mhalle

Your comments are quite good and accurate. I would add that funding from private foundations is in many cases only viable because NIH overhead is paying for the supporting infrastructure. Very few biomedical researchers are supported solely by private funding (including the building and lab facilities).

These changes also directly impact research hospitals, not just universities.

That doesn't mean indirect costs shouldn't be reformed. This is just simply the most disruptive possible way to do it.

Some important points that this article glosses over.

The FAA Academy where all flight controllers are trained is way over-subscribed. Recruiting policies aside, I can find no evidence that the FAA wasn't training as many controllers as it could through its academy. This fact remained true through the Trump 1 administration into the Biden admin, except for COVID. The pandemic was understandably a huge disruption, as were government shutdowns.

We can know this from the FAA Controller Staffing reports from 2019 (Trump 1 before the pandemic but after Obama) and 2024 (Biden). The 2024 report has been scrubbed from the FAA website when I last checked, but is available through the wayback machine:

2019: https://www.faa.gov/air_traffic/publications/controller_staf...

2024: https://web.archive.org/web/20241225184848/https://www.faa.g...

There appears to be no urgency in Trump 1 about this issue in the report. Things changed in 2023 when an external safety report revealed the staffing problem and suggested improvements.

https://www.faa.gov/NAS_safety_review_team_report.pdf

As a result, hiring almost doubled between 2010 and 2024, with 1800 controllers hired in the last year. More importantly, the FAA followed the report recommendation to use CTI schools as additional academies:

https://www.govexec.com/workforce/2024/10/amid-hiring-surge-...

It seems like the Biden administration took real action to address a problem that had been unfortunately present and unacknowledged for many years.

See a chatgpt analysis comparing the two reports here: https://chatgpt.com/share/679eb87f-c4fc-800a-a883-3b7f79e06d...

XML would require a schema to express the concepts in Preserves, or JSON for that matter.

The reason JSON is lower friction than XML for data representation is that you get basic data representations (numbers, strings, arrays, maps) for free in a natural native syntax that happens to parallel multiple programming languages.

XML, in contrast, is a meta-language that allows schema to express different data representations. You've got to use attributes and elements to represent data and data types. XSD is a common datatype schema, but it's quite verbose, and data serialization looks very different from what it looks like in a programming language representation.

Preserves looks like a superset of JSON. It includes additional data representation concepts through syntax extensions, but the idea is the same.

What I don't see is a standard way to map record types (like "irl" in the tutorial) to a unique identifier like an URI/IRI, or something like a CURIE. That kind of feature would allow Preserves to better describe standardized record types.

Note that the Supreme Court decided the argument based on national security grounds, not content manipulation grounds.

Justice Gorsuch in his concurrence specifically commended the court for doing so, believing that a content manipulation argument could run afoul of first amendment rights.

He said that "One man's covert content manipulation is another's editorial discretion".

I completely agree that communication skills are critical in extracting useful work or insight from LLMs. The analogy for communicating with people is not far-fetched. Communicating successfully with a specific person requires an understanding of their strengths and weaknesses, their tendencies and blind spots. The same is true for communicating with LLMs.

I have actually found that from a documentation point of view, querying LLMs has made me better and explaining things to people. If, given the documentation for a system or API, a modern LLM can't answer specific questions about how to perform a task, a person using the same documentation will also likely struggle. It's proving to be a good way to test the effectiveness of documentation, for humans and for LLMs.

Unfortunately, that's a big challenge for a volumetric display, particularly a spinning one. There's a mass problem (heavier spinny things are more difficult than lightweight spinny things). There's also a wire problem if the elements on the spinning element are active and need to be addressed, since the wires generally need to come through the hub. If a projector is used and the spinning element is passive, then there's a tricky alignment issue.

And in any case, there's a bandwidth issue, since every 3d point is now addressed by a 1D or 2D array. That's an enormous amount of the data that ultimately needs to be uncompressed by the time it hits the display hardware.

Sure, it's possible to approximate occlusion for a fixed viewer. At that point, though, the display is competing with displays like looking glass (lenticular, parallax barrier, or stereograms) that are higher resolution, cheaper, and have a rendering pipeline that's more compatible with standard computer graphics.

"non-isotropic radiation profiles" is probably better expressed as the ability to produce light that can be modulated differently in different directions. That's what's required to show proper view dependent shading on a surface, and what's required to turn off a voxel that's behind another voxel from a particular view angle.

That's arguably a different class of display, and an extremely complex one.

To build it you'd likely want to sacrifice vertical parallax (to allow the use of a spinning lenticular sheet or something else, and to reduce the information capacity required), but there's no fundamental reason that prevents a full parallax display of this type (a spinning integral photograph).

From an engineering complexity point of view, it's the hardest of all worlds.

This device is a volumetric display, similar in product placement (if not exactly the same illumination technology) to the volumetric dome display from Actuality Systems about 20 years ago.

Volumetric displays have their place, but they can't display general occlusion of far objects by near objects. That restricts their application to non-photorealistic scenes that often look like clouds of points.

Since occlusion is one of our strongest depth senses (much stronger than stereopsis), that's a significant restriction.

Big spinny things are also hard to scale.

While other autostereoscopic display technologies like the parallax barrier displays from Looking Glass Factory give up the ability to walk around the scene, they work with no moving parts and can display photorealistic scenes (either synthetic or photographic).

Again, different display technologies have their place, but volumetric displays have historically struggled to compete in the 3D display market.

Lippmann photographs are like holograms, specifically reflection holograms. The image is exposed using self-interference, which works in what we would typically consider incoherent light due to a very short distance between the light and its reflection.

Development of the plate produces a superposition of many different volume diffraction gratings mostly parallel to the surface of the plate. If the plate is bleached, these diffraction gratings become high efficiency phase gratings.

For playback/viewing, light from the illumination source is both diffracted and filtered in wavelength by the volume diffraction gratings. In a hologram, the diffraction gives the multiple perspectives that make the medium so cool. For Lippmann photographs, the camera has removed most of the perspective information and the dichroic or interference filtering of the gratings is the primary effect.

In either case, the final image is a layer free image bearing volume in the emulsion.

That's why it can't be effectively copied using 2D techniques. Since the image is 2D (maybe slightly 3D? I'd have to think about that), it can be copied using a standard photographic technique. But the interference gratings in the emulsion have some angular dependence on the light source and viewer's angle that wouldn't be present in a 2D copy. In this way they also look a bit like a reflection hologram.

The use of {} for strings without substitutions and effectively for scope are actually interrelated, remarkably.

Scopes are special syntactic forms that delegate variable substitution to the associated procedures. For example, a "for" statement evaluates the body of the function after substituting the loop variable. In a string based language, that's basically the same as expressing the scope in a string form where the string content is passed verbatim as an argument without variable substitution or function invocation.

Having used Tcl extensively back in the day, I am not sure that this syntactic cleverness really was a major impediment to adoption. It was just something to learn. Same was true with [] meaning lisp-like function calling rather than array definition.

Baseline Tcl's biggest challenge, in my opinion, was providing mechanisms to write modular code for larger programs and data encapsulation. Core Tcl put off decisions about the appropriate mechanisms to do so by only providing "namespaces" as a building block for higher level third-party syntax. That led to fragmentation at a time when other languages were gaining popularity.

NFA == National Firearms Act

https://en.wikipedia.org/wiki/National_Firearms_Act

Nice work. From another comment I see that you chose this project in part to learn three.js, but you might also consider using deck.gl or kepler.gl.

They provide higher level visualization primitives and are optimized for performance with large datasets. They would also allow you to try out different visualization methods more easily, or even overlay multiple ones.

Both packages allow for very fast prototyping (with kepler.gl, you might be able to mock up something with no code at all), so trying out this approach shouldn't involve an enormous amount of work.

Good luck!

When I had my wallet stolen in London about a decade ago, Amex was far faster, easier, and friendlier replacing my credit card than was Citibank. Did it at an in-person Amex office.

Michael Crichton's book Airframe provides a fictional but well-researched look at many of the more hidden issues involving aircraft safety and maintenance, as well as the relationship between carriers, manufacturers, and the FAA.

Anyone who gets a transplant needs to take immunosuppressants for the rest of their life (except some older patients with liver transplants many years post transplant).

In general for many patients, immunosuppressants have relatively little impact on daily life.

Source: liver transplant recipient.

I think the true future of anatomy atlases is to address the kinds of applications you're describing. Open source anatomy atlases can become toolboxes of visual anatomical knowledge. Subsets of atlases can be used for patient education, with structures labeled and styled to meet the needs of the audience. Atlases can also be used as visual indexes into collections of medical data.

Finally, we need atlases that represent the breadth of humanity. It's far too easy to view the "typical anatomy" represented in a major atlas as "normal anatomy" with any variations being abnormal, rather than just less common. Open source atlases can fix all of those problems, ultimately providing a comprehensive and collaborative view of what human anatomy really is.

The title is not correct, it should be:

The first open source 3D atlas of human anatomy in Blender.

There have been a variety of open source human anatomy atlases, though few of the entire human body. Not to take away from the great work that has gone into Z-Anatomy or the underlying data from the BodyParts3D project in Japan. The BodyParts3D data is licensed Creative Commons Attribution-Share Alike 2.1.

Disclaimer: I am the PI for the Open Anatomy project (https://openanatomy.org), which is working to build open source anatomy atlases for worldwide use. We also have an anatomy term viewer https://ta2viewer.openanatomy.org that uses the same nomenclature that Z-Anatomy uses, and which is linked from many Wikipedia anatomy articles.

[Edited to add the bp3d license.]

Beyond the risk of medical bankruptcy in the US, lack of health insurance can prevent you from getting some lifesaving treatments.

If you are uninsured, for instance, you stand almost no chance of getting an organ transplant. There are not enough organs for too many people, the operation is very expensive, and lifetime support medication is also expensive.

In my experience, transplant centers won't even start the evaluation without proof of insurance.

While it might be possible to get insurance after diagnosis for people like the OP who choose not to have insurance, that isn't always feasible. Sudden organ failure, while rare, certainly happens. And should it happen, it is hardly the best time to be scrambling to find insurance.

Yes, the system is screwed up. But it is the system (in the US).

I believe that the 737NG is not the problem aircraft. The new engines and the problematic MCAS are on the 737 MAX, a different model.

This American Life had an episode that challenges the premise of this book: https://www.thisamericanlife.org/806/transcript

A related story is that the Walgreens CFO admitted during a recent earnings call that the company may have overstated "shrinkage" of inventory and that it had spent too much on "largely ineffective" private security to combat it: https://www.cnbc.com/2023/01/05/walgreens-may-have-overstate...

It is hard to fully understand what caused the disconnect from reality.

Here's a good NPR piece discussing the issue: https://www.wbur.org/onpoint/2023/12/18/is-shoplifting-getti...

Kids who go to summer camp where they (thankfully) aren't allowed to use phones are a great market for mosquito itch reduction devices.

The non-USB Beurer BR60 works very well, doesn't require a phone, and is about half the price. We have used them and found them very effective.

I think your statement "adding to existing fulltext functionality" is subtly important: embeddings provide semantic search that complements traditional search algorithms.

Specifically, many applications are heavily dependent on names or other proper nouns, often without much context. You might refer to your dog by name without explanation, and a particular embedding might not pick that up. Proper names (people, places, street names) may have outsized importance for anchoring personalized or domain-specific search, but modest generic language models won't know about them.

Is there a specific way of dealing with this problem?

I was recently very pleased to find pre-cut "LED light dimming stickers" on Amazon.

My family enjoyed a half hour early evening "search and stick" mission to attenuate our home's luminous loudmouths.

Also note that the Amazon padded packaging says to remove paper labels before recycling, but it practically impossible to remove the Amazon paper labels (at least in my experience).

Here's another in-depth article on the safety issues surrounding a particular generic for the drug tacrolimus (Prograf is the non-generic drug).

https://www.bostonglobe.com/ideas/2019/05/24/ideas-katherine...

https://archive.is/2023.02.27-233237/https://www.bostonglobe...

Except that pedestrians face the same or greater risks than bicyclists from vehicles from higher front ends.

The biggest challenge is that pedestrians, bicyclists, and drivers tend to live, work, and shop in the same locations. Even with separate facilities (sidewalks, cycle tracks), there will be conflicts (intersections and driveways). Lack of visibility and a disparity of mass and velocity are always going to be hazardous at these conflict points.

I would assume that there is at least some constant personnel or time cost per audit that the IRS incurs. I would also assume that even with recently increased funding, the IRS only the capacity to audit a small sampling of the taxpayer base.

In that case, from a revenue recovery point of view, going after the highest income / highest discrepancy taxpayers produces the greatest return on investment, with some randomness involved to keep everyone on their toes.