Today's Artificial Intelligence is nothing to fear about; it is not AI

LLMs are training hungry.

From there the need and surge of "prompt engineering". Adding a "human taste" to their responses. A simple analogy is having humans full time training a parrot to regurgitate an answer according to instructions of its master. There is absolutely no intelligence there.

All the fuss we have seen lately is extrapolating some good answers by the LLMs as thinking of them as "intelligent entities". I suppose is part of the subjective, ghostly, supernatural part of our human brains that want to believe in that, as per Carl Sagan maintains in "The Demon-Haunted World: Science as a Candle in the Dark."

In addition to that, merchants that spot a market of believers will try to profit on that, and transform a million dollar machine learning product into scores of billions, scare the bejesus of governments, and bring the public into an state of anxiety.

I think the best thinkers on AI are true scientists with no commercial interest in the AI commercialism. Usually crossing disciplines such as physics, neurobiology, mathematics, statistics, linguistics.

The argument they make is that we don't yet have the technology for Artificial General Intelligence, or quasi-human sentient machines.

1. We haven't yet invented the computational system fast enough to create thought. All of our computational systems are still based on silicon and are digital. The most rudimentary animal intelligence -including humans- are analog intelligent. Quantum is a temporary patch; still is digital. Can do things faster but still is digital.
2. We don't have the operating system that could learn by itself in the evolutionary way like other basic intelligence in the planet. If the operating system is able to learn, is able to create sentences by itself. If it's able to create sentences then will be able to create a language to communicate. And then, thought.
3. The current AI is incapable of auto or self-correction. Code and algorithms have to be manipulated.
4. Today's AI cannot take decisions by itself; it acts according to a set of instructions and algorithms fed by data. In AGI we are looking for an entity that can take not only one decision at a time, but many.

The "AI" we are seeing daily mentioned in the news is pretty good machine learning, and highly dependent of human intervention. Think of a marionette. Of course, people who sell it will tell you is "intelligent", so you can buy that bridge. This is the commercial part that has nothing to do with science.