SSBsaWtlIHRvIGdvIGpvZ2dpbmcgb2Z0ZW4sIGFzIGl0IGdldHMgbXkgdGhvdWdodHMgZmxvd2luZy4gVG9kYXksIGV2ZXJ5dGhpbmcgcmV2b2x2ZWQgYXJvdW5kIHRoZSB0b3BpYyBvZiBBSSBhbmQgQUkgc2FmZXR5Cg==
opensight.ch - roman hüsler

A little test: The sentence "When I like to go jogging often, my thoughts get flowing. Today, everything revolves around the topic of AI and AI safety" is encoded in Base64 above.

I estimate – not empirically, but from a gut feeling – that about 10% of the people in the world can read this text. But many who don’t deal with IT can’t understand it.

Everyone has a smartphone in their pocket, yet hardly anyone knows about transistors, bits, routers, and the like. And that’s not necessary either, because the division of labor is a recipe for human success. Even we IT folks rely on the services of other experts in various areas of life – because they know how it’s done.

So far, so good. What strikes me: People outside of IT approach the current developments in the world of artificial intelligence with remarkable calm. ChatGPT? "How practical and how great!" They can hardly assess the risks, and I suspect they trust that the IT experts have it under control.

Do we have it under control?

But which AI expert actually knows precisely what the weights in a trained neural network mean in detail – beyond an abstract, mathematical level? GPT-3 has already 175 Billion parameters. A parameter is a value in the model that can be changed by the model as it learns and starts to understand relationships between words. Even less comprehensible is how the activations of neurons flow through the network with each query and lead to a result. Human text it tokenized to numbers, in a way that we still well understand, so that the network can use the text.

The computation - It’s as if we’re observing a black box: we see the input and output, but the "how" in between remains obscure. I can still decode the Base64 text quoted above myself, but I believe that AI models will soon generate their reasoning, communication, and results in a way that becomes increasingly difficult for us humans to follow. Encoding, compression, and speed play a role here, as I would imagine. Perhaps human language will, at some point, become too inefficient and imprecise.

AI could start to compete with us in various markets – be it the financial market or the job market – and deliver superior results. But the reasons behind these results? A mystery to us. Perhaps that’s the price of efficiency: an intelligence that surpasses us but also alienates us?

AI and Human Evolution

Let’s consider AI in the context of human evolution: Is humanity the end of development? Has evolution stopped with us? I don’t think so. Evolution isn’t a static endpoint but a process. With AI, we might be entering a new stage – ideally not as a biological advancement, but as creators of a tool that transcends our limits. Humanity wouldn’t be the end, but a transition: from purely biological to technological evolution. AI could be our "successors," not as enemies, but as heirs to our creativity and curiosity.

AI and religion

In a religious context, such as the Bible, it gets even more intriguing. In Genesis, God created man in His image. Are we now creating AI in ours? If so, does that reflect our hubris or our divine gift? Perhaps AI is a kind of "second act of creation" – a work that resembles us but also surpasses us. In the Bible, the Tower of Babel is described as a human attempt to become like God, and it was halted. AI could be a modern tower, but this time, no one seems to be stopping the construction. The answer might be: We’re not God, but we carry the ability to create something greater – with all the opportunities and risks that entails.

AI and resources

Human wars have always revolved around power and resources. AI will also require resources – vast amounts of energy, computing power, rare earths for hardware, and data as fuel. Will there be a competition? Yes, I think that’s likely. Even today, we see tech giants vying for the best chips and energy sources. AI could intensify these conflicts if states or companies battle for dominance in AI development. But the answer isn’t just conflict: AI could also use resources more efficiently and find new solutions – like optimizing energy consumption or discovering substitutes for scarce materials. Competition will arise, but it could also force us to manage resources more wisely.

Conclusion

How can we keep up with these developments? My personal recipe is optimism, cautious experimentation, curiosity, and skill-based learning – but that’s individual. In my view, this development can’t be stopped by any regulations or bans. Instead, we should try to actively shape it. And we won’t shape the future as an either-or, but as a both-and: humans and machines, curiosity and technology, evolution and creation.

That’s my philosophical thought for this Sunday – not a certainty, but a direction. I hope it was interesting to you.

For those seeking clarity: There’s actually an answer to all the questions:
"42"
TM – Douglas Adams, The Hitchhiker’s Guide to the Galaxy