Hacker Newsnew | past | comments | ask | show | jobs | submit | lalaland1125's commentslogin

Immigration isn't the issue. Population growth is actually down compared to historical levels. Immigration numbers (and births) are lower than they have ever been and housing costs are higher than they have ever been.

https://www.stlouisfed.org/on-the-economy/2021/january/us-po...


What middle class SFHs? There are no middle class SFH neighborhoods remaining in Los Angeles or the Bay Area. Take a look at Zillow. Your average young person isn't buying anything anyways.

Your information is at least two decades, maybe three, out of date.

But this bill will help lower rents, which is a very worthy goal in and of itself.


I think you are incorrectly missing that many larger units (both 3+ bedroom apartments and houses) are currently filled with singles or couples with roommates who would rather live alone in 1 or 2 bedrooms, but can't due to inadequate supply.

Building 1/2 bedrooms would help those people move out, freeing up larger units for families.

> I fear it's just going to enrich the property management class

The property management class benefits most from the current system with no construction and high rents. Building a bunch of 1/2 bedrooms, triggering lower rents, would cause them to lose money.


I wonder if this is true. There is significant risk in price changes for renting (or even HOA fees for condominiums), such that many middle class people might feel more secure living in their home with a near zero interest rate mortgage, if not a paid off mortgage.

On top of that, most jurisdictions (in the US) subsidize property tax rates for senior citizens, so there is a lot less price volatility for simply remaining in one’s home (or even moving to a different, but smaller detached single family home).

Unless a person specifically wants an urban lifestyle in a shared building, I don’t see much impetus to move out. Worst case, they get to stay in their home they have gotten used to and have space for visitors, best case they save a bunch of money and sleep easy knowing their costs are more controlled.


As little as one experience is worth,

When I lived in socal, almost every person under 35 i knew there was living in a 2-4bd with roommates.

1bd and studios were very scace, and almost often prohibitively expensive.

So I second that creating a real offer for 1bd and studios would definitely free up family housing.

On top of that 1 and 2-person households still need housing. Building some for them is a good thing in my book


I was talking about renters moving out of larger apartments or shared homes into 1/2 bedroom apartments.

Middle class homeownership is basically dead in California due to the absurd price of housing. Almost everyone young who didn't inherit wealth or earn 90th percentile income is renting


I don't think we are ready for the risk of mass manipulation through AI relationships.

Imagine if Elon decides tomorrow morning that he wants to encourage all of the users of his AI relationship app to support Trump?


A drop in a bucket compared to what already exists via algorithms and the media at large.


What do you mean imagine? He's already using his AI to spread his political beliefs.


> I don't think we are ready for the risk of mass manipulation through AI relationships.

We are not ready but we're ripe for them.

The advertisement industry has fingered our brains and raped our attention while eating up all the information they could get, now "AI" is harvesting our "open relationship with sharing information about ourself". The merger between the work the Ad Industry has done in preparation with the new data will collect will be catastrophically successful.

Meanwhile, we willingfully slept on digital literacy. The effects ripple already through many aspects of our societies. Causing havoc.

We're running toward an really "interesting" peak in Western Societies and I don't see how that might stop or even slow down.


Right that would be extremely dangerous to our democracy[1] Only the TV and mainstream media are allowed to manipulate opinions.

[1] https://www.youtube.com/watch?v=ZggCipbiHwE


Does Elon still support Trump? I don't follow the presidential soap opera closely but thought they had a big falling out over the beautiful bill and he was toying with the idea of a third party?


He literally just paid for Trump’s reelection, why do we think these people need AI?


Potentially this is like giving an axe murderer an uzi. They don’t need it, but they’ll probably be more effective.


Cheaper, general purpose, and more effective than other means.


It would be a political disaster to transition people off their employer plans.

The short term effect would be an increase in costs but no increase in pay for most people.


With a small tweak it would have the opposite effect. Just mandate that employers must add what they previously paid in health insurance to people's salary. So it would look like significant pay raises to a substantial portion of the populace. I bet that would be popular.


I don't know if you would even need to mandate it. Stop mandating it, and start taxing all benefits as income, and you remove the biggest incentive for employers for pay for them.


That would suck for employees. They just effectively got an effective decrease in pay.

A switch from employer-pay to government-pay should be a no-op for employees with employer health insurance. But in a naive scheme, it isn't. The burden for paying for health insurance moves from the employer to the employee (through increased taxes). The employer benefits because they stop paying for health insurance, the employee pays the costs.

Voters, who are mostly employees, would hate it.

OTOH, an on-paper pay raise for employees that doesn't cost the employer anything? That'll be much more politically palatable. "Both your taxes and your salary go up 10%" is a lot more palatable than "Your taxes go up 10%".


Health insurance contributions are pre-tax. Take home pay is post-tax.


It was a political disaster to keep people on the status quo of employer provided coverage, which the ACA did.

There's no safe path for any of this, with such effective propagandists ready to slander whatever move you make.


> Thats always the claim, but they have stolen SSN's to get those jobs in the first place (those without SSN's aren't paying income tax).

Where are the criminal prosecutions if that is actually happening? For all of Elon's efforts, even he was unable to find much (if any?) social security fraud.

> Not to mention states like California which explicitly spend federal money on illegal immigrants through Medicaid.

With the exception of pregnancy and emergency care (which are relatively minor), this is not the case.


People get prosecuted for stealing SSN’s all the time (identity theft), just generally not for paying in taxes under said SSN. The gov’t loves free money.


> It's hilarious that the dashboard is claiming that illegals pay taxes at the same rate as everyone else

The dashboard is explicitly not doing that. They cite, and use numbers from, research reports that explicitly estimate the taxes that illegal immigrants pay.

You can disagree with their analysis, but they are not making the assumption you are claiming they are making.


This blog post is unfortunately missing what I consider the bigger reason why Q learning is not scalable:

As horizon increases, the number of possible states (usually) increases exponentially. This means you require exponentially increasing data to have a hope of training a Q that can handle those states.

This is less of an issue for on policy learning, because only near policy states are important, and on policy learning explicitly only samples those states. So even though there are exponential possible states your training data is laser focused on the important ones.


I think the article's analysis of overapproximation bias is correct. The issue is that due to the Max operator in the Q learning noise is amplified over timesteps. Some methods to reduce this bias, such as https://arxiv.org/abs/1509.06461 were successful in improving the RL agents performance. Studies have found that this happens even more for the states that the network hasn't visited many times.

An exponential number of states only matters if there is no pattern to them. If there is some structure that the network can learn then it can perform well. This is a strength of deep learning, not a weakness. The trick is getting the right training objective, which the article claims q learning isn't.

I do wonder if MuZero and other model based RL systems are the solution to the author's concerns. MuZero can reanalyze prior trajectories to improve training efficiency. The Monte Carlo Tree Search (MCTS) is a principled way to perform horizon reduction by unrolling the model multiple steps. The max operator in MCTS could cause similar issues but the search progressing deeper counteracts this.


https://news.ycombinator.com/item?id=44280505 I think that thead might help?

Total layman here, but maybe some tasks are "uniform" despite being "deep" in such a way that poor samples still suffice? I would call those "ergodic" tasks. But surely there are other tasks where this is not the case?


Good clarification. I have edited my post accordingly.

There are situations where states increase at much slower rates than exponential.

Those situations are a good fit for Q learning.


Is this essentially the same difference as between vanilla regular-grid and importance-sampling Monte Carlo integration?


Feel the Majorana-1


> What does that part about VAT have to do with this? VAT is essentially a sales tax with a more involved collection process.

To add some further context that helped me understand VAT:

Sales taxes are great, with minimal dead weight loss and distortion, but have the downside of encouraging black markets since it's easy to avoid reporting final sale transactions.

VATs are designed to be mathematically the same as sales taxes, but robust to black markets. The sales tax is captured on the manufacturing end, which is much harder to avoid reporting for a variety of reasons.


Sales taxes have the disadvantage of being regressive: the less prosperous spend a relatively larger proportion of their income on taxable goods. There are states that offer deductions from income tax for sales tax paid on food and perhaps other items, but as I recall it took a very disciplined filer to claim it.


Some states don't require tax collection at point of sale, so things like basic food or essential clothing are not charged sales tax. There's no need to apply for refund when computing personal income tax.


> Sales taxes are great, with minimal dead weight loss and distortion

That's not context for VAT. Or a fact. As far as deadweight loss and distortion go, a uniform VAT is the same thing as an income tax.

If you want a tax with minimal deadweight loss, you're pretty much stuck with a capitation tax.


VAT only reduces consumption. Income tax also reduces investment.

For example, if a distiller buys a tank, income tax is immediately paid. But VAT only generates revenue many years later for the country when the beverage leaves the store. So it's really a consumption tax.


No, there's no difference. Whether all prices rise by 25%, or all incomes fall by 20%, the system will reach exactly the same equilibrium. When a VAT is enacted, you can model this as everyone with an income paying the corresponding income tax. (For full accuracy, you'd also need everyone with monetary holdings to pay a one-time wealth tax, but you can safely ignore this because the amount of wealth is so small relative to the amount of income.)

The effect of the VAT is to make all money less valuable. This means that people will seek to earn less of it.

> For example, if a distiller buys a tank, income tax is immediately paid. But VAT only generates revenue many years later for the country

VAT is paid on all transactions; that's the whole point of VAT. You're thinking of a sales tax that exempts business purchases. As soon as the tank is purchased, its seller must pay the appropriate VAT.


Yes, VAT is levied on the sale of the tank by it's manufacturer. But the distiller can claim back that VAT. This continues up the value chain except for the consumer who is not allowed the claim back VAT.

https://en.wikipedia.org/wiki/Value-added_tax


What do you want me to read in that link?

On its face, it directly contradicts you:

> Using invoices, each seller pays VAT on their sales and passes the buyer an invoice that indicates the amount of tax paid excluding deductions (input tax). Buyers who themselves add value and resell the product pay VAT on their own sales (output tax). The difference between output tax and input tax is the amount paid to the government (or refunded, in the case of a negative amount).

Though of course a distiller isn't reselling its still; it is the final consumer of the tank.

> [example of] 10% VAT:

> At each stage of production, the seller collects a tax and the buyer pays that tax. The buyer can then be reimbursed for paying the tax, but only by successfully selling the value-added product to the buyer at the next stage.

That reimbursement comes from the customer, not the taxing authority.


This link explains it better. And it proves my point: VAT on capital goods (in this case helmets) bought by VAT registered businesses are effectively refunded in full.

https://taxation-customs.ec.europa.eu/taxation/vat/vat-direc...


That link says... exactly the same thing as wikipedia, still directly contradicting you. There is a question mark here that I'll take up later.

But the example goes:

1. A vertically integrated helmet producer produces helmets out of nothing and sells a batch of them to a mine for a total price of €240. This producer has added "€200" of value, despite the sale price of €240, and owes VAT of €40 at a "20%" rate, which it pays out of the revenue from the sale of helmets.

2. The mine has purchased €240 euros of helmets and converts that purchase into €1200 of ore. This might look like adding €960 of value, but it isn't. It sells the ore for €1200, and since it has added "€800" of value, it owes VAT of €160, 20% of €800, which it pays out of the revenue from the sale of ore.

3. Because the ore sold for €1200, the government is owed "20%" of that, or €200, which it receives in one payment of €40 (when the helmets are sold) and another payment of €160 (when the ore is sold).

So far the terminology here is deeply stupid but logically coherent.

The question mark I mentioned before is what happens if the helmets fail to be consumed in the production of this batch of ore. Presumably they'll be reused to produce more ore. It would be odd if they still counted against the mine's "value added" for the second batch of ore.

I tend to suspect that the mine is considered to have added more value to the second batch of ore, but the page doesn't address the issue. This would mean that capital goods depreciate in full immediately.

Now, your original comment:

>> For example, if a distiller buys a tank, income tax is immediately paid. But VAT only generates revenue many years later for the country when the beverage leaves the store.

This is clearly false. Both of your links say it's false. The taxing authority receives revenue on exactly the same schedule that it would if the VAT were an income tax. Try running all the same transactions again with a VAT of zero and an income tax of 17%.


You have a lot of words here, but it still seems like the VAT is irrelevant here as it applies equally to both domestic and foreign production.

I agree the 10% tariff on US cars that is problematic, but that has nothing to do with the VAT


I'll try to summarize it for you. My first point was that with from the EU buyer's perspective, the German car can be delivered a bit cheaper because VAT is totally removed from it up until the end. The US car entering the EU market might have has sales tax embedded into the cost due to the differences in the tax systems.

To illustrate my second point, let's imagine both countries have a 10% tariff, and it takes $50,000 to build a car (ignoring point #1). In Germany, the German car's after tax cost is 50000*1.19=59500 and the US car's is 50000*1.1*1.19. The difference between the two cars in the German market is 5950.

In the US market, taking the average sales tax, the German car's after tax cost is 50000*1.1*1.085=59675 and the US built car is 54250, a difference of 5425. The difference between the cars is $525 tighter in the US market compared to the EU market. The German car makers get an advantage by virtue of how VAT is systemically higher and the resulting multiplicative effect, even when the tariffs are the same. When the EU tariffs are relatively higher, as they are now, the effect is even greater.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: