The purpose of this agreement isn't to actually restrict access in the Pentagon. Its main purpose is to provide a plausible legal means to prosecute journalists who write things that the administration disapproves of. It is meant as a threat, not as a restriction.

Unless there is a hot military war. Or a hot economic war. The US government targets the rights of individual Americans to advance personal political ideologies. While the Chinese don't care so much about those rights, they are very much interested in stealing industrial and military secrets. Some of these secrets are advantageous in a potential hot military war, but a lot of those secrets are already advantageous in promoting Chinese companies and jobs over American companies and jobs.

The title talks about a software quality collapse. That implies that software quality was better in the past. While we all bemoan the current state of software quality, how many of us actually think, "Remember the good old days when software quality was good!"? Or is this complaint-fest simply an acknowledge of the ever-present challenges of engineering software quality, challenges that are not necessarily different than before. As with all things in life and not just software, it's easy to complain about the present and nostalgically reminisce about how things were better before.

Regulations aren't the only problem. There are also other issues such as lack of institutional knowledge, lack of skill workers, huge capital costs, very long ramp-up times to create facilities and supply chains, and challenging economic viability especially in the face of cheap Chinese competition. Regulations often address environmental issues, and these exist at many levels, including federal, state, local, and private interests. Sometimes these interests manifest as regulations, but they can also manifest as lawsuits, efforts to interrupt required ancillary resources, public relations campaigns, etc.

Isn't this sort of like arguing that IP should have always-on encryption? Not everyone wants or needs encryption, and some may prioritize speed or power instead.

Part of the predatory part is the availability of alternatives. As a pathological case, consider a cable ISP that only allows 10-year contract with early termination fees. If there are no alternatives (other than just having no ISP), then the cable company is predatory.

Another part of being predatory is whether the upfront terms are easily digestible. Part of the California bill addresses this aspect, where terms are either hidden or otherwise presented to be either misleading or not easily understood by normal non-lawyer consumers. Buyer's remorse only applies when the buyer initially understood the original terms. Unless otherwise prohibited or punished by law, it's in most companies' interests to push the consumer into thinking he's getting a better deal that actually exists.

This article was written by an Indian student studying in the US. So, he's just citing an example based on his personal perspective. Aside from English, the one language that would be sort of a natural fit for AI training is Chinese. Up to 17% of the world's population can read/speak Chinese, which is close to the up to 20% that can read/speak English. Plus, a large percentage of AI researchers, companies, and models are located in China.

The article's author looks at Common Crawl, but that may not be representative. I suspect that the knowledge sources used for Chinese model training have a significant non-overlap with models trained in the US. I can believe that much of the world's knowledge still lies untouched by any AI model, but the distribution of that untouched knowledge across languages, technology domains, time of creation, etc. is likely much more complex than a simple discussion of just the language dimension.

"If NVDA has to provide the capital that becomes its revenues in order to maintain growth, the whole ecosystem may be unsustainable"

Lucent did this in the late 90s, covering up their ruse with clever accounting that wasn't so clever when it was discovered. This tanked the company and the stock. However, what Lucent did is more like AMD's recent deal rather than what Nvidia's deal with OpenAI. Neither Lucent nor AMD did quid pro quo deals but rather a large outright gift to the customer just so they could technically record a large sale. Nvidia's deals are also geared towards pushing sales. However, instead of giving away NVDA shares, they are receiving OpenAI shares, more like a quid pro quo.

"We are in a phase of the build-out where the entire industry's got to come together and everybody's going to do super well," OpenAI CEO Sam Altman told the Wall Street Journal on Monday.

This makes no sense. There is no emerging market where "everybody's going to do super well." That's always a lie. There will be a bloodbath, which has been true of every single emerging market ever. Most players today will fail. Only a few will eventually remain, and they will be the only ones that will do "super well."

OpenAI is way behind its main competitors (the hyperscalars) because it loses money and has no cash flow cow. OpenAI is likely one of the ones that will fail unless there is a significant paradigm change.

Perhaps there was disdain for Nvidia SoCs in gaming consoles. However, the feeling was mutual, as the profit margins on these chips are significantly lower than for discrete GPUs. For that profit, significant engineering resources were required. Nvidia's feeling was that it was a competitive advantage to have AMD stretch engineering resources to gaming consoles in addition to CPUs and discrete GPUs, all with fewer total employees than Nvidia, which was focusing almost solely on discrete GPUs, especially after the realizing about ten years ago that its SoC forays into phones and tablets were never going to amount to much.

The study mentions covariates, including income, ethnicity, and maternal education and gives population descriptions for some of these covariates. However, the report then almost completely fails to discuss covariates for the remainder of the paper, including in the discussion. The only statement in the results is that the results were "adjusted" for covariates.

Kudos to the study authors for tabulating covariates and a huge minus for asking the reader to just trust them that they did "the right thing" in considering covariates. Absent the covariate analysis, it's not clear that the results mean anything. If income, ethnicity, and maternal education weren't correlated with test scores, that would be surprising. The question is how strong the screen time and test score correlation is after accounting for the covariates. If the results are after accounting for covariates, then the authors needed to devote a significant portion of the paper to that analysis.

Also, boiling everything down to tidy little numbers usually masks the more interesting and insightful parts of the data. I would rather see scatter plots of the the covariates along with screen time and test scores to see what the distributions are.

Why isn't more favorable tax treatment a reason for moving? Because these people have so much money that they can hire lawyers and accountants to give them favorable tax treatment without moving abroad.

Moving abroad for investment opportunities seems suspect. Seems like it's not hard to invest in foreign countries without moving one's residence. Same with better educational opportunities. People who are already rich don't worry about their resumes like ordinary people, and if they want to learn, they know there are better and faster ways to learn without moving to a new country.

The only plausible reason is the desire for better quality of life.

Ideally carbon production would be avoided on the first place. Secondarily, it would be nice to capture and sequester carbon at the source. However, there are so many sources that it's simply impractical due to price and scaling constraints. Carbon capture from the general environment bypasses the need to identify carbon producers and to engage their cooperation.

Furthermore, it could be philosophically argued that the world is already engaged in mass terraforming with the extreme carbon release of the last few countries and certainly the last few decades.

Perhaps a nitpick, but neither China nor the US cares about free trade as much as they care about favorable trade. Both countries employ various tactics to restrict trade to their advantage. All countries use subsidies. We're seeing countries use strategic export controls and even embargoes. Also dumping, import restrictions, forced technology starting or transfer or theft, industrial espionage and more.

Free trade rhetoric is just marketing/propaganda. All countries are for free trade when it benefits them but cry foul when it's a disadvantage.

Seems like the lamentations are about how people aren't being laid off. Shouldn't the goal be getting the same people to be more efficient and do more work in the same amount of time or at least with less mundane effort? AI that summarizes emails or reports saves time for a worker, but it wouldn't make any sense to let go of someone because of that time savings. Similarly for things like code generation. No, AI won't write a complex piece of software by itself, but it can reduce typing and creating the simple, straightforward parts and allow for time and effort to be concentrated on the more complicated parts.

I had the opposite experience. To me, a workstation was a linux/unix machine. When I finally got a laptop (big and slow Toshiba), it was a Windows machine. Of course, starting about 20 years ago, the workstations went away. I had servers and workstations in the lab, and those all ran linux, but the office desktop went away and was replaced by a Windows laptop with a docking station and a monitor.