Deportation Dramas & Workplace Waste, FEMA's on Notice

Originally published at: https://peakprosperity.com/daily-digest/deportation-dramas-workplace-waste-femas-on-notice/

US Politics

President Trump signed an executive order to review the Federal Emergency Management Agency (FEMA), establishing the Federal Emergency Management Agency Review Council to evaluate the agency and propose changes, including the possibility of its dissolution. This action follows criticism from Trump and other Republicans, who have accused FEMA of withholding disaster relief from Republican-afflicted states. However, some Democrats, including Senator Elizabeth Warren, have criticized the move, arguing that dismantling FEMA could leave millions vulnerable during natural disasters.

Meanwhile, Illinois Governor J.B. Pritzker has openly challenged the Trump administration’s deportation orders, emphasizing that local law enforcement will protect law-abiding undocumented individuals. This stance underscores the ongoing conflict between state and federal authorities over immigration enforcement, with potential legal ramifications for defying federal directives. The Trump administration has warned of legal consequences for state officials who obstruct federal efforts, highlighting the complexities of balancing federal authority with state autonomy. Legal experts note that while states have some autonomy, federal law on immigration remains supreme, potentially setting the stage for a significant legal battle.

In a series of events, President Trump initiated and resolved a trade conflict with Colombia in under 10 hours. The conflict began with Trump’s demands for Colombia to accept the return of illegal aliens from the U.S., which Colombia eventually agreed to, averting proposed tariffs and sanctions. Colombian President Gustavo Petro initially responded by increasing tariffs on U.S. goods, indicating potential future trade tensions. The rapid resolution was met with mixed reactions, with some viewing it as a victory for Trump, while others criticized the approach for its implications on future trade relations.

In a viral video shared on X, a whistleblower from the Army Corps of Engineers has reported inefficiencies and a lax work culture within federal agencies. The former employee described a work environment where job security is abused, with employees neglecting duties and misusing remote work privileges. This revelation raises questions about the effectiveness and accountability of federal operations. In response, the American Federation of Government Employees argued that these issues are not representative of the entire workforce, emphasizing the dedication of many federal employees.

Geopolitics

Concerns have been raised over the influx of Chinese military-aged men entering the U.S. illegally. Brandon Judd has highlighted potential dangers posed by these individuals, who are reportedly smuggled by cartels and forced into labor to repay their debts. The purpose of their presence remains unclear, with suspicions of espionage or other undisclosed activities. However, some reports suggest that many are fleeing economic hardship or political oppression in China, seeking asylum in the U.S.

Meanwhile, China’s DeepSeek AI models, particularly DeepSeek-R1, have emerged as competitors to Western tech giants, challenging Silicon Valley’s dominance. These models, developed despite U.S. sanctions, are notable for their efficiency and open-source nature, making them accessible and adaptable. DeepSeek’s success has prompted major AI labs to study its techniques, potentially reshaping the tech landscape and impacting global power dynamics. While DeepSeek’s advancements are impressive, the AI landscape remains competitive, with ongoing developments from multiple players.

Health

Recent studies have suggested a potential link between IgG4 antibodies, diet, chronic pain, and allergies. Research indicates that certain foods can trigger IgG4 antibodies, contributing to inflammation and health issues. A study involving individuals with chronic pain found that removing specific foods from their diet led to significant pain reduction. Another study on children with allergies showed similar results, suggesting dietary restrictions could be a valuable tool for managing these conditions. Experts caution that while dietary changes might help, the science is still emerging, and individual responses can vary widely.

Energy

Intel has called on the Irish government to subsidize renewable energy to address rising energy costs that are impacting its operations in Ireland, particularly at its Fab 34 facility. The high energy prices in Europe, compared to Asia and the U.S., are driving deindustrialization as companies relocate to regions with more competitive pricing. Intel’s advocacy highlights the broader challenge manufacturers face in maintaining competitiveness amid soaring electricity rates and emphasizes the need for investment in renewable infrastructure to reduce costs and support industrial operations. However, environmental groups have questioned the efficacy of such subsidies, arguing they may not represent the best use of taxpayer money. In response, the Irish government has indicated that any subsidies would need to align with national economic and environmental strategies.

Sources

The Unseen Threat: Concerns Rise Over Influx of Chinese Military-Aged Men in the U.S.

Whether they’re spying on us or laying in wait
 for some other purpose, we just don’t know.

Source | Submitted by Chris Martenson

DeepSeek’s AI Breakthrough: A Small Startup’s Big Leap in Efficiency and Performance

DeepSeek has made profound advancements not just in model quality, but more importantly in model training and inference efficiency.

Source | Submitted by nickythec

Trump Signs Executive Order to Potentially Abolish FEMA Amidst GOP Criticism

FEMA is getting in the way of everything.

Source

Colombia’s 10-Hour Trade War: Trump Triumphs, Tariffs Tabled, and Twitter Tiffs

Trade War Ends In Less Than 10 Hours After Colombia Agrees To All Of Trump’s Terms

Source | Submitted by Tycer

Intel Urges Irish Government to Subsidize Renewables Amid Rising Energy Costs

Intel is advocating for the state to assume part of these costs during the development phase to ease the financial burden on manufacturers like itself, the report says.

Source | Submitted by John Goodwin

IgG4 Antibodies: The Mysterious Link Between Diet, Chronic Pain, and Allergies Unveiled

This is literally, according to them, the very first time, and this is only in 2022, that this type of medical experiment has ever been done where food was removed from the diet for people with pain because of the type of IgG4 antibodies that they were producing to foods.

Source | Submitted by Mavryk

China’s DeepSeek AI: The New Epicenter of Tech Innovation Shifts from Silicon Valley to Hangzhou

DeepSeek is a nuclear bomb detonated in the heart of Silicon Valley.

Source

Illinois Governor Pritzker Challenges Trump on Deportations: “Law-Abiding Undocumented” to be Protected by Local Police

Gov. Pritzker Defies Trump Admin — Declares Local Police Will Defend ‘Law-Abiding Undocumented People’

Source

Government Whistleblower Exposes Alleged Inefficiency and Lax Work Culture in Federal Agencies

Our government is filled with the most incompetent and most lazy people and an occasional hard worker, and those hard workers are severely punished every time they outwork their colleagues because then the colleagues realize people will see that they’re lazy and they don’t want to have to work more.

Source | Submitted by AaronMcKeon

In addition to sources submitted by community members, the following were also used in the creation of this report: The Washington Post, Elizabeth Warren on X, Department of Justice, Gustavo Petro on X, Bloomberg, American Federation of Government Employees, CNN, Fox News, South China Morning Post, Greenpeace Ireland on X, The Guardian, The Volokh Conspiracy, Chicago Tribune, and Office of Personnel Management on X.

Interesting take from zerohedge.com https://x.com/zerohedge/status/1883881677778219034

1 Like

DeepSeek is CCP propaganda all the way. It does not outperform ChatGPT, nor is it anything new on top. And for people who think this somehow will tank nVidia’s shares: Why? DeepSeek is the exact same technology requiring the same hardware to run as all the other LLM’s. Even if it was trained using less hardware – and we have no proof of this – that doesn’t have any bearing on running it.

The amount of uninformed opinions swirling around “AI” is such a pain in the head because I could spend every second of every waking hour running around trying to inform people how these models work and why the hype isn’t real – for both DeepSeek and ChatGPT and all the rest – and it’s a losing battle because trillions of dollars worldwide are invested in making it something it isn’t.

sigh

When this bubble bursts I’m going to hear it from China and the US and the rest of Europe all the way over to my mountain home here in Norway, where I’m running small language models on my own GPU and know exactly what I’m dealing with.

3 Likes

I think the point is that this has sown doubt in the requirement for brand new expensive chips. If some chinese MacGyver can get a functioning AI without the latest and greatest chips and without the power draw then what is the reason for the high price of nVidia?

1 Like

First off it builds on all the work done in the west already, including Facebook’s open Llama models.

Second said Chinese “MacGyver” were already working on another project and had a cluster of 50+ GPU’s they claimed were all they needed to train this. Even if we assume this is true, it still doesn’t match current state-of-the-art from ChatGPT, nor is it a leap forward.

Right now it’s a lot of claims with a lot of incentives – political and financial – for a CCP-backed company to claim this. And we can look at the stock market and media to see how effective that is.

In regards to nVidia, like I said, you still need GPU’s to run these models. The larger they are, the more professional-grade GPU’s you need. DeepSeek’s hardware requirements during runtime are comparable to other LLM’s: GPU Requirements Guide for DeepSeek Models (All Variants)

Do note that nvidia are the only ones still making hardware that can run them. For example, to run the DeepSeek V2 236B model it’s recommended you have 6 NVIDIA H100 80GB GPU’s. These cost ~$30 000 each. For the largest variant you need 12 of those, and that’s the recommended spec. If you want to serve a few thousand users, you’ll need more than that in your datacenter. If you want to serve hundreds of thousands, well
 You see where this goes.

How does it then make sense to tank nvidia’s stocks? :thinking:

4 Likes

Ah, some say once trained, dont need but fraction of compute anymore. How it is using AI compute when serving users with trained model? Or is it as I suspect it has to be in real world, constantly re-learning, “updating” that model based on new data, thus it is constant demand for AI compute based on changing dataset
 I had some online AI course but those rarely answer this question. Best model is useless if I have to buy and install new model tomorrow every time.

BTW, do you have info how chinese get these newest hardware? Or are they smuggling rtx 3000 and 4000 series due to sanctions? They stockpiled as much as they could when window was closing
 was it during 2023
 time flies.

3 Likes

The hardware requirements I linked to are for runtime. That means what is required to run the model interactively, not for the training. The training is significantly more expensive and a whole different thing. But yes these requirements for runtime are a fraction of the training requirements.

I can run the smallest models locally on my RTX 3060ti card. It can do about 3-5 tokens (read: words) / second when responding. So that’d be the 7B parametere Llama model, for example.

3 Likes

Emotion

1 Like

Nvidia is frothy anyway. Hopium.

https://tribe.peakprosperity.com/t/deep-seek-edits-answers-to-align-with-ccp-position/43659/8?u=friedrichs_teeth

2 Likes

They have the industrial-grade professional GPU’s, which do not have iterations that often. The same ones all the datacenters are still running, which would be the H100 Tensore Core GPU’s

The commonly used version of these for AI and professional, high-end 3D offline rendering are the 80 GB RAM versions. They released in March, 2023.

It’s only the commercial-grade gaming GPU’s that have yearly iterations. None of these, even the latest 50xx series, can run the larger parameter language model versions. They simply do not have enough memory. You can get some commercial RTX cards with up to 16gb but that’s not nearly enough to run anything above ~14B parameter training sets at most. A lot of people seem to not know this difference.


An aside, but my personal take on this whole DeepSeek thing is it’s a great financial weapon against the western AI bubble – as is clear in the markets right now. Well played by China, but it neither solves any of AI’s practical problems nor does it “change the landscape”.

People that question everything their own government is doing are being taken for a ride by the CCP. I also see a fair bit of appealing to authority on this topic. I’d hoped that we’d learn that lesson by now from Covid 19.

2 Likes

Can you explain where that is.

I think most people are saying that Nvidia is frothy because the markets are irrational. No one is saying that they think that because authority told them that. For me it is self evident the price per earning on Nvidia is too high, which it really is.

1 Like

The market as a whole, PE ratio is way too high
 I look at it like trading cabbage patch dolls or pet rocks.

2 Likes

DeepSeek is just the tip of the iceberg

12 years ago in a discussion with a prior art searcher that helped large US corporations get patents, he was astonished that most of the prior art from his search results was Chinese. At that time I was getting patents for US and Japanese companies in a variety of fields. I was similarly surprised to find that the biggest impediment to getting a patent was often due to the Chinese having already made many of the inventions that I was trying to get patents on. 12 years ago.

More recently my wife has been working as a patent attorney for Japan’s NIH and I got to know some of the researchers at her research facility. Young ambitious and bright researchers were emigrating to China to continue their careers due to stultifying political impediments in Japan. Another personal experience that I want to mention occurred two years ago in the middle of covid when I was waiting in line at the Shanghai airport to leave China. A young woman next to me in line told me that she was returning to her R&D job in California where she was helping develop a state of the art new type of chip. She explained that all the researchers there were non-Americans, mostly from China, and that no young American engineers less than about 60 years old were available to talk about their work due to their lack of training. Instead, old American engineers from the head office in Boston, who had learned basic electronics back in the day when American schools focused on hardware and physics, were needed to evaluate and manage their work. These experiences fit a pattern wherein only immigrants seem to advance American technology to create companies such as Nividia (from Taiwan) or SpaceX (from S Africa).

America media nonetheless is filled with stories about how China has new technology only because of IP theft from America. After Google spends 20 billion dollars a year developing AI (and similar expenditures by Meta etc.) the techno oligarchs, led by Sam Altman convinced Trump that we need massively more electricity, bigger chips, and 500 billion investment to out compete China in the AI field. Meanwhile, due to export controls, a small startup in China, from a 5.6 million dollar expenditure, and limited to non-embargoed chips outperformed the massive American industry with a superior AI that is 30 times cheaper, runs on 15 times fewer chips, and is given away for free. This demolishes the expensive financial plans (and new feudal systems of top-down control) of silicon valley. More importantly, this free open source software, which can run on a cell phone opens up a world of equality and freedom, in contrast to the silicon valley approach that was based on expensive, yet hidden secret software controlled by a person at the top. Someone remarked that the total development cost of DeepSeek AI is similar to the salary of a president of some of the contemplated American AI startups.

The American universities cannot compete with China because they are focusing on things other than free exchange of ideas and cutting edge research. The real scientists (those who search for truth regardless of politics and their own career advancement) are increasingly pushed out of universities and government laboratories and sometimes emigrate to China. Sabine Hossenfelder often discusses this problem on youtube. I recommend a very accurate youtube video on this topic by an Elon Musk AI fake at https://youtu.be/i-bJRGktOHc . Basically, there has been a brain drain into China for some years and China is far ahead in most technology, in the generation of wealth, and in influence around the world. China has leapfrogged the US.

People in the US are living in a dream state. We all know that the truth shall set you free. But belief in lies will enslave you. Americans are swimming in a sea of mass media lies comprised of often ridiculous, but feel good beliefs contrived to make some rich people richer. This make believe (America is greatest, make America great again etc.) will crash at some time, and China will be blamed when that happens.

I can see how we could in turn leapfrog the Chinese in the future, but that is the subject of another conversation.

5 Likes

The appeal to authority isn’t prevalent here on PP I think, other than perhaps taking the developers behind DeepSeek at their words when there’s no way to prove that it’s true. What I mean by that is that even if the models are open-source, the weights behind them and the pre-training data isn’t. We simply have no way to verify what they actually trained them on. What corpus of data it is, and whether they really spent “only” 2.664 million H800 GPU hours on the pre-training per their official docs, is something we actually cannot verify: GitHub - deepseek-ai/DeepSeek-V3

Everybody is reporting this as truth, but what we know from the output is that often DeepSeek refers to itself as ChatGPT. So there’s some theories that it’s been using ChatGPT as the source of the data. If so, the whole story would be completely different.

All that aside, they’re still using nVidia’s professional-grade H800 GPU’s (a variant of H100 I mentioned earlier, released at the same time) at large scale to service the runtime use of the model. There’s no reason therefore that any of this should tank nVidia’s stocks unless there’s other factors in play – like for example the media reporting verbatim all these claims and not investigating it and traders that would like to snatch up nVidia shares abusing this current news cycle to drive down the price before snatching up shares at a lower cost before it inevitable goes up again.

Some raw data to chew on: A professional-grade H800 GPU costs ~$30 000. If the training took 2.664 million hours, and say you wanted to finish training in 3 months, that’d be 2160 hours. You’d need 1233 H800 GPU’s to do the purported pre-training of DeepSeek. Running at full-capacity these cards drain 500 watt/hours. That’s $37 000 000 worth of GPU’s in up-front hardware cost draining 616 500 watt/hours. You can of course halve that by doubling the training time (spending 6 months) etc., but let’s not pretend this isn’t datacenter-scale operations.

To answer your question specifically: I have no idea what the “true” value of nVidia’s stock should be. I don’t own their stocks either, to be clear. But what I know is that in contrast to other company evaluations of dubious nature – ChatGPT’s evaluations, for instance – nVidia actually produces real physical hardware that is in tremendous demand not just for AI but also for worldwide film production and professional 3D rendering (think most of Hollywood’s CG), not to mention being leaders in commercial-grade gaming GPU’s as well. Their stocks at least reflect that they have a deep moat technologically that their only major competitor, AMD, have yet to breach.

2 Likes

Mots, it’s good to read of your thoughts and experiences.

neofeudalism remedievalization

2 Likes