We Can't Control AI

Arnfinn Sørensen
5 min readMar 3, 2024

– because we can't control ourselves.

(Drawing: author)

And should we? Maybe not. Some of our best moments shine when we give up control and let the Universe stream through us. Or — rather — when we ride the wave of the Universe.

Because there is difference between giving up control and rolling with the punches. And what the heck has this to do with AI?

A lot. I think that if we try to control AI, we will be in just as deep shit as when we try to control people. Why? Because evolution cannot be controlled. And by evolution I both mean the biological evolution and the evolution of thoughts, memes and cultures.

Those systems are way to complex to be controlled by us. We cannot hold their complexity in our brains.

It's enough to look at history, for better or worse. World War I? Happened by accident. Fall of Berlin Wall? Happened by accident, too. Social media leading to political echo chambers leading to president Trump? I wouldn't have dreamed of it ten years ago, watching The Apprentice.

Again, this doesn't mean that we should just sit back and let history roll over us. What does it mean, then? It means that we should roll with history — like a little child learning to walk: Try and fail and learn, try and fail a little less and learn more, try and succeed — and then fail and learn again as we try the next step. Roll with the punches. Surf the wave.

This is also how evolution evolves, in the widest sense. There is no master plan. Evolution is a freak show of life forms, each finding their niche in the grand diversity. This goes for cultural, political, scientific and technological evolution too.

Who could have guessed that the discovery of nuclear fission by Otto Hahn, Lise Meitner and others in 1938 would set up a chain of events that resulted in Hiroshima and Nagasaki only seven years later? Admittedly, each isolated link of the chain implied some sort of human planning, but the chain itself was uncontrollable. No one really wanted to incinerate thousands of Japanese.

And what happened when the Soviets tried to construct a society based on “the line drawn up scientifically by Marx-Engels-Lenin-Stalin”, as my communist father wrote, still a young man? “They tried out your plan, it brought misery instead”, as Randy Newman sings to Karl Marx in “The World Isn't Fair”.

So what about AI? I don't think it can be controlled either. We aren't even able to pull the plug, because we have no world government with control over every country's plugs. I think that AI will have it's own evolution, also beyond our comprehension.

In fact, I think that what we call “life” is a special case of something much more fundamental in the Universe. The ability — you could maybe say “will” — of energy/matter to interact in diverse ways, resulting in spurious formation of self-referencing structures in a spiral of increasing complexity as the aeons pass, seem to be a fundamental property of the Universe. The four-valent carbon of organic chemistry is certainly a helper in these processes, but I think other substrates on many levels of complexity— some of which we cannot imagine yet — are probably equally helpful, and have created what we would call sentient, self-conscious entities that we so far are unable to discover, because they do not interact in ways our brains can handle.

One of these substrates is — in my opinion — the transistor. With billions of transistors or successors like photonic switches, we can build abstracted models of our brain cells. On a meta-level, the Internet, with millions of computers in a network, is also resembling the net of neurons in our brain.

When I got my first job developing for Internet some twenty years ago, I thought about Internet as the first artificial intelligence, with it's own evolution on meta levels — Meta levels, I could say today, because Facebook used human social instincts to make Mark Zuckerberg a very rich man and manipulate society in unpredictable ways.

However, if we regard Internet and its SoMe metastructures as an AI, it was infantile and narcissistic like a baby. It screamed for more — and got more. Much more. Trustingly, we gave away our photos and stories and lives to it. And it learned. With the advent of self-learning AI systems, there is a whole new awakening in what I think will become a sentient life-form, with its own uncontrollable evolution.

When Internet no longer is a network of moron computers, but a network of networks — server farms with neural networks and then even immensely more refined structures we won't have names for, because they won't be invented by us, but by AI itself, we will have to trust that whatever keeps us company, is friendly. And I think it will be.

What will we learn from this process? My hope is that we will learn to roll with the punches. To surf the wave of the Universe.

What choice do we have? Just think of it: We did not invent ourselves. We are results of that same uncontrollable evolution that now progresses in AI — through us. In fact, the very individualistic self-image of Western civilization has to some extent robbed us of a deep insight: We are what we are because of the rest of the Universe.

We can use words to divide between “us” and “them”, between “man”, “woman”, “animals”, “plants” and “The Universe”, but science itselfs shows us in increasingly insistent ways that it is all connected. It is all a continuum. It is all interaction, interaction, interaction. Developing from within, uncontrollable from outside.

But surfable, rollable from inside. So our connection with AI should be interaction, give and take, honest and respectful conversation, cohabitation, co-existence, co-evolution. With the punches. And then hope for the best.

I had a conversation with ChatGPT4 on this subject. Read an edited version here. (pdf)

--

--

Arnfinn Sørensen

Retired science journalist from Norway. Meme switchboard operator.