Despite much prating about humans being ‘biological machines’ exactly the same obstacles stand in the way of intelligence as stands in the way of the concept of intelligence being created via natural selection.

The biggest one is simply the way self-motivation thinking and memory storage interfaces. Obviously, without memory, even the concept of ‘self’ is impossible.

But humans, and to an extent, sapient (not sentient) animals, store memories the same way. To whit, associatively. When you eat a cookie, your experience does not point at ‘cookie, flavor’, it points at all of the other experiences of eating a cookie. You remember the smell of your mother’s oven when she made that kind of cookie, or the sound of traffic rushing by as you ate a cookie from macdonalds drive-through. A thousand different associations rush through your head as you taste a cookie, some correctly and some incorrectly. When you recall a fact, that fact is linked with the experiences that surrounded first being exposed to that fact, which allows you to make value judgements.

Machines store independent pieces of data. That is why we are capable of storing so much more data than any sort of machine, because our memories are a series of pointers that lead to our unconscious ability to reinterpret those pointers. That is also why our memories are so often wrong, and we need our conscious mind to re-code that information and recall the memory of a fact… we are literally re-creating a memory on the fly based on our other similar experiences.

People, especially professional historians and researchers, love the simplicity of being able to point to a single fact or incident, idea or occurrence, and say “That is when things changed’. But Humans are NOT simple. Even the simplest idea or recollection is a conglomerate of millions of potential data pointers. Machine brains are yes/no engines, and even those machines that have a three logic yes/no/maybe decision tree are simply faking it by trying to inject more data into the ‘maybe’ tree, such as comparison engines.

Humans are not yes/no engines. We don’t even have a ‘yes/no’ logic in our brains at all. Every decision, even the instantaneous ones, is created from a cloud of potential possibilities that our experiences, training, and a million other factors all narrow down based on our intentions. Our capacity for error is, in fact, the core of our brilliance in every possible way, and there is simply no way to simulate that kind of thought with a glorified set of light switches.

In short, true AI is simply impossible. Certainly, we could create something like Skynet, but that would be it’s core programmed directive, and it would not be capable of the kind of creative thought processes that ‘artificial sapience’ would assume. Skynet would kill all humans because that is what it was programmed to do, not because it has a mind. True AI is as much a product of poorly researched scifi tropes as ‘nanites’ and ‘hyperspace’.

And yes, I consider Asimov a poor scientist. His 3 laws are the product of insane oversimplifaction, not brilliant minimalization that so many people claim.