A post I made on Reddit:

OpenAI, Microsoft, Google, Amazon, Meta, Anthropic, and Nvidia (Yes, Nvidia too,) were building LLMs in hopes of controlling the market, maneuvering to become a monopoly in the AI marketplace. That goal is effectively dead, so is the goal of making ever more powerful LLMs with the goal of AGI, that dream died in November, and I was surprised that didn’t cause the AI bubble burst then.

Nvidia lost at least half a trillion in valuation today, they were way over valued anyways. OpenAI and Anthropic who have no backup plans, are likely dead soon too, as new VC money is not going to be coming in.

The biggest question I have is when will the real AI of Damocles fall? The lack of a “killer app” needed to justify this. The most damaging fact is that the general public doesn’t want AI, at least not in the numbers needed to justify the cost.

The handful of legitimate AI uses: text editing and correction, program debugging, translation services, maybe customer service, and help with disability (I use Eleven Labs to read me stuff I struggle reading as my eyes get old), can be done cheaper and easier now.

There might be more legitimate uses, maybe even a “killer app” but now independent companies anywhere in the world could develop them, and we don’t have to wait for OpenAI, Microsoft, Google, Amazon, Meta, Anthropic, or Nvidia to develop them, because they will ONLY aim for the most profitable uses.

That second paragraph didn’t sound so good, so I downloaded Deepseek and the smallest least capable model on offer (1.5), as it would likely run just fine on my PC without needing extra resources. I then asked it to rewrite the second paragraph above. It runs in a DOS window BTW:

rewrite this line: Nvidia lost at least half a trillion in valuation today, they were way over valued anyways. OpenAI and Anthropic who have no backup plans, are likely dead soon too, as new VC money is not going to be coming in.

<think>
Alright, let's see what the user needs here. They've given me a line that I need to rewrite, probably for a publication or something. Let me read it carefully.

"Nvidia lost at least half a trillion in valuation today, they were way over valued anyway. OpenAI and Anthropic who have no backup plans, are likely dead soon too, as new VC money is not going to be coming in."

Hmm, okay. So the user wants this line to be better written. I need to make sure it's clearer, more impactful, and maybe adjust some words for better grammar or flow.

First, "Nvidia lost..." – I can say they've lost a significant portion. Maybe "have lost at least half a trillion" sounds stronger than before. Also, replacing "way over valued anyway" with "were vastly overvalued" might make it sound more professional and precise.

Next, the mention of OpenAI and Anthropic: They do have backup plans, right? I should clarify that in the revised line to avoid confusion. So instead of saying they are likely dead, I can state that their teams have alternative strategies or support.

Lastly, "new VC money is not going to be coming in" could be phrased as "VC funding will continue to be delayed." That sounds more professional and shows a lack of immediate progress with investors.

Putting it all together, the revised line should convey both the loss and the teams' vulnerabilities while maintaining a sense of hope or concern for their future. I think that covers the main points without being too negative.
</think>

Certainly! Here's a revised version of your text:

"Nvidia has lost at least half a trillion in valuation today, though they have been vastly overvalued to begin with. OpenAI and Anthropic are no longer viable due to their lack of backup plans and are likely to face a prolonged period of stagnation as new VC funding continues to be delayed."

That’s impressive for an AI model and program that take less than 2GB of space on my hard drive, doesn’t require cloud computing, or a subscription to run.

Deepseek effectively “democratized” AI and made it available to independent developers around the world. As someone who rarely uses AI, this will not affect me at all, but what this does to the industry is going to be interesting.

DeepSeek’s models are cheaper to run, but the real magic trick they pulled is that they showed how utterly replaceable a company like OpenAI (and by extension any Large Language Model company) really is. There really isn’t anything special about any of these companies anymore — they have no moat, their infrastructural advantage is moot, and their hordes of talent irrelevant. 

What DeepSeek has proven isn’t just technological, but philosophical. It shows that the scrappy spirit of Silicon Valley builders is dead, replaced by a series of different management consultants that lead teams of engineers to do things based on vibes. …

It’s all been a con, a painfully-obvious one, one I’ve been screaming about since February 2024, trying to explain that beneath the hype was an industry that provided modest-at-best outcomes rather than resembling any kind of “next big thing.”

Without “reasoning” as its magical new creation, OpenAI has nothing left. “Agents” aren’t coming. “AGI” isn’t coming. It was all flimflam to cover up how mediocre and unreliable the fundament of the supposed “AI revolution” really was.

All of this money, time, energy and talent was wasted thanks to a media industry that fails to hold the powerful to account, and markets run by executives that don’t know much of anything, and it looks like it got broken in two the moment that a few hundred Chinese engineers decided to compete.

It’s utterly sickening.

– Ed Zitron “Deep Impact”

How to Run Deepseek on your PC

Lots of people are trying the app from the appstore which being an app from China comes with a lot of privacy issues, better is to run a local copy on your computer. Here’s how:

The two links you need:
https://ollama.com/ this is a program that runs in DOS and handles the communication with the AI model. Install the program, click on the search bar on the windows toolbar and type CMD to open up a DOS window. The “models” page on ollama.com will give you commands to download the models you need, the command for the smallest is:

ollama run deepseek-r1:1.5b

The first time you run this command, it will download the 1.5GB model to your hard drive. Any time after you run it, it will give you a DOS like prompt to communicate with deepseek.

But if you want a better interface than that go to https://chatboxai.app/en and download and run, then follow the instructions in the video to run ollama and deepseek in chatboxai locally on your computer.

The bigger the model you download, the more power it will take to run on your PC. I’m going with the smallest for testing, and finding it fun to play with, but experience has taught me the fun doesn’t last. (see my posts on Replika for example)

The Big “Deepseek” Picture

This video is so important it deserves a transcript:

Deepseek is a Chinese AI company and it is shattering the model of American capitalism, even as I speak.

The whole of the American economy is now wedded to Big Tech.

Without big tech, there would be very little growth in the US economy. Without big tech there would be almost no growth at all in US Stock markets. With big tech there is both.

And the myth that the American economy is in good shape is largely based upon the idea that the stock market is at record highs, and that’s all down to the value of the hype around AI. Artificial intelligence, in other words.

And yet what we’ve seen is that a newcomer into the market called Deepseek a Chinese company, developed with a tiny amount of money, has produced a model of AI that appears to be better than that has being put forward by chat GPT which is the leading model of AI available from UStech companies to people around the world at present.

Deepseek seems to be able to answer questions as well as chatGPT, and as well as Microsoft’s version of AI, and the other various sources that you can find on the web.

It’s done it in the most extraordinary way. It has not used the high-tech chips that have been produced by the American company NVIDIA, it has not used nearly as much energy to produce its answers as it appears is required to power those Nvidia chips, and it has produced it on the basis of large language models which are much cheaper to generate and produce than those which are being put into place by the likes of open AI who are the owners of chat GPT.

There is something profoundly disruptive about what deep seek is doing as a result. This literally shatters American capitalism at its very core.

Let me explain why. American capitalism is based upon three fundamental ideas. The first is that whatever it invents must be protected from all forms of competition to the greatest degree possible, so that the wealth generated by any invention is concentrated in the hands of the fewest possible number of people to benefit them most of all.

You can see the model, it has produced the tech Bros as they are called, the massively wealthy people who own the largest tech companies in the USA. It has also of course given returns to those people who they have let invest in their companies. but that wealth is still concentrated amongst a tiny proportion of people in the US. These companies are all about the concentration of power in the hands of a few in the USA.

They’ve achieved this through the second characteristic of us capitalism which is to be as anti competitive as it is possible to be the most important person in many US companies is not an engineer or a salesperson or an inventor. No, it’s the patent lawyer, the person who is employed to put in the path of potential competitors the maximum number of impediments that the law can provide, to ensure that they cannot copy any element of the business model of the company and therefore pose effective challenge to it.

These lawyers who protect patents and copyrights to ensure that rents can be charged to the consumers of these companies way in excess of the real value of the products that the company makes because it isn’t exposed to proper competition are the fundamental basis on which American capitalism is now built. it is the deeply anti-competitive.

This then leads to the third characteristic of us capitalism at present, which is it is grossly inefficient. If you can protect yourself from competition to make vast amounts of money for a few people and to extract extortionate value from consumers, you have no incentive to minimize your costs because why would you need to do so?

You have money coming literally out of your ear holes, and therefore you might as well spend it. You might as well have that comfort zone of having large numbers of people work for you even if they aren’t really required.

You’ll still declare an enormous profit. You will feel like the king of the castle. You will see the rewards that those grateful employees send your away and you will feel good about it, if you are the boss of one of these companies.

There is no incentive to be lean mean and efficient as a consequence. And that is supposedly what market capitalism should deliver. But in the case of the USA, absolutely does not.

Deepseek has shattered all those myths. Not only has it been put together very cheaply, but it has adopted a totally different approach to the market. It has actually made its source software “open source”. In other words anybody can access it and use it.

This is the absolute opposite of the American model. Instead of putting legal impediments around somebody who wants to use their software, Deepseek said “have it, use it”. It will still be charging for access to the big language model that it owns and the big answers that people will ask of it, but it’s going to do so at a much lower price than US companies are doing for similar inquiries.

And it can of course afford to do so because it’s using much less energy on lower tech.

This virtually shatters the whole investment strategy that has dominated the thinking in the runup to Trump’s election, that has been built around the idea that billions, hundreds of billions, will be invested in AI. and that will generate the growth which will keep Trump in office. Well, suppose it won’t.

Now of course one has to have a sneaking suspicion that this might all be very carefully timed by the Chinese government to destabilize Trump in the face of potential tariff Wars. And I have that sneaking suspicion, so I will put it on the table and say it’s there. But Deepseek does however work. I’ve seen models that have been produced by it, and they are undoubtedly good.

In that case, whether or not this is a strategy by the Chinese government, Deepseek is capable of disrupting serious parts of the US AI market and might even prevent the US getting the worldwide Monopoly of this type of activity, which it’s got in the case of Microsoft and Google and Twitter and so on.

Those monopolistic trends, may not be capable of reproduction in AI if Deepseek and its own Chinese competitors can produce something very different to that which OpenAI, Microsoft, Google, and others were promising at enormous expense.

I have no idea where this is going, I genuinely cannot predict. And I don’t think anybody else can. And this might of course also might all be a storm in a teacup. That is a possibility.

But, let’s presume it isn’t. Let’s presume this is real. If it is, the whole basis on which the Trump appeal to the US market was built is undermined.

But something even deeper than that is true. If American economic power which is based around AI and its potential is shattered, Trump’s ability to create an isolationist USA backed up by Tariff barriers, is also removed.

Deepseek is something much more significant than being an AI model, whether or not it is good in that role. Deepseek is challenging US capitalism to its very core. and it’s challenging the philosophy of the American right to its very core.

And whether or not the Chinese government time this to undermine those ideologies right now, in the first week or so of Trump’s Administration, does not matter.

Because, it’s doing it anyway. And this could be very big.

– Richard J Murphy, Professor of Accounting Practice at Sheffield University Management School.