dwallin

They aren't eliminating gig work, they are just ensuring that if you want to categorize your workers as contractors, you have to treat them as contractors. Companies are taking advantage of loose definitions to blur the lines, they want the benefits of employees without the liabilities. This forces them to decide, do you actually want contractors, or do you want employees? The workers that stay as contractors will have more freedom and ability to set their own work boundaries. The workers that get converted to employees will get better benefits and stability. On average, workers will be better off. There will be companies going either route so gig workers should be able to find a company that matches their needs.

Urban density doesn't cause the issue of traffic. The issue of traffic is caused by cars. Urban density allows public transportation, the only current viable alternative, to be financially sound on a large scale.

LLMs don't lie, they are instead ALWAYS hallucinating. Sometimes those hallucinations will coincidentally match a fact from training data, but a LLM has no concept of truth. They are far more reliable when you restrict them to transforming information than when you allow them the latitude to generate it.

More so than just compressing the data, I feel like the biggest gains for splats will come from a chunked LOD system, and streaming data into memory. Regardless of how efficient you make the representation, you are still going to run into fundamental limitations without this.

On the low hanging fruit side of things, tools should really start integrating a Spherical Harmonics skybox into the training steps to better handle large scale distant details.

I felt the article pretty thoroughly explained why the author feels that hives are not currently respected as an extended phenotype. Many studies of bee behavior are run on bees in artificial hives. They are treating artificial hives as interchangeable with their natural equivalents and assuming that the behavior observed reflects normal bee behavior. This can lead to many erroneous conclusions, the same way you might get misleading studies on beaver behavior if you studied a bunch of beavers living in a rabbit hutch. To quote from the article:

> Honeybee (Apis mellifera) colonies don’t hibernate. In the wild they overwinter in tree cavities that keep at least some of their numbers above 18°C in a wide range of climates, including -40°C winters. But popular understanding of their overwintering behaviour is dominated by observation of their behaviour in thin (19mm) wooden hives. These man-made hives have very different thermal properties compared with their natural habitat of thick-walled (150mm) tree hollows.

This is really cool to see, I didn't expect anything at this level for a while even if the performance is a ways off. Supposedly they are getting ~30 FPS for simple scenes?

System prompt have proven time and time again to be fallible. You should treat them as strong suggestions to the LLM not expect them to be mandates.

Are you truly going to proclaim that we have explored all options for how to structure a global economy? We arrived where we are via a mix of self-interest and local optimization and it seems unlikely we have arrived at even a local maxima.

I'm not going to claim I have the solution. Anyone who tells you they do is either deluded or has something to sell you. And that definitely includes those who want you to believe that modern capitalism is the end-all be-all of society.

What I want is to stop the self-imposed helplessness. What are some things that might work? How can we validate them? How can we roll them out?

"Corporate headquarters are easily moved, and Corporate income is easily shifted between corporations and individuals in different tax jurisdictions all over the world."

This is a consequence of policy choices, not some inherent natural law.

Except human beings are not optimization machines and the best way to get great people to bring all of themselves to work is to allow them to be all of themselves at work.

The best and brightest have diverse interests and if they get bored they can just move on. The cost to the company to let them mess around with generative AI for a couple of weeks is far cheaper than having to replace them if they jump ship elsewhere.

Plus, this builds internal expertise with a new area of technology that may or may not end up relevant to their business in the future.

This is like looking at an algorithm, thinking of the worse case scenario, running a single step, and deciding that it's not worth it from that alone. This is not a one-and-done rule, and the same dynamics that might push an old lady to move in one situation, might make things much better for her in another, and more importantly, over time might lead to a better situation for her grandchildren when they grow up.

In practice, that grandma is stuck in a city where her children and grandchildren can no longer afford to buy real estate in. Friends and neighbors who rent are priced out and have to move regardless. The local goods and services rise in price and taxes go up, all under the current system. Her support network and community evaporate and soon she finds nothing left but to sell to a corporation, who is the only entity left that can afford the prices her house demands.

The first order effect of a Land Value Tax is that it shifts who pays the biggest burden of taxes. Grandma A who owns a big house in the center of a vibrant city where lots of young families are looking for homes near their jobs might see an increase in her taxes, but the Grandma B who owns a Condo in a larger complex will likely see her taxes drop. Grandma C who lives in a nice well-maintained house way out in the burbs will likely see a drop as well, most of her value is in the nice home she has been tenderly caring for her whole life.

A second order effect of LVT is it creates a suppressive effect on real estate prices. When prices in an area are going up across the board, this is entirely in the value of the land, which means LVT taxes are going to rise correspondingly. This additional tax liability significantly helps counteract a speculative rise in prices, leading to a lower and more stable pricing, and pushing out those who are just looking to speculate on trends. The value that a government creates for its citizens is recaptured and able to be reinvested, instead of going directly in the pockets of land speculators.

The third order effect is that this shifts the incentives around value creation vs value capturing. Perhaps the house next door to Grandma's is currently owned by a deadbeat landlord, Pennybags, who lets the house fall into disrepair, hoping to turn around and sell it as the market rises. Now all of a sudden Pennybag's plan is quickly becoming a bad investment. With LVT his property value is more heavily correlated with the actual quality of the house and less with value of the land. He is finding that despite an increasing demand in the area, his property is actually dropping in price. Meanwhile good ole Grandma has been diligently been taking care of her home, and adding value and the price of her home has actually gone up faster than the rise in taxes. She's able to use to use the increase in value to take out a loan to build an accessory unit she uses to now help her cover her increased taxes, which further increases her property value. Despite all of that, her taxes have not gone up at all!

Seem fairly tractable to use Segment Anything or a similar method to derive plausible semantic clusters of splats.

#2 seems very dependent on the competitive landscape in your market. If you have well-funded competitors growing quickly and taking market share you could absolutely fail due to not scaling rapidly enough.

Great, you gain more sales and immediately go out of business since you were already pricing your product near cost.

Prices then rise back up as your nearest competitor was smart enough to not lower price below their cost.

Or do you somehow believe that price competition currently only exists everywhere EXCEPT on Amazon's platform?

Clearly you are not a believer in free markets, because by the theory of free markets these prices would naturally arrive at the lowest possible price all on their own.

You are right that this policy is very simple, it's very clear both mathematically and logically. Companies cannot lower their prices lower than their costs. If costs are variable across platforms but prices are fixed to a single value than this will de-facto result in INCREASED prices. If this policy causes someone to leave a platform, that STILL doesn't result in prices any lower than they would absent the rule.

In fact, the sheer existence of the rule is a clear statement from Amazon that they believe they have significant and durable market pricing power, since if they didn't, this policy would clearly lead to sellers and buyers leaving their platform.

You are a seller. Amazon makes up 40% of your sales. You have your own first party website where you can profitably offer your product for 5% less and still make an additional 10% profit but it only makes up 5% of your sales currently.

What's your move in this position?

You can't just imagine all companies as perfect frictionless spherical interchangeable cows. In reality companies as large as Amazon can have significant market distorting effect, while much smaller companies might have to choose between their principles and going out of business.

It's not pro-consumer in any way. If the seller can profitably list it for a given price on a given platform, competition will almost certainly drive their price down to that level. The only time a policy like this becomes significant is if another platform, or the sellers own platform, can offer the product for cheaper than Amazon (eg. maybe they can ship straight from the factory), in which case this policy would force them to raise their prices, while simultaneously removing potential competitive pressure on Amazon.

Ideally you decide by voting. It’s either that or you decide things by who has the most money / power / guns.

Seems like the real utility of this technique will be as a way to vastly improve the temporal stability with a variety of generative video techniques. For example, if you are trying to use a video as a base for a new generative video: Take the first frame of your video and run it through SD with the control net of your choice. Then take that initial image and run it through this process to generate a new base model and then use that to generate your second frame. Now you can use that second frame to feed back into your model and rinse and repeat, always using the past few frames to inform the latest.

It’s not an either/or situation. You have a bunch of tools with pros and cons and you have to use the proper tool where each one is most appropriate.

The best transportation systems are a blend of transportation types. Use rail where the benefits are worth the cost and buses where they are not. There also many incremental options in between.

In this case it’s likely not. The advantage of Gaussian splats is that it allows you to bake in advanced lighting effects for a static scene. If you already have detailed renders there are plenty of existing approaches that perform plenty well and can be far more optimized.

The devil is in the details (which was the problem with the original terms). The original plan seemed reasonable at face value but when you dived into the details it was really problematic. The revised terms are transformatively better, but might seem similar just looking at them shallowly. The key bits are

1) It is no longer retroactive. Existing games on the market using unity can continue distributing without their business getting immediately impacted.

2) It's no longer per install, it's now "new people engaging with your game each month" and the numbers are now all self-reported. Originally Unity was going to calculate installs on their end in an opaque way.

3) You can choose to just use a revenue share arrangement instead of per install. Since it's the lower of the two you can effectively ignore whichever one isn't favorable to your business.

Immigration is such a straw man, an influx of people with generally low capital availability and incomes is clearly not the cause of expensively priced housing and rents. Saying government regulation causes it is also meaningless unless you specify WHAT government regulation you believe is causing housing issues, historical evidence clearly indicates that no regulation doesn’t lead to the outcome you desire. It’s very clear that the key is figuring out what sorts of regulation, taxes and laws would lead to an optimal solution and in order to begin to do that you really need to break down how the demand and supply of housing actually works.

The demand side of rent prices is generally based on the surplus value created by an area for its residents. The value of rent prices is all about providing signals about relative utility to potential residents. This is key because it means that absolute rent is largely unimportant, the relative rent is the key. All rent paid above actual services provided does not create economic value, it instead captures surplus value from its residents and delivers it to rentiers. The issue with rent control and other similar regulation is not that it limits overall rent levels, it’s that it distorts the relative signaling. However this needs to be balanced with the negative effects large rent increases have upon tenants. A world where people are constantly forced to move due to a rapidly changing market might be price efficient, but it creates lots of negative externalities. The key here is establishing rent rules that are loose enough to create accurate price signaling, but tight enough to minimize these externalities.

The demand side for purchasing for individuals is based on individuals and companies evaluating on a range of criteria which largely boils down to the utility it directly provides, the rent you can charge others for their utility in the properly (or conversely, the rent it allows you to avoid), and its value as a speculative investment. However it is clear that in the problematic markets we are discussing, the prices are extremely dominated by the latter two.

This leads to the supply side of this also being somewhat counterintuitive. People seeking to invest their capital in property actually have two primary ways of profiting: via rent collected or by appreciation in property value. If you own property, and you have additional capital to invest, do you invest in improving your existing housing stock or into buying additional property? The answer is you invest in whichever has the highest rate of return. In low housing stock environments the answer is going to typically be to invest it in additional property over improving existing properties. You might think that rising rents would change this ratio, but this gets it backwards. Rising rents leads directly to increased property values and so largely has the opposite effect, it encourages investment in buying property over improving it.

This makes it clear that to increase investment in housing stock, your best method is to discourage speculative investment in favor of investments that raise the quantity and quality of housing stock. The way you structure regulation, incentives and taxation should reflect this. Everything else is noise, and doesn't actually solve the problem.

Side note: An exception here happens in cases where development leads to massive differential between start and end value. This is why in these sorts of high speculation environments development becomes highly slanted towards turning low density buildings into massively high density, high margin buildings. This is generally not ideal, as this type of construction has lots of negative externalities on its surrounding neighborhoods, and tends to make communities averse to further development. Making lots of small incremental improvements in housing stock is far less disruptive to neighborhoods but will only really happen in environments slanted towards improvement over speculation.

Just because a behavior may end up profitable for its owners doesn’t mean we should be just accepting it as a society.

Also viewing a company as just a collection of assets owned by capital can be somewhat limiting. Companies are generally made up of human employees, and many schools of thought treat employees as one of several stakeholders in a company and assign various rights to them and various responsibilities to the company for them.

Looking at it this way, the “owner” of a company can easily be described as looting the company if they are destroying a lot of value in the company for marginal benefit for themselves.

It’s not the fact that Unity wants to make more money from its engine that has gotten people so up in arms about it, it’s the way it was structured.

If they simply took a percent of revenue, everyone would have shrugged, you might have had some indies complaining, and the world would have kept on turning. Instead, they made a couple of compounding missteps. An install based fee creates a massive amount of uncertainty about how much you will owe at the end of the month, creating potential cash flow issues. You are at the whims of Unity in terms of how they calculate it, and praying that they properly handle all the edge cases and potential abuses, despite all their incentives giving them plenty reasons not to care that much. And then on top of all of that they made it retroactive to all existing games on the market using their engine. Companies had a very reasonable expectation that they could rely on the terms their games were released under. All that trust has now been upended. Unity has shown they are willing to change terms retroactively with little notice. Now when evaluating Unity as an engine you will have to take into account not only the current fee structure, but also the risk that they will add additional fees retroactively that have the potential to completely disrupt your business model.

Penalizing private car owners in favor of taxis in a borough where less than a quarter of households own a car seems about right.

In addition, it gives the United States and EU the chance to catch up and build up their domestic semiconductor capabilities without being undercut by chips from China. Countries have too slowly realized the importance of this and are playing catchup after years of underinvestment.

No, you just need a properly set land value tax to discourage simple rent seeking behavior from driving up the price of land / housing.

It's not even really inefficiencies, it's just straight up queuing theory.

As you approach maximum throughput, average queue size – and therefore average wait time – approaches infinity.

In order to have a quick and nimble organization you need to leave sufficient slack in your utilization of employees.

Serving size is an often arbitrary and gameable number. This is a case where the fda need to improve their definitions. Eg. zeroing-out amounts could take serving size into account.