I'm not sure why I'm responding to you other than i can't barely stand how misinformed you are.

Yes bubbles happen in US stock markets. The MBS situation in the 2007-2008, the Enron and Lucent situation in the late 90's, the SPAC system of the last several years, and now AI.

But you're seeing the trees and missing the forest. In the US, stocks are required to report their financials according to specific rules and are audited, all of which is regulated by the SEC, have many large institutional investors who are more analytical and less prone to emotional buying, and most importantly unlike a single-party (and now nearly single-member) system like China has, we have a bureaucracy with a transparent form of the rule of law. While the process for making laws in the US can be messy, the open debates about those laws and regulation changes means the market is informed of when the rules are going to change and has time to adapt. Transparently, consistency, and trust are key to predictability, which allows investments to be made.

China has none of that. The bureaucracy can change on a whim. if a company is too successful, they can "disappear" the CEO for months until he returns and swears fealty to the party. Taxes can change in a moment without any warning. Most of the investors are retail or individual investors, prone to panic and flights, which is why institutional investors are far more rare in China; it's too unpredictable and risky.

Even now, when the SEC is gutted by the Trump administration and the AI hype is in full swing, the markets are reacting rationally. Most of the public companies today in AI besides NVIDA are these neo-clouds, and they're getting hammered by the markets. Anthropic just filed their S1, and while some are buying the hype others are pointing to the fact that the company per it's own disclosures is in deep financial trouble and is majorly risky. It'll go public, but not for $2T like they want, and there will likely be major caveats on them. Even when the regulators are asleep at the wheel as they are now, there are too many players who have a stake in the game to keep it consistent. ANd yes we're in an AI bubble, but the bubble appears to be in the private markets mostly, not the public markets. And when it crashes, yes the US markets will come down, possibly as bad as 2008. But again, there are too many players in the game in the US; after 2008 the market bounced back as people focused on companies with good fundamentals. There are plenty of companies on the US stock markets that are doing very well; Berkshire Hathaway, Waste Management, Costco, Coca-Cola; they're doing quite well and even in a crash will continue to do quite well. Energy is going well, FInancial Services is booming, and Industrial stocks are all doing great, primarily with the Defense subsector as the military goes to retool.

All of those companies are doing quite well, following the rules, are predictable investments and good growth prospects. When teh AI bubble bursts, there are still great sectors who follow the rules and will cushion the market, and capital will flood back in, rebound the market as it wipes out the AI nonsense, adn things go on.

You just can't say that about Chinese markets, because there is no trust and they're highly unpredictable; the US markets are vastly different.

No one, not even people in China, trust the Chinese stock markets. The companies aren't required to be as transparent, and most of their financing is more likely to come from a State-Owned bank and debt, not the stock market. Because of that, this market tends to be a bit more like a casino than how we in the Western world might view the NYSE or the NASDAQ or the London Stock Exchange. That also means that it's highly prone to overspeculation. As someone posted above, the NASDAQ on average is priced at around 30X earnings; by comparison the STAR50 index was trading at 250X the average earnings of its' companies. Many of the companies were domestic hardware suppliers, but as more and more barriers are starting to rise against Chinese hardware going into AI Infrastructure, investors spooked, got out and the index fell.

For perspective:

The Shanghai Composite Index dropped 72.8% during the 2007-2008 global financial crisis. For comparison US stock markets fell around 53%, and began recovering quickly.

The Shanghai Composite again crashed 49% in 2016, when State media was encouraing people to invest in their stock market to try and grow it to become competitive to Western stock markets. Novice retail investors jumped in, and through that encouragement heavily invested with margin loans. That spooked the government, who started putting regulations on those loans; everyone panicked and the market crashed within a few weeks as everyone tried to get out to cover their margin loans.

Afterwards, the government put in these circuit breaker regulations, to stop trading if the market dropped more than a certain percentage to mitigate those steep drops, again in 2016. The first day those regulations went into place, everyone panicked, triggered a big sell off, and trading halted within minutes. Investors saw the market freeze, panicked even more, and the next day everyone rushed to get their money out, erasing roughly US $1 trillion in value and forcing Beijing to remove the circuit breaker regulations.

This market is very prone to hype, panic attacks, reactionary regulations, over confidence and underperforming. So you have to take these kinds of stories from this market with a bit of skepticism; this just appears to be a feature of the Chinese stock markets.

I see where you're coming from, but given that I fundamentally disagree with you tells me that today we're all just simply speculating.

Again, this is where I come from, but I think that a cheaper python coding assistant trained on python code sounds good, but is it polluting the statistics if it's also trained on Shakespeare, children's stories, poorly written peer-reviewed science papers and reddit posts? Language is used in certain ways for certain purposes; in my opinion you could skew the statistics by incorporating every use case of language into every general purpose use case.

Anyways, I guess we'll find out eventually which conclusion is correct.

That's some cyberpunk style concept there, but I would argue the answer is never. The reason is simple: that wouldn't do anything. These hijackings are not like a Fast and Furious movie, done at high speed on the highway. It's GPS spoofing, or inserting false drivers with fake IDs, who send the truck to some wrong destination, the goods are stolen and the truck dumped or sent on it's way. What is a drone going to do? There's no world where a drone-based law enforcement is going to be allowed, so they're not going to shoot the hijackers like it's some Robocop style situation.

To be perfectly frank, this is priced in. GPS trackers, RFID tags in the shipments, and insurance coverage covers it sufficiently. But I don't think there's a world where the cost of adding some sort of drone to stop some sort of hijacking ends up being justified by the overall financial loss or gain, particularly when modern hijackers are already good enough to spoof GPS systems or use AI to create false licenses. They could hack the drone, or bring their own to defeat it.

So as cool (dystopian) as that sounds, there's really no reason for it so I doubt it would happen.

I think you're coming at the critique of self-hosted models from the wrong direction. You're talking from the bottom-up approach, not the top-down approach.

if you start with end applications, SLM's start making more sense. if i want an app that can make decent legal documents, I want it trained on legal documents to write good legal documents. Conversely if I want an app that can write children's stories, I want it trained on children's stories. Both of those would be Language models scoped to an appropriate size for the end-use, but what isn't useful is a general purpose AI that can "do everything", because we've already seen that none of the frontier models can do everything well, they can simply do everything moderately well to poorly.

Each of those applications would be small enough to run locally, not needing a big data center, and would likely be better at those applications because they're trained on what they need specifically, not generally. Meanwhile the generalist frontier models have yet to prove a use case where they can deliver meaningful value, but they absolutely require the high-cost data centers to be baked in to the price.

As a result, self-hosted becomes very straightforward and easy. If the entire SLM can deliver what I need from my phone or desktop, then the cost of that particular application comes way down as you don't have the data center costs baked into the pricing, plus you've got far more opportunities where your data and queries can remain yours if hosted. Regardless of the privacy issues, the fact that in the recent Subnautica 2 lawsuit showing that AI queries are not private and subject to subpoena, a locally hosted SLM would be far more likely to skirt that issue.

So I see just from a practical application perspective as well as all the privacy and legal aspects surrounding this that SLMs start making real sense really quick.

I worked in construction a few years (decades actually) ago. When copper surged in price, which was mainly due to China joining the WTO and starting it's massive infrastructure build out, several materials shot up in price. Copper in particular was notorious because there were plenty of recycling centers who would pay cash and asked no questions as it was unregulated at the time; it was easy to convert to cash quickly. We had guys get caught with 50lbs of copper wire under their shirts, because the paid price for copper went from $1.00/lb to around $4.00/lb; for a guy doing general labor 50lbs of stolen copper was another $200 cash he made that day. The petty theft guys were a nuisance, but the industrial scale theft was enormous.

Copper is now $6.80-ish/lb. A spool of copper wire headed to a data center might be worth $8,000-$10,000 at a scrap yard.

Some research is showing that the number of incidents is relatively flat, but the value of incidents has nearly doubled because data center infrastructure is far more valuable than retail goods which were historically hit. This report shows that truck theft incidents are down around 26%, but overall losses are double. So it's technically a valid story on Slashdot, as it is the overall value of the goods in the truck headed to data centers that has shifted the pattern just because of it's far higher value and relatively easy way to turn the goods into money.

Everyone else is calling you out for your misinformation, and they're right to do so. I'll add more.

mRNA vaccine research was funded for 9 years before COVID broke out, by DARPA. It turned out it was highly effective at stopping aggressive immune responses, which is what actually kills people with infectious diseases and pandemics; our body's response to the disease.

Moderna had already done some generic testing with both animals and hundreds of human studies for years before COVID; we had nearly a decade of non-stop research that included human trials before Covid even happened. Framing it as untested and not passing safety trials is simply FUD.

From the article:

"The benefit appeared even larger after the six-month vaccine regimen, with a separate analysis showing a 58% reduction in risk of recurrence or death. The companies say it is still too early to determine whether the treatment helps patients live longer overall."

It's far more promising than people expected. What happened was the analysts, who while they know pharma companies still barely understand the science, thought it would be higher and were estimating that so the stock pumped up. when it came in lower than expected, the market reacted.

But who cares? Based on that quote alone, the market is reacting stupidly. Buy the dip; if you have some free cash by Moderna stock. It'll come back up soon, as just as you said this is a great drug.

I'm sure that feels good to say, and you probably feel like you're having a mic drop moment, but you're entirely wrong proven by the long march of history.

Tibet pursued a policy of peaceful isolationism influenced by Buddhist pacifism. In 1950, they got curb stomped by the PLA, conquered and the Dalai Lama sent into exile.

In 416 BC, the Island/City state of Melos chose neutrality during the Peloponnesian War. Athens (the Democracy I should add, not the authoritarian Spartans), demanded they submit and become a vassal state to the growing Athenian sphere of influence. THey refused, Athens invaded, killed every man, enslaved all the women and children, and repopulated the city with Athenian citizens. This was famously summed up in the Melian Dialogue by Thucydides, the strong do what they can, and the weak suffer what they must.

Denmark signaled to Nazi Germany they intended to stay neutral and built a small army, to show they were no threat, and signed a non-aggression pact with Nazi Germany in 1939. In 1940, Germany broke the pact, invaded, conquered the country in 6 hours.

There are many others all throughout history. The only examples of entities surviving without a strong military are countries like Costa Rica and Mexico, which they could do because they fell inside the security umbrella of the United States, who was adamant that the Western Hemisphere would be peaceful, or else. Other small countries like Lichtenstein or Iceland did it through alliances, like Iceland joining NATO with a special caveat that they would build no military, but all NATO could use it's ports and airfields as bases.

Greater men than us today, who experienced enormous hardships, understood this. Si vis pacem, para bellum has been proven true time and again in history.

Fun fact: it used to be called the Department of War, so technically you could be referring to George Washington. It was renamed the Department of Defense in 1947.

Regardless, your point is entirely valid.

Clearly you missed the part where i call out the current administration.

But honest truth: it isn't Trump. Trump is just the symptom of a broader movement; he wouldn't stand a snowball's chance in hell of getting elected if there wasn't a broader movement happening. The reality is that the Baby Boomer and Gen X'rs who were in charge during the 80s, 90s, and early 2000's, whether they were Neo-Con Republicans or the newer Democrats, were globalists. They built a system that built wealth, security and prosperity for them, but missed two key aspects: they locked the door behind them for the next generation to participate in that wealth, and they used the middle class's wealth to fund that security. As a result, the middle class and upcoming generations got fed up, lost faith in both parties, and elected an outsider. About 1/3rd of my relatives are Trump supporters; they all despise him personally, but their vote for him was a vote against all others; frankly many of them would have elected a Golden Retriever as long as he wasn't part of the insider political class, but Trump was what they got. The globalists of the world built a globalized world, but didn't manage their base well and the base revolted when they perceived they weren't being taken care of, and as a result we're seeing a crisis of confidence in government leading to the election of "anything but them".

I say that because Trump is not a US-only phenomenon. Nationalism is growing all around the world right now. France is having major riots that started with youth and spun way out of control, and following the polls, Marie le Pen is the most likely next Prime Minister of France. She is a French female Trump, and has said Day 1 if she gets the top job she will stop all funding and support to Ukraine and drop sanctions on Russia. Germany is seeing another revival of nationalism, so is Italy, the UK, Hungary, Brazil, India, Turkey; all are reviving nationalism. Heck even China is using nationalism to shore up domestic support of the government, and you're even seeing growing nationalism in Canada of all places. This is a global phenomenon; Trump is just the US' version of it.

The US military is still strong. Nationalist tendencies grew in the Philippines about a decade ago, leading to the removal of US troop presence; now they're asking for it back because of Chinese belligerence. The US military is still the most capable fighting force on the planet and in history, but it's strength is being tested by autonomous systems which allow countries to dominate their own local region way above their ability previously and for quite cheap. To maintain it's role as the strong stabilizer, the US Military does need to retool, and quickly (and it is, I have many old and current contacts in the DoD and there are some fascinating projects going on) to counter this rapid growth in technology, which goes back to my point: it's foolish of these people to quit Google in protest over supporting the US military right when the US military needs new tools to project strength and actually promote peace.

I'll reply and give you extra points; you're absolutely right.

This is something I see a lot; people who are intelligent but not really aware of how the world works, but also completely unaware of their own lack of understanding.

The facts of the world are:

1) The US military's enormous strength has kept the world relatively peaceful for 70 years. Yes there have been wars and other conflicts; you can't solve every problem. But compared to the World Wars, the Napoleonic Wars, the Franco Prussian Wars, the Seven Years' War (which some define as the first real World War), and the Thirty Years' War, the back half of the 20th century by comparison was relatively quiet. That's because the US gave the world a choice: trade with us and prosper, or challenge us and be destroyed. You can do that with a powerful military, and occasionally it has to be used, but we haven't seen a conflict even remotely on the scale of the major World Wars until Ukraine. There are issues and problems, but the US' military strength made going Total War again far too costly, when it was just cheaper and better all around to trade. So as much as they are taking a stand on morals, the reality is that the very strength of the US military historically has suppressed conflict.

2) That doesn't mean people won't try to push, again as we've seen in Ukraine. Almost all forms of conflict became asymmetrical; deadly, and it makes headlines, but for the most part asymmetrical warfare could not existentially threaten a nation. But, again as we've seen in Ukraine and now Iran, there are limits to the US military's strength, which built a system of high efficient, high cost systems for global power projection, but are significantly challenged by middling efficient, ultra cheap systems in large numbers. And while you may not want the US military to have AI tools, all you're doing is hobbling the one military who's very power ensured an era of relative peace and prosperity will be unprepared when potential adversaries are empowered with those tools. And quitting Google will not stop Iran or China or Russia or anyone else from developing AI weapons.

Given that the US military's very strength kept the world relatively peaceful when looking at the long view of human history, the moral argument against supporting the DoD falls apart pretty quickly. I could understand not wanting to support the current administration which is absolutely a warmonger, but that changes.

It's Microsoft, and it's an AI world and MS at a corporate level is heavily invested in AI. I'm curious when it's announced how much of this new product will incorporate Copilot in some fashion (and I'm not advocating for it, just expecting it).

I didn't know that, and probably never would because I wasn't planning on watching it. I actually spent time in this industry around when Theranos was first publicly known and then during the fall from grace. Theranos didn't kill diagnostics as a startup field on it's own; it was already headed that way. But the company certainly made it hard for people doing real science that worked; the entire story evaporated. Even the trailers bring back memories of a hard time I'd rather not revisit, so didn't realize they brought in Seyfried too. Makes sense though, I imagine it's very weird to have someone portray you while you're still alive particularly when that portrayal is about your failings, so that's a whole other dimension to this story.

Yep, that unblinking stare, the large size. She's like a predator trying to calm it's prey before it kills.