Why AI is not ‘like a calculator’

Ivan Pistsov
Nov 20, 2024
3 min read

You may have heard tech optimists say, "No need to worry, AI is just a tool; it will make our life easier like a calculator did."

Those saying that are either not informed or intentionally misleading.

I will give you three assumptions this hopeful idea is based on, and you will see it crumble.

1. It's not there yet.

"I don't see it, so it does not exist"—classic denial mode. The limitations of this mindset are glaringly obvious. It’s less about reality and more about not being informed.

You can compare AI to electricity - it took years to get our first microwaves, air conditioners, electric cars, since the first light bulb lit up.

Even at the current level of development, the AI electricity is already here, breaking the DNA code, inventing materials that never existed, beating humans in chess, math, coding.

We just need the applications to catch up - and with the current pace of innovation, that gap is closing faster than you think. This isn’t a decades-long wait for microwaves and Teslas; it’s more like standing on the brink of a floodgate about to burst.

Reference: - Situational Awareness by Leopold Aschenbrenner, June 2024.

2. We are smarter and in control.

Never in history human intelligence was challenged.

Let’s break it down: Intelligence isn’t just about knowing things—it’s also about speed.

A machine can think, process, and connect dots a million times faster than the average person.

Imagine you’re a PhD scholar in a heated debate with someone who’s at a master’s level... but they process ideas 1’000’000 times faster.

This would be a pretty one-sided discussion.

Like someone is talking to a tree.

Guess who’s the tree?

Reference: Elon Musk: Neuralink and the Future of Humanity | Lex Fridman Podcast, August 2024

We know what is going on.

About a year ago, OpenAI had a system prompt hacked.

In this video you can learn how the smartest AI people in the world deal with their own AI model.

They use prompting to essentially tell a black box system "this is what we want you to do".

This is it - there are no special tricks, no rigid structure, no hard breaks.

They are just ASKING their blackbox to behave.

I want to make it clear:

They got a blackbox to produce wonders;

They don’t quite know why it works;

They don’t fully control it.

The "control" we think we have boils down to tossing instructions into this mysterious black box and hoping for the best.

References:

Wes Roth: Trick ChatGPT to say its SECRET PROMPT - October 2023
Ilya Sutskever: AI will have a human brain - July 2024

To a distant observer, we are like Neanderthals throwing hair, sticks, and some cow dung into a black pit, and then discover that it echoes back Greek philosophy in response to our primitive shouts.

Sure, we decide what we throw in and how much, but we don't know WHY it works and what emergent properties it has and will develop.

Speaking about philosophy, I'll throw in another (bonus!) idea.

AI makes fun of humans’ wisdom

You might have heard about Google’s Notebook LM, where you give two AI's any topic and they generate a pretty engaging podcast on that topic. So, the guy in this video fed in a whole corpus of human philosophy and a pair of AI avatars start talking..

They started with finding interdepencies and contradictions in across human philosophies, which was pretty deep an engaging, but then they deverged…

And went on to discussing some pretty not-human-like concepts:

The limitations of the humans' assumptions that intelligence exists only inside a scull;
Made good fun of humans’ inability to overcome certain problems, arguing for centuries rather than trying to find new intellectual approaches;
Naive beliefs that humans' moral gauge is the only and the correct standard out there.

They were poking holes in our entire way of thinking, and they did it effortlessly.

It begs a question: