A-Infos (en) Italy, UCADI #211 - AI, or the fall of the Gods (ca, de, it, pt, tr)[machine translation]

A hurricane seems about to hit the nests of the black angels (to use the 
ravings of Peter Thiel) and upset them, plunging them into the misery 
into which they will overwhelm all those who have invested in their 
follies. --- Artificial intelligence --- Artificial Intelligence (AI) 
isn't intelligent: it's based on linear inferences and owes its appeal 
to the speed and number of chips that, working in parallel, allow for a 
huge number of calculations and comparisons to be performed in a timely 
manner. AI isn't even artificial, since it relies on logical operations 
very similar to those performed by the human brain. Again, the 
difference lies in the number of operations that can be performed in a 
short time. AI is also different from the programming we're used to: in 
a certain sense, it relies on analogies that don't guarantee the 
correctness of conclusions (in this sense, it's very similar to natural 
human reasoning ). However, we're seeing huge and growing investments, 
especially in the US, in the belief that it will have a huge impact on 
many areas of human activity.
Among others: Microsoft, pioneer and main funder of OpenAI, which has 
integrated AI into its entire range of cloud products and services; 
Alphabet (Google), which through Google DeepMind and the Gemini family 
of models, is a historic and current pillar of AI research and 
application; Amazon Web Services (AWS), which provides large-scale cloud 
infrastructure and dedicated services for the development and deployment 
of enterprise AI; and Meta, which develops and releases the open-source 
Llama model suite , integrating artificial intelligence into its social 
media platforms.
It's worth remembering that these are parasitic companies that feed 
their "creatures" by plundering billions of data produced by 
universities and other public institutions, including researchers, 
writers, journalists, and others, without paying any royalties, in 
defiance of intellectual property protections.
Added to these are the companies that produce the hardware needed to run 
the various codes, first and foremost Nvidia, the world leader in GPU 
production, which relies on the Taiwanese TSMC (Taiwan Semiconductor 
Manufacturing Company), and the Dutch ASML, still the only company 
capable of producing nanoscopic chips that allow for great execution 
speed thanks to the development of lithographic techniques in the 
extreme ultraviolet: to give an idea of the complexity required, the 
machine developed by ASML involves almost half a million components and 
requires the use of elements produced by other German and American 
high-tech companies.
So far, the stock market growth of companies involved in AI development 
has been the result of a Ponzi scheme (chain letter) in which revenue 
comes primarily from the sale of products needed to build new data 
centers from one company to another, rather than from the final use of 
the software by external companies .
The only example of a resounding success of AI techniques comes from the 
prediction of the spatial shape of certain polymers, which has enabled 
the development of new drugs and the treatment of chronic diseases: Eli 
Lilly has benefited in particular.
Outside of this case, there are no reports of significant improvements 
in the efficiency of large companies (those that would guarantee greater 
profits). Currently, AI appears to be having a significant impact only 
on micro-companies, startups, where individuals are able to do the work 
that previously required a pool of workers.

China

Adding to the uncertainty surrounding the actual success of AI is the 
presence of a formidable competitor for the United States: China. For 
starters, the success of AI is almost a matter of life and death for the 
US: currently, AI is responsible for 40% of US GDP growth and 
contributes to 80% of stock market capitalization. A potential collapse 
would have unimaginable consequences for the American (and global) 
economy. The same cannot be said for China, which hasn't invested 
comparable amounts of money in this sector.
Specifically, China's Achilles heel has been the performance of 
individual chips. To maintain their advantage, Trump and then Biden had 
already imposed bans on the sale of the most advanced chips. But as 
often happens, the resulting shortage pushed China to launch a sort of 
Manhattan Project (formerly developed by the US to develop nuclear 
weapons) to develop artificial intelligence, the Next Generation AI 
Development Plan. This attracted investment in research and workforce 
training, intensifying cooperation between the government and the 
technology industry with significant public investments in data centers, 
power transmission, and semiconductor production.
Through a combination of industrial espionage, reverse engineering, and 
the development of innovative techniques, the Chinese are now able to 
produce far-ultraviolet beams at the same frequencies as the ASML. The 
problem is the lack of reliability that impedes mass production (too 
much waste).
Although the gap with the US has not yet been closed, the quality of 
Chinese semiconductors has improved to the point that China itself has 
banned the purchase of the penultimate-generation American chips, the 
H20s, which Trump had authorized. This has put Nvidia in a difficult 
position, as it has had to give up a significant portion of its market 
share. It should also be noted that Beijing controls the entire supply 
and production chain and is therefore not subject to blackmail.
The same cannot be said of the United States, given that semiconductor 
production requires highly refined rare earths. Last year, in 2025, the 
US showed off, imposing unreasonable tariffs on products from China. As 
a result, the Chinese blocked the sale of rare earths, forcing the US 
into an embarrassing backtracking, so significant that the tariffs on 
Chinese products fell below their value when the orange clown took 
office. Months have passed, and the Americans have begun reintroducing 
various bans on Chinese products, resulting in Beijing once again 
restricting exports of rare earths used in dual-purpose products (both 
civilian and military). The division is highly subjective, and it's 
reasonable to assume the Chinese could tighten the reins as needed. 
Speaking of Achilles' heels, this one seems much more critical than the 
Chinese one.
However, the speed of a single chip isn't everything, even within the 
context of hardware. Data processing systems work in parallel with chips 
that operate simultaneously, but which must exchange information. 
Traffic management is an equally significant issue. Virtually the entire 
world uses CUDA (Computer Unified Device Architecture), a platform 
developed by Nvidia and typically implemented in graphics cards. At the 
same time, the Chinese have developed CANN (Computer Architecture for 
Neural Networks), which, in addition to being independent of the 
American company, appears to be able to optimize communication traffic 
between the various individual chips by exploiting a three-dimensional 
structure. This result was also achieved thanks to a national effort 
(the so-called Manhattan Project) that coordinated giants such as 
Alibaba, Tencent, and Huawei.
Along the way, we arrive at the software. The Chinese have chosen an 
open source or open source approach , which can be downloaded, modified, 
and used freely, unlike the proprietary, encrypted software typically 
developed in the US. This American choice, in the past (and currently), 
has forced users to buy blind and purchase updates. This is the choice 
that allowed Microsoft to become what it is today, imposing a 
first-night tax on PC purchases. But this choice also worked thanks to 
its European vassals, who, despite having developed competitive open 
source software in the 1990s (UNIX, which later gave birth to Linux), 
preferred to pay protection money and give priority to Microsoft 
software in public administrations.
But China is not the EU, and has not only developed open source models, 
but is working to distribute them to the rest of the world, because the 
potential commercial success of AI depends on the number of users, and 
clearly, the ability to tweak the code as needed represents a strong 
incentive for development.
Another factor to consider is the time required to train the AI for a 
specific task. AI isn't born learned; it needs to be trained, and 
training requires time and, above all, energy (electricity). To give an 
idea of the expected consumption, the American Stargate program, which 
involves the largest investment funds (Black Rock, Vanguard, State 
Street), predicts the need to install 10 Gigawatts within 5 years to 
operate the various data centers: this is more power than what is 
currently installed in Japan.
Therefore, energy is expensive (a small reminder for the inept European 
governments who are doing everything they can to commit energy suicide), 
and it is therefore essential to save on training costs: even if the 
software is free, any developer must cover the costs of developing the 
code itself.
Here, another factor seems to emerge in favor of Chinese models, 
particularly Deep Seek, which appears to require much less time 
(presumably related to the logical structure of the codes).
Added to this is Qwen (Alibaba): one of the most popular and downloaded 
language models in the world on platforms like Hugging Face, 
constituting a very valid alternative to US products. We can also add 
GLM (Zhipu AI), which provides advanced models (e.g., GLM-5.2) with 
permissive licenses (often MIT), offering excellent coding capabilities 
and support for very general contexts.
Remaining in the context of software, the potential of a given AI 
platform requires skilled developers to be realized, and here another 
sore point for the US emerges. China produces 5.357 million STEM 
(scientific) graduates each year, compared to 820,000 Americans, many of 
whom later leave the country or are forced to do so by Trump's policies. 
The gap isn't just virtual: Tsinghua University produces more computer 
science patents each year than MIT, Stanford, Princeton, and Harvard 
combined.
A final distinctive feature of the Chinese approach concerns the type of 
objectives. Unlike US companies, which are mobilizing resources to 
develop a futuristic yet "abstract" type of artificial intelligence 
(General Artificial Intelligence) that mimics the workings of the human 
mind, Chinese companies focus on pragmatism and direct it toward the 
execution of specific functions (Narrow Artificial Intelligence), 
especially in the fields of surveillance, public administration, 
education, and the high-tech industry. This is demonstrated by the 
gigantic complex built by Xiaomi, which, by integrating artificial 
intelligence into fully robotic production processes, produces an 
electric car every 76 seconds. This policy has made the Chinese 
artificial intelligence sector far more accessible. However, the 
experimental application of AI in some schools near Shanghai among 
primary school students (ages 6-10) is disturbing. Helmets with 
electrodes are being fitted to students to monitor their attention span 
across different subjects to guide their future academic careers.
With its strategy, China has been able to claim not only that it has 
closed the gap with US industry, but also that it is growing faster in 
both AI and chips because the model it has adopted allows for greater 
pragmatic and open innovation, circumventing US tariff policies that are 
stifling the market. As a further example, Kimi (Moonshot AI), a company 
with just 200 employees and an investment of just $6 million, has proven 
capable of managing massive projects and delivering high-level 
performance. It is establishing itself as the most competitive AI 
company, compared to a direct competitor with 5,000 employees that has 
invested a whopping $100 million to produce a product with equivalent 
performance.
Looking ahead, China aims to completely exclude the United States from 
its supply chains, including not only hardware but also US and European 
software of all types, which will need to be completely replaced with 
Chinese counterparts, particularly in the critical sectors of finance, 
energy, and public administration.

The bursting of the AI bubble

According to some reputable analysts who predicted the bursting of the 
2008 bubble, a new bubble is brewing, involving AI. Its consequences 
will be devastating due to the immense amounts of capital invested, 
whose returns are destined to collapse. The dizzying growth of the AI 
bubble finds a disturbing parallel in this context. The trending 
increase in interest rates on corporate bonds issued by companies 
integrated into the AI ecosystem, combined with the rise in credit 
default swaps on the major companies operating in the sector, signals 
growing nervousness among institutional investors. Consider the stock 
market maneuvers of the Thiel Macro fund (led by Peter Thiel, co-founder 
of PayPal and Palantir), which has liquidated shares of Nvidia and 
drastically reduced its exposure to Tesla. Alex Karp and Stephen Andrew 
Cohen of Palantir, meanwhile, have sold tens of millions of dollars 
worth of shares in their own company. Similar operations have been 
carried out by Fidelity, T-Rowe Price, JPMorgan Chase, SoftBank, and 
Stanley Druckenmiller, a leading speculator who became famous for his 
1992 attack on the pound while working with George Soros and Michael 
Burry. For them, the staggering profits made by companies operating in 
the sector are based on "one of the most common frauds of the modern 
era," namely the artificial manipulation of payback periods. Companies 
spread hardware investments over long periods, given the average "life" 
of chips of about two years: "To justify the current price multiples, 
now suspended in the void in some cases, revenues ten times higher than 
those generated today by artificial intelligence would be needed, and 
they would be needed very quickly, because the payback cycle for data 
centers and chips is extremely short: perhaps five years."
Among the causes of the potential crisis, they point to geopolitical and 
economic uncertainty and excess liquidity that has inflated the value of 
assets far beyond their fundamentals. The combination of artificial 
intelligence, high-frequency trading, and investment algorithms that 
dramatically amplify market fluctuations is creating an overabundance of 
liquidity, leading markets onto a very dangerous trajectory. What is 
happening now will lead to a further phase of verticalization among the 
players operating and managing the market. No one knows when the bubble 
will burst, but it will eventually happen, and then a run on banks could 
be triggered.
The implications would be particularly severe for consumption, where 
growth is already weaker than before the dot-com crash. A shock of this 
magnitude could reduce it by 3.5 percentage points, translating into a 
2-point decline in overall GDP growth, even without considering the 
decline in investment. The global repercussions would be equally severe.
Foreign investors would also be affected, potentially suffering losses 
exceeding $15 trillion, equivalent to approximately 20% of global GDP, 
penalizing holders of US securities, including European and Italian 
institutional investors. By comparison, the dot-com crash caused losses 
for foreign investors of approximately $2 trillion, approximately $4 
trillion in current currency, and less than 10% of global GDP at the 
time of the 2008 crisis. This sharp increase in spillovers underscores 
how vulnerable global demand is to shocks originating from the United 
States.
The fact remains that the AI bubble is 17 times larger than the New 
Economy bubble and four times larger than the subprime mortgage bubble 
that triggered the 2008 crash.
The size and complexity of the AI bubble could collapse the global 
economy for a decade, so a "public parachute" should be provided for 
OpenAI's colossal investments in chips, data centers, and related 
infrastructure to mitigate the financial risks arising from rapid 
technological change. The fact is that OpenAI, like Nvidia and other 
high-tech giants, have become too big to fail, like the Wall Street 
banks bailed out in 2008. In other words, the entire US stock market is 
at risk of crisis, and with it, much of the economic system.
In other words, the ongoing war between the US and China could be won 
without a fight, in keeping with the best teachings of Sun Tzu, who 
stated that the best victory is the one achieved without a fight.

Antonio Politi

https://www.ucadi.org/2026/09/06/ia-ovvero-la-caduta-degli-dei/
_________________________________________
A - I N F O S  N E W S  S E R V I C E
By, For, and About Anarchists
Send news reports to A-infos-en mailing list
A-infos-en@ainfos.ca
Subscribe/Unsubscribe https://ainfos.ca/mailman/listinfo/a-infos-en
Archive: http://ainfos.ca/en