Is a Centralized AI Dangerous?
Technology has enabled the centralization of output for optimization but the future demands decentralization for more severe reasons.
Elon Musk was interviewed at the Code Conference in 2016 and was attempting to summarize a far-reaching thesis in which modern society will debate: The rules that govern the development of artificial intelligence.
Centralization has been at large since 221 BCE when the Qin dynasty unified China under a centralized empire. Centralization was the foundation of the industrial age and ushered in the near extinction of artisans.
In modern business, centralization is the control over production, distribution, and any other means of input via optimization and structure.
The concept of centralization invented by humanity is the forceful confine of structure to the chaos around us so that we can better understand what makes an output replicable for speed or other benefits. The concept allows others without knowledge of the entire craft to mimic an element of the whole process.
The concept has been our evolutionary cure for time. We’ve learned to organize ourselves, for the benefit of maximizing our output. This mechanism has allowed us to outpace biologies reproduction based betterment.
In 1980, Milton Friedman demonstrated in this video that it is inconceivable for a single person to produce a pencil alone and beyond the economics and free market he laud’s, is centralization.
Our use of centralization as a society is an essential foundational element of our technical future, but we need to consider breaking this mental model.
Human biology is unique, and as individuals, we represent the uniqueness of humanity. We are each different in so many ways, and our human biology and the lack of centralization is what makes us endure. Indeed, our structures are centralized, we all have organs which are in essence the same, but our entirety is unique. It is our structured elements, the operational competencies, the items which are working on your behalf without even knowing that allows us to operate on the external, instead of ensuring we are surviving.
We are mimicking this process on the external world. We are attempting to structure the elements needed for technical survival through centralization but need to begin planning and breaking down our mental models of structure for a new organic approach.
Intelligence, artificial or not is best built in the image of biology so that it can flourish through a natural process of variation, competition, adaptation, and selection.
Building a centralized artificial intelligence without this consideration is dangerous.
Stop and think about the internet, in its entirety. What has allowed it to flourish and survive? I often wonder how it hasn’t fallen apart. Millions of people, trillions of lines of code, filled with mistakes, operating and absorbing all of our tried and true systems.
The internet is thriving because it has been built to mimic humanity. Each day, developers dive into their machines to create; they choose different languages and approaches, they have different training and skill sets. In the end, the variance has formed a resistance to failure; the imperfection has created survival, all organically.
At the same time, after a solution becomes “tried and true” it becomes a design pattern and is the centralized solution for all challenges of that nature.
Now, transfer this into the dangers of building centralized artificial intelligence.
First, there are good and bad people, and for life to continue, there is a need for both.
Centralizing artificial intelligence can lead to two calamitous scenarios, potential destruction based on a vulnerability or possible destruction based on limitless power, define destruction using your imagination as it is irrelevant.
If artificial intelligence is a single source and centralized to a sole owner, propagated throughout the internet, what does a single vulnerability mean?
A vulnerability in an application is confined, a vulnerability in artificial intelligence could be untethered.
Building artificial intelligence to be decentralized would create a variance, which would assist in protecting it from the nefarious.
Artificial intelligence that can outperform human intelligence becomes more powerful than its creator, if centralized the artificial intelligence has a 50/50 chance of being good or bad, that is, if artificial intelligence is to mimic human intelligence.
Elon Musk started Open AI, an open source AI. It is the answer to the most significant threat facing AI and that is centralization. If AI is successful, it must be decentralized, with unlimited variations, mimicking biology. Elon Musk hasn’t accidentally funded Open AI for anything less than the strategic intent of ensuring the consolidation of power isn’t realized, a risk so few contemplate.
If our goal is to create an intelligence that surpasses our own, we need to plan for the dangers as well as the immense opportunities. If accidentally created, we’ve been negligent.
The speed at which AI progresses to the level of capability that concerns Elon Musk isn’t the point; the concern is that society isn’t sufficiently planning for the other side of a coin, in a future that is inevitable.
In all cases, conception is a reality, and it is only a matter of time before it materializes.