Deepseeking supremacy
The AI pie in the sky
I finished my last post with a promise that I will return to the subject with the following prompt:
AI has two possible directions: to become the best tool humanity ever had to understand and control the world, or to become an instrument of our oppression and eventual self-destruction. AI will not make the decision about its direction. One way or another, we will. Like it or not, you are part of it.
When I started writing the post, I didn’t know about Deepseek, but everything I learned about it so far fits perfectly into the paradigm that has an essential presence in this Substack. The open source Deepseek seems to be a possible distributed decision-making alternative to the proprietary control of the centralized American versions (LLMs).
Just look at the three clowns on the picture above.
I always had a visceral dislike for Larry Ellison and Sam Altman, especially Altman. I did not know about Masayoshi Son, but from what I learned about him in the past few days, he is a perfect complement to the other two. They are not geeks; they are not engineers; they are venture capitalist hustlers. They are salesmen. All three are just trying to ride the wave of the hype. Their understanding of the technology is as deep as that of a typical salesman. What they do understand is who they are selling to.
The announcement was mostly a publicity stunt. These three hustlers just started a fundraising with the public blessing of the American president. As Elon Musk pointed out in a tweet, they do not have the money. There is no federal money involved, the only role of the federal government is to accelerate permitting, which is something they were planning to do anyway. It was just a photo-op with the blessing of the President and the prestige of the office in which it was given.
This YouTube video has some telling clips from the press conference that followed the White House announcement. All the dreamy, miraculous, Sci-fi BS promises of the AI pie in the sky. The comments of the three entrepreneurs were nothing more than empty promises in search of investment dollars.
The second aspect of the sales-pitch was to show prospective investors that they have the most important potential buyer of their services lined up, once the products they are busily working on is available.
The ultimate buyers of centralized controls are governments. The lesser buyers are the companies who want to mine vast amount of data most intelligently so that they can sell you something. Larry Ellison has even said so much in the press conference after the announcement.
The video linked above also makes this point. Larry Ellison owns Oracle, the world’s largest database SaaS (software as a service) provider.
A dystopian society, a totalitarian government and the AI marvels that are about to come, needs a lot of data and computing power. Providing such services is what Ellison is doing already, the AI software pie in the sky is what Altman promises to deliver and Son pledges to find money for.
The only thing that is missing so far is a profitable, bug-free and working product.…or a real demand.
A little aside:
Contrary to what the name implies, OpenAI is anything but open. It is a proprietary system with closely guarded secrets. Elon Musk was one of its founders, but broke with the company when the rest of the founders decided to make their product proprietary.
The promise of Deepseek is that it is actually Open Source.
All of what I said here so far is just the circus. Politics, salesmanship, hype.
But why does it matter? The answer is in my previous posts, especially in parts #2 & #3
Their relevance to this post is this: AI is NOT one thing!
Narrow AI is a tool to solve specific but complex problems.
It is distributed, evolutionary and clearly limited in its purpose.
Applications are: self-driving, problem solving in various disciplines, pater recognition in medical tests, in material sciences and engineering. AlphaGo and Alpha zero are perfect examples. Simple rules, nearly infinite amount of data and complex decision making.
Narrow AI has wonderful potentials.
General AI is an attempt to mimic human intelligence with the promise that with enough power it can become infinitely smarter than we are.
It is centralized and generic in its purpose. It will reach self-awareness first then it will become Artificial Superintelligence. We will need to give absolute control to its handlers to keep it safe for us.
Maybe we will be able to ask it (or whatever the proper pronoun of a superintelligent entity is) about the meaning of life, the universe and everything, even though we already know the answer to that.
It is 42.
Just listen to Larry describing the promising dystopian future where both the cops and the people will know better than to misbehave.
Where just by waving the magic wand of AI, the existing power structures of health-care will be rearranged to serve the interest of the people instead of the medical professions and the drug companies. Easy-peasy.
As I am writing this, I still cannot understand how can anybody take this bullshit seriously. It is all about control, it is all about politics and all about power. That is what the hustlers are selling. Saying in effect that “we will give you absolute power over your subjects for a good chunk of money and a seat at the table.”
I would never suggest that the Chinese are any better. What makes Deepseek better is the fact that it is open source; that it is small, efficient and therefore accessible to a large number of people creating a pool for AI evolution. A great number of people will create a great variety of tools in a competition where the different concepts and approaches will have to compete with each other. What makes Deepseek great is that it is a wake-up call for the LLM hustler utopians.
Its approach is more conducive to distributed decision making and specialized tools.
Are you with me in the differentiation between the two approaches and their larger implications?
Let me know in the comments
More from here
References
I could give you a dozen more, but I will let you find them yourself
Like everything else on Substack, this is a reader supported publication.
You can help it by following or subscribing.
You can engage with it by clicking on like and/or commenting.
A ‘like’ costs nothing and is worth a lot.
You can help this Stack grow by sharing, recommending, quoting or referencing it.
You can support it by pledging your financial support.
Any and all of it will be much appreciated.
Thanks for writing this article and others related to AI. As one who knows nothing about AI, this has started the process of opening my eyes. Obviously, I have a great deal to learn.
Thanks Zork. This is a great piece to help people to see the hucksters behind the AI hype.