dwallin

It's probably better to just link to the Rand Report: https://www.rand.org/pubs/research_reports/RRA2680-1.html

I think a better example of this type of moderation would be Wikipedia.

Except that there are significant switching costs. A person moving has to pay moving costs, might have to replace furniture that doesn't work for the new place, can affect your children's schools etc. This means the value of a unit to someone living there can often be significantly higher than the market rate.

This can led to situations where the most profitable move for landlords is to take advantage of the discrepancy to regularly raise the rent for existing renters by more than the market, in attempt to maximize profit. Sure they might have to deal with the hassle of finding a new tenant every couple of years when someone gets priced out, but if it leads overall to slightly higher profits it's the winning capitalist move.

The rentiers are ostensibly following the law, but the overall cost to the population and quality of life loss to renters can be significantly outsized compared to a sliver of additional profit for the rentiers. This is a great example of externalized costs in a free market and exactly where government should generally be attempting sensible regulation.

Unfortunately, the way we structure, fund, and manage modern corporations creates tons of poor incentives all on its own, no legislative influence needed. (unless your goal is to maximize short term profit signals at all costs)

This is not a binary situation. There are plenty of reasonable approaches that help limit abusive landlord behavior without damaging the prospect of profitable real estate development.

Humans have a natural tendency towards social hierarchies. If you don't provide structure people will instinctively create it; so attempting to entirely remove structure from an organization is idealistic and inevitably fails. This usually leads to hidden power structures and counter-productive popularity dynamics. A great classic read about this topic is essay "The Tyranny of Structurelessness" by Jo Freeman: https://www.jofreeman.com/joreen/tyranny.htm

The simplest way to encode a structure is with a basic hierarchal structure where power gets delegated and directions flow down from a single individual at the top, and information gets filtered and flows up to enable decision making. This is one of the most simplest and common structures you see across society. It is a structure great at quick, efficient decision making, but has numerous flaws that make it suboptimal in many cases. Notably, the single-directional flows means it's particularly bad at self-regulating, and therefore it's susceptible to corruption without a significant outside influence.

However you can leverage systems and technology to engineer and enable novel and durable alternative structures and power dynamics. On a societal level, democracy is a hugely successful example of a system like this. Elections create a loop from the bottom to enable accountability for the person at the top, helping solve a number of failure states. You also have techniques like creating multiple structures that operate in tension, structures that operate entirely via democratic consensus, etc. Each structure has different strengths and weaknesses, and combining them well can be used to create high-functioning governments.

At the level of corporations however, you see minimal exploration of optimal structure. The modern corporation has seen some innovation, but this happened almost entirely at the ownership level, with boards and shareholder elections etc. The actual executive functionality of most companies is almost entirely your standard hierarchy. It beggars belief to think that this would be the optimal operational structure across all industries. The reality is it's the structure that maximizes shareholder control (and therefore shareholder value). Other structures might enable an organization to better serve the market, employees, etc ; but these are not the concerns of the people setting up and funding said corporations.

Apple is selling a platform to end users and they hold durable market power with that platform, due to high switching costs and minimal alternatives. End users consume services by various third parties on the platform. In addition Apple is also offering services on the platform that compete with those third parties.

It's entirely fair to say to Apple (and companies in similar oligopoly positions) that if you want to compete with third parties on your platform then you need to do so on a level playing field.

It's not then about forcing Apple to build a feature at no cost, but rather declaring that Apple built their NFC feature improperly, by building it in a way that prevented competitor use, illegally privileging their own competing services. They can either remove the feature or build it properly, according to relevant law.

Apple has had no problems with providing their macOS libraries, tooling, and associated updates for free, even with a drastically smaller user base, and it appears to have been sustainable for decades.

The truth of the matter is Apple would almost certainly continue to do offer their iPhone services, updates, app store, etc; even if they were statutorily prohibited from making any money off of any of it, because in the end it makes for a better user experience and helps them differentiate and sell their phones at a premium.

What is happening, is the more financial-minded types within Apple have realized they have substantial lock-in and market power, and they can leverage it to take a % of profits from other market entities.

If you actually care about Apple products, and want them to continue to deliver great products well into the future, then the best thing that could happen would be for them to get the message and accept a loss in this battle. Otherwise you will see increased financialization continue until they are thoroughly hollowed out of great product/engineering culture, and collapse under their own weight, like so many before them.

https://www.youtube.com/watch?v=P4VBqTViEx4

Assuming a US perspective, welcome to a world of unlimited corporate political spending (Thanks Citizens United). If you stick your head up to try to bring such a bill to the table expect your opponent next election season to come with a massive PAC-fueled war chest.

IMO, this is the best think that could happen for Firefox. Mozilla as an organization has been thoroughly corrupted by huge buckets of money being thrown at them by Google. "The Resource Curse" (https://en.wikipedia.org/wiki/Resource_curse) doesn't only apply to countries.

Scale and context matters. No one else could afford to pay anywhere remotely near what Google is paying Apple. If it was truly competitive bidding Google could just offer some small reasonable percentage more than the nearest competitor could afford and Apple would take it. The reason google is paying so much over where a market price would be is because they want to discourage anyone, particularly Apple, from even thinking about competing with them. Google is paying Apple 36% of the revenue (not profit) generated by those users.

A pretty good argument could be made that Apple would likely have built up and/or acquired it's own search engine by now if Google wasn't paying such obscene amounts. Apple was previously involved in talks with Microsoft to purchase Bing, and even acquired an ai search startup founded by ex-googlers (LaserLike).

This is a misleading title. They said no to whether they are currently considering any additional US joint ventures. The title implies that they are saying no to a specific joint venture, which is not the case, all existing joint ventures are continuing as planned.

You are missing the key bit right after that where they basically scolded the 5th circuit for not having a basic understanding of free speech. They then proceeded to outline in detail why said analysis should almost certainly lead to the law failing to pass constitutional muster. This was not the Supreme Court throwing up their hands, this was the Supreme Court saying "This is basic stuff you shouldn't be getting wrong, try it again."

> It is necessary to say more about how the First Amendment relates to the laws’ content-moderation provisions, to ensure that the facial analysis proceeds on the right path in the courts below. That need is especially stark for the Fifth Circuit, whose decision rested on a serious misunderstanding of First Amendment precedent and principle.

I don’t put much value in polling claims that don’t disclose how their question was phrased. The difference between slight differences in wording can be transformational. What this also doesn’t capture is how strongly people feel about these policies. It’s very easy for people to casually support the status quo.

Although by adding small amounts of possible allergens to a bunch of common food items you help reduce the frequency of future generations developing the same allergy. So it's a bit of a mixed bag there.

I disagree. The dissent focused way too much on emotional appeals. It shouldn’t matter if the remedy was popular amongst the primary parties if the law doesn’t allow for such a remedy, “it was popular” is not a solid legal footing.

If we want to make sure that the victims of people like the Sacklers can get some justice then let’s actually write some laws for that purpose. Let’s remove the loopholes that let them make a ton of money at the expense of a nation, and them take it all offshore. Let’s add transparency to corporate structuring. Let’s fully pierce the corporate veil for egregious situations such as this.

What we shouldn’t do is allow the powerful yet another tool to escape consequences. Whether or not this particular deal was a net-positive for victims is besides the point. It gives too much power to bankruptcy judges and is a ripe avenue for corruption. Go to a “friendly” judge and have them absolve you of any personal liability, then send them a nice gift basket full of Benjamins, now fully legally (Thanks Supreme Court!).

I would maybe say that it is almost assuredly one of the most featureful and widely used layout technology. But I'm not convinced it's the most powerful, flexible, or even advanced. It suffers from significant performance issues due to legacy baggage and a poor underlying abstractions. There is a reason anyone building things like visual editors, spreadsheets, et al tend to drop into lower level rendering technologies.

They tested their models on both a conventional gpu as well as a custom FPGA.

Yeah, there are some applications in the area of SLAM that benefit from quickly generating plausible gaussian representations.

I’ve heard about this one but don’t have an appropriate setup to test it myself: https://www.gracia.ai/

I do agree that new use cases are emerging and it will probably enable tons of new businesses. I'm very gung-ho about the technology myself as well. I guess what I'm trying to say is that the new businesses that emerge because of this are not necessarily going to advertise that they use gaussian splats to do it, it's not a buzzy enough term, and many of the industries it serves just care about the results it delivers. Your average tech person is unlikely to hear much about it. Your average graphics engineer will have probably heard about it, but not know about all the use cases that are leveraging it. And your average person in the industry it is changing won't know what is causing the change (they will probably assign it to the nebulous ai bucket). I fully expect gaussian splats to be a quiet revolution.

It's important to disentangle this specific approach of generating a gaussian splatting model from 2d images, with gaussian splats as a general rendering technique.

There is nothing that prevents gaussian splatting from being used dynamically. There are a variety of approaches to extend gaussian splats into the time dimension to capture and represent a 3d scene over time. The challenges here about how to capture sufficient scene data (or use ai to fill in insufficient data) and how to compress it. There are also techniques that enable dynamic simulations, or real time animation of collections of splats.

Also, adding un-baked lighting to gaussian splats is not particularly hard, you can already throw slats into several game engines / 3d renderers and add new lights to them. The hard part of relighting is taking an existing capture of a scene with baked-in lighting and deriving the resulting material properties and lighting sources. This isn't directly related to gaussian splats themselves though, you would have a similar problem recovering the base materials and lights from a 3d mesh with baked-in lighting textures. This really falls under a separate category of techniques called "inverse rendering". If anything, gaussian splats give us a new tool to help with these sorts of problems.

Honestly the biggest remaining roadblocks to more elaborate and widespread uses of gaussians as a rendering method are probably storage and performance related. And I'm optimistic these will be convincingly solved, triangle rasterization has had many orders of magnitude more research, optimization, and custom hardware built around it.

There's a lot of use cases, and companies already using photogrammetry techniques have adopted it extremely quickly, but for many of these businesses the usage of gaussian splats is just a technical implementation detail, resulting in quality / feature gains, but doesn't as of now unlock entirely new business models.

No the main difference is looking at the problem from a new perspective, relating it to a large body of existing work in statistics (see https://en.wikipedia.org/wiki/Markov_chain_Monte_Carlo and https://en.wikipedia.org/wiki/Stochastic_gradient_Langevin_d...). Then they are able to use this new perspective to add several improvements that lead to significant quality improvements, clearly establishing the validity and utility of the new theoretical underpinnings. It's likely this will be massively influential in the direction of future research in the space.

The actual changes this led to are:

1) As you mentioned, they added noise. But notably, they state it was "designed carefully" to conform to the requirements of SGLD and they detail how they designed the noise.

2) They simplified the original operations of "move, split, clone, prune, and add" and their related heuristics, into a single operation type. They do so guided by existing knowledge about MCMC frameworks, leading to a simpler model with stronger theoretical underpinnings (huge win!).

3) Adjustments to how gaussians are added and pruned to better fit with the new model. This seems more like housekeeping rather than something novel in and of itself.

One major problem with the Vision Pro as a devkit is it's extremely locked down. The only significant AR experiences you can create are those already anticipated by Apple and compatible with their privacy expectations.

IMO this was a huge mistake. This is a completely new form factor, not an established market. You need people doing significant experimentation, and to be able to create compelling new experiences that your in-house staff haven't even dreamed of. It's fine if you have a user-friendly default, but there should be some way to opt into a professional mode that opens up more possibilities for novel experiences.

(IANAL) Assuming it's not a copyright violation, just a contract violation, if you get access to the model weights without entering said contract you could make a strong case that none of the restrictions apply to you. However, as far as I can tell this is not settled case law, and the entities distributing these model weights are definitely hoping for them to qualify for some level of copyright protection. However, even if thats the case, the resulting images are not currently subject to copyright in the US, and can be used freely by non-parties. Even if that changes down the line, these images would almost certainly be considered derivative works from the model, and these models already assume that training is fair use anyway.

You solve this by implementing the LVT as a small tax rebate on top of traditional property taxes. Something along the lines of an "Improvements Rebate" or whatever polls best. This is much more politically palatable, you can pair it with a small increase in tax to some currently unpopular sector. Very slowly over time you can increase the property tax, while increasing the LVT rebate. Now those who might want to kill LVT now have a more complicated job of negotiating simultaneously to remove a popular benefit while adjusting overall tax rates, a much more complicated proposal.

LVT also has the drawback of being more complicated to calculate than property taxes, and therefore makes convincing constituents that it is being implemented fairly harder. When it is a negative (cost), this complexity makes it easy to attack, while a positive is more defendable (sure it's a tricky calculation, but it gives me money back in my pocket).

Generative AI is progressing at an extremely rapid rate. What many are experiencing though, is that every company wants to be perceived as being on the cutting edge. Being on the cutting edge gets you mindshare, free publicity, investment, stock price bumps, etc. This incentivizes all this bluster and false promises. If you just take a shallow look at everything, it would seem like an unending wave of this, and it's easy to get jaded. However right behind that is a not so slow march of people doing actual work, building non-trivial products, or using the technology to create tons of small, sometimes invisible, but cumulatively transformational, improvements to their products.

If this is true, it's almost certain to lead to a shareholder lawsuit.

Oh, you are right, I missed that inversion. Although it's worth noting, the absolute employment rate for graduates doesn't seem particularly out of line historically speaking. It might be more accurate to say that the current low employment rate is affecting primarily existing workers in the market.

I don't think looking at the date of the crossover point is particularly helpful though, probably better to look at the derivative of the ratio between the two to see when the trend began. At a glance it looks like it might have began in 2009 during the recession, which actually matches up really well with when labor participation rates saw a big drop off. My off-the-cuff hypothesis would be that this effect could be explained by existing workers leaving the labor market, which wouldn't necessarily affect recent grads. I wasn't able to find a historical labor participation rate for recent grads though, which would help support that hypothesis.

For general labor participation rate information: https://fred.stlouisfed.org/series/CIVPART