I find nothing particularly incomprehensible about laws in general. In an experiment where you ask a bunch of amateurs to write legal documents I'm not sure you can apply any real interpretation to those results.
They use the example of DUI laws. Here's two. There's nothing particularly complicated about them, and the "center embedding," to the extent it is present, is entirely comprehensible.
I have looked up multiple British laws over a wide range of areas (from landlord and tenant to education to corporate) for practical purposes and it is mostly reasonably simply written given the need for precision and to cover edge cases.
The article says "researchers plan to analyze British laws to see if they feature the same kind of grammatical construction." Not in the last half century or so. I cannot recall having read any older legislation recently.
It's one of the examples presented in their article, I did not pick it at random. Perhaps when people without subject matter experience write laws or contracts they end up as the paper suggests, but when people understand the problem domain clearly, the laws end up on paper with equal clarity.
The paper notices the fact but draws the completely wrong conclusions.
Fair enough but now this name leaks out into code I have to type. I don't dig highly opinionated languages with tooling that complains at me over style guidelines breaking their own style guidelines to solve a problem that they themselves have created.
They painted themselves into a corner and their solution is for me to hit myself in the head with a hammer.
Generally people ignore the per PUT and GET pricing on S3 along with the higher latency since it's a "global" service. If your objects are small then you're almost always benefited from using DynamoDB as the GET pricing and latency are far more favorable, as long as you don't mind the region dependency or the multi region setup.
Even if AGI suddenly appears we will most likely have an energy feed and efficiency problem with it. These scaling problems are just not on the common roadmap at all and people forget how much effort typically has to be spent here before a new technology can take over.
> Is there some expectation that these things won't improve?
Yes. The current technology is at a dead end. The costs for training and for scaling the network are not sustainable. This has been obvious since 2022 and is related to the way in which OpenAI created their product. There is no path described for moving from the current dead end technology to anything that could remotely be described as "AGI."
It's a group with no real authority. Their only option is to create reports. In some cases the Administration may act directly on them. In the majority of cases it will be a report delivered to the legislature that they will then completely ignore.
The largest budget programs are entitlements which the Trump team seems powerless to change and thus has no stated desire to engage with them at all or the Pentagon which is _the_ pork barrel machine as far as congress is concerned.
> Can you imagine the costs for corporations for email legal compliance?
7 years of storage in an environment controlled by SOX. There is no requirement that the storage be online or accessible by the originating user, and in most cases, there's good reason to specifically avoid this. As a result it's not as bad as you'd think.
They use the example of DUI laws. Here's two. There's nothing particularly complicated about them, and the "center embedding," to the extent it is present, is entirely comprehensible.
https://leginfo.legislature.ca.gov/faces/codes_displaySectio...
https://www.revisor.mn.gov/statutes/cite/169a.20
Particularly clear both to the layman and to officers of the court.