because VC is so illiquid, but for sure it provides diversity and has historically been an excellent investment. The variance across VCs is also huge relative to other types of asset managers, so you have the potentially for really obscene returns. If you have enough money that you don't need to be liquid it is a good investment.
What you're describing doesn't sound like a particularly good investment. You replace the problem of picking stocks with the problem of picking VCs and have less liquidity. You have the potential for really obscene returns picking stocks too but there's no real reason to think picking VCs is any easier.
That is correct. There are not ETFs for these kinds of investment classes, so similar to hedge funds you would need to believe that you are capable of picking the right managers. You also need to be an accredited investor to gain access, and funds like A16Z mostly take institutional money (endowments/sovereign wealth funds etc). Like hedge funds, management fees are very high. But folks want the allure of pre-ipo tech companies...
This people do not understand the math of startup comp. If you get equity instead of cash, you could have taken that cash and just put it into as high a risk investment as you want. As such, startups are really offering access to an illiquid investment (but how sure are you it's better than other options), and forced risk-taking with some short-term tax benenfits (most people wouldn't put 10-20% of their income into one stock and you pay tax up front). You could go to google, put 30% of your income into crypto, and have some thing a lot like a startup risk/reward profile w/ a lot more liquidity. It's just that that feels riskier to people. Startup is only a great deal financial for the founders.
