46 Comments
founding

Matt, thanks, again, for your work, another monopoly history lesson, and for keeping us up-to-date on things monopoly. Special thanks for pushing back against the Khan Derangement Syndrome. I was offended by Ms. Wilson's piece in the WSJ and said as much in the Comments section, but your take on it is way more powerful indeed. Folks in my world are simply ignorant of the monopoly power in their lives, which is why it is so important that stories of every-day Americans, under the boot heal of monopolies, need to be told and shared. Information is power, but obviously not enough, but it is a great start.

Expand full comment
Feb 18, 2023Liked by Matt Stoller

“I’ve been doing research into OpenAI and search…”

Yesterday Matteo Wong of The Atlantic wrote an article titled “AI Search is a Disaster”. However, the real gem was his reference to Ted Chiang’s analogy of a flawed Xerox copy from an article Ted Chiang wrote for the NewYorker regarding large language models and lossy algorithms.

https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web

For those unfamiliar with Ted Chiang, he obtained a degree in computer science at Brown University and once worked at Microsoft. He has written articles about the dangers and conceit inherent in big tech and its relentless pursuit of domination/monopolization. E.g. https://www.buzzfeednews.com/article/tedchiang/the-real-danger-to-civilization-isnt-ai-its-runaway

He still works as a freelance technical writer, writing documentation for programmers, but may be better known to the general public for a Nebula Award winning sci-fi short story he wrote that was later made into the movie Arrival.

I hope you find this interesting in your research, Matt. I continue to enjoy and appreciate your work.

As for Ms. Khan, she’s proving extremely effective and it’s apparent. That’s why Wilson et al. are going nuts. Ah, schadenfreude…..

Expand full comment
Feb 18, 2023Liked by Matt Stoller

Smart piece thanks for writing it. I’m struck by the similarities with how the CDC reacted to dissent within the immunological profession about covid policies. In both instances with complete hysteria towards those with opposing views and attempts to silence the dissenters. Wilson wants Khan removed for “lawlessness,” Fauci & Collins organized public “takedowns” of kulldorff and bhattacharya and influenced the White House to work with Twitter to blacklist them. These insular technocratic communities capture regulatory agencies and policy making, hold on for dear life and react viciously to anyone who deigns to challenge their paradigm. Not surprisingly it’s very much like academia but with far higher stakes.

Expand full comment
Feb 18, 2023Liked by Matt Stoller

When the Soviet Union collapsed economically these ghouls came out to pretend long standing laws and regulations against corporate monopolies were not in fact a traditional part of America's capitalist system, but instead were "communist".

Expand full comment

Thank you Matt. Great work

I can't shake the "feeling", that neoliberal/libertarian/absolute free market economists, like hayek, buchanan, friedman, et al, were or are, less driven by actual beliefs, and more, opportunistically attuned to the idea of justifying wealth's dominance. what a great way to raise one's own dire economic condition, than to sell to generational wealth, stores and theories, that substantiate, their obnoxious, self interested delusions about dominion, their inherent right, to claim the commons, defend their claims of the commons, and control more and more. why they obviously were anointed by god to rule us all, simply by being born into perverse generational wealth.

of course, once one spends enough time selling and publishing their thereois, it would be predictable that eventually, they drank their own koolaid, and became zealots for the cause.

or maybe they were just that blind to all of recorded history, maybe.

Expand full comment
Feb 18, 2023Liked by Matt Stoller

I find it unlikely GOP voters will be energized by the Wall Street conservative faction straight up lying to them.

The anti-Big Tech push is too powerful and widespread to kill that easily, and the average person understands it fairly intuitively.

Expand full comment

Thanks for your writing and thoughtful analysis, it reminds me of why I pushed the buy button, and why I will keep this renewing.

Expand full comment

The Chicago school types would point out that anti-trust law assumes the consumer is being harmed by the monopolist, yet even Standard Oil kept retail prices very low. This was the justification for allowing monopoly in the 1980s and 1990s, a time of great deflation thanks to the rise of the electronic office and Chinese manufacturing. By ignoring the market factors driving deflation the myth of the benevolent monopoly was allowed to take root.

The other problem is that no one paid any attention to copyright or patent law, despite the fact that the major growth engine of that time period was software and technology. By permitting tech firms to build defensive portofolios, and extend copyright to obscene lengths, the stage was set to quash any upstart competitors. Now tech firms have a perfect little Cold War where none can afford to take on the other while small players are locked out of the game completely unless they set up their company for acquisition.

Expand full comment

Matt, I enjoy reading your work. I'm also a big supporter of libertarian ideals, Hayek, Friedman. In my estimation Ms Kahn is correcting 30-40 years of consolidation. Our current system is a reflection of socialism, bureaucracy, oligarchy. It is certainly not what libertarians believe in. No state always leads to anarchy a limited state that doesn't interfere in the lives of it's citizens is ideal. Libertarian ideas are completely incompatible with corporate wealth influencing governance. That would be a obvious violation of logic and the rights of citizens. The example of corporate, government surveillance of it's citizens clearly illustrates this point. 🙏

Expand full comment

As someone who works tangential to the AI/data space, my view is the current hype of LLM (large language models) or diffusion models (like art or music generation) is at best delusional Wall Street trying to spur any tech rally they can or at worst a violation of multiple software licenses, copyright law and theft of other peoples work.

We already see Github Copilot being sued for scraping code others have written and Stable Diffusion (image generator) being sued by Getty for violation of copyright. The application of LLMs as search engines means scraping the entire web of its content and regurgitating with implications on how revenue is being generated, shared, or stolen from the actual writers. To see how this plays out somewhat, Google ran into this issue with Wikipedia, where Google search would scrape Wikipedia articles, and churn out answers in a dialogue box instead of having the user visit Wikipedia. Wikipedia caught on and told Google to start paying it and Google did.

In terms of concentration, the case is clear. To create such models, companies need an astrononmical amount of data and an astronimcal amount of compute resources/server infrastructure.

On data, its obvious Big Tech is the clear and only winner here: Google, Meta, Microsoft, Apple, Amazon, Yandex, Baidu. Now this has been known for sometime in the AI space, and so, much work was done to create open source data sets in hopes of levelling the playing field. Without going into the ethics of creating such datasets, like how much of the data was scraped without permission and may violate copyright, these datasets allowed some AI startups to create some good results, like OpenAI or Hugging Face.

However these models give huge performance gains when trained on larger datasets, and compared to the amount, frequency and informational structure of datasets Big Tech gets, these open source variants hold no candle. In terms of data, its a Big Tech game. OpenAI, not so open anymore, took a $1B injection from MSFT, to exclusivly use MSFT's cloud infrastructure and for MSFT to be the first prefered partner.

In specifics to LLMs as search engines, well one would need search engine level data, permissions and a somewhat existing product. In this space there is an even smaller section of Big Tech viable: Google, Microsoft, Yandex, Baidu, maybe Apple (rumblings of them having a search product and how the $15B+ Google pays Apple every year to be the default search engine on Safari on iOS is actually also implicity to dissuade Apple to create a rival search engine).

In terms of compute resources or infrastructure its even smaller. Only Amazon, Google and Microsoft have the data centers at such scale and Amazon doesn't really have a search crawler. You get the idea, its a MSFT and Google duopoly, and extension of the current search market even more in favour of the two tech giants.

Nvidia exists as a crucial monopoly in this space as well. Their high end GPUs are used to train these models and Nvidia has a monopoly in this space right now due to their bleeding edge compute but also due to software frameworks being built around these GPUs for decade+, resulting in other GPU's having interoperability or comparable performance issues. At this scale, any performance issue is going to be very very costly. Google tried to dethrone this Nvidia monopoly with its TPUs and there are numerous startups appearing, creating custom processors built even more specificall for AI tasks.

A fair part of the recent boom in SMC startups is in one part due to new global scale use cases in AI/computational modelling and in another part a paradigm shift from monolithic chips to a more modular chiplet. A paradigm shift in the monopolised sector that should also be investigated to ensure new SMC companies can make it.

In conclusion, its a Google, MSFT game with MSFT really desperate to take marketshare from Google. I don't see it working out, LLM has a very diferent function to search and the business model doesn't make sense (nothing new with Silicon Valley, start something unsustainable by violating anti-trust/sector specific regulation, and hope it becomes profitable once you get near monopoly marketshare).

Even with such dominance, MSFT and Google are acquiring LLM specific startups. Regulators should be watching. MSFT just bought 49% of OpenAI and Google just bought a $300M stake in a LLM startup called Anthropic. Both MSFT and Google already have world class internal AI teams by the way. Google even has two: Google BRAIN and DeepMind (acquired in 2014 iirc).

Here's an excellent resource on the details of this space:

https://www.semianalysis.com/p/the-ai-brick-wall-a-practical-limit

https://www.semianalysis.com/p/nvidiaopenaitritonpytorch

Might be worth reaching out to the author there and see their thoughts on the matter. Extremely reputable and knowledgeable in this space.

Expand full comment
founding

Kahn = a good person.

Her success = a model for all to study.

Keep writing so we can follow most clearly ! Many thanks !

Expand full comment

Meanwhile that comically arrogant and knuckle-headed faux-intellectual economist Herb Hovencamp (who Matt - and commenters 😉 recently trashed soundly) just read Christine Wilson’s garbage take in the WSJ Garbage Take (“Editorial”) page, and he must be all like… 🧐😳😏 🎶 “Who’s that lady? (Who’s that lady)… Beautiful lady (Who’s that lady)… What a fine fine lady… 🎶 ...and he’s so smitten doesn’t even notice that his monocle has fallen atop his soft-boiled egg.

Christine Wilson & Herb Hovencamp... now that’s what I call a pair that can’t be ace high.

Expand full comment

It’s Billy Graham. Kind of a big error, actually to people who know any history.

Expand full comment
Feb 18, 2023·edited Feb 18, 2023

searching Bloomberg's archives for Mark Pittman's reporting on the fraud orchestrated by Phil and Wendy Gramm is well worth the effort. Today Phil Gramm has retired from UBS as its VP for wealth management and Wendy Gramm, formely chair of ENRON's Audit Committee [can't make this stuff up] now has a sinecure at the Mercatus Center, a wholly owned subsidiary of Koch Industries[you really, really can't make this stuff up.]

—Benchmark for successful bank heist is clean getaway.

HOME / GLOBE / OBITUARIES

Mark Pittman, 52; reporter who foresaw subprime crisis

By Bob Ivry Bloomberg News / November 28, 2009

Mr. Pittman joined Bloomberg News in 1997, wrote stories in 2007 predicting the collapse of the banking system.

That year, he won the Gerald Loeb Award from the Anderson School of Management at the University of California, Los Angeles, the highest accolade in financial journalism, for “Wall Street’s Faustian Bargain,’’ a series of articles on the breakdown of the US mortgage industry.

Mr. Pittman suffered from heart-related illnesses. The precise cause of his death was not known, said his friend William Karesh, vice president of the Global Health Program at the Bronx-based Wildlife Conservation Society.

http://archive.boston.com/bostonglobe/obituaries/articles/2009/11/28/mark_pittman_52_reporter_who_foresaw_subprime_crisis/

Expand full comment

There is a physiological and psychological explanation for "conservatism" and the behavior of Donald Trump and his gullible enablers.

Despite Trump's obvious behavioral flaws, the proper question people should be asking about Donald Trump is not whether he has dyslexia or is incompetent or is blatantly corrupt and a con artist. The proper question which needs to be asked is "why do so many gullible people believe him?"

The "explanation" is that individuals with dyslexia (many of whom are not even aware of their vision problem) and an authoritarian/stressful childhood tend to prefer an "authoritarian leader" regardless of that "leader's" competence or credibility.

Without the support and enablement of his gullible followers, Donald J. Trump would likely be universally regarded as a dangerous clown and con artist. Instead he is regarded by some as their “anointed leader” and potentially next President. However, I think I have made a discovery which will allow individuals trained in psychology to join the legal and political efforts to stop Trump and the cult of insurrectionist from taking over our country, and explain why Fundamentalism seems to be taking over the rest of the world.

See the web page describing the behavior of Donald Trump and his supporters/sycophants.

http://www.trumphasdyslexia.com/

About 87% of 2018 Trump supporters have a combination of dyslexia (Near Vision Stress) AND an authoritarian/abused childhood. Only about 32% of the 2018 Clinton supporters had dyslexia (Near Vision Stress) but they all had a supportive/nurturing childhood.

Expand full comment
founding

Dennis, I don't think he'd want to be president. If he had a job, it would be Treasury or Commerce Secretary--someplace where he can help re-order American industrial policy.

Expand full comment