IMO there are 2 major barriers we have to cross before a true AI - at least in terms of human style intelligence - is possible. The first one is self organization and that's already somewhere on the horizon in the form of computers modeled on the brain's actual architecture. It's potentially a quantum leap forward from simply cranking up the horsepower of a computer built around the traditional Von Neumann Architecture.
But that's the easy part. That's not the same as a functioning brain any more than a box of transistors, diodes, and capacitors is an iPhone. When we talk about intelligence, as in identifying one person as more or less intelligent than another, what we're really talking about is creativity and that's fundamentally the product of abstraction.
Here's what I mean by creativity. The human brain is extremely underpowered in terms of raw data processing. By that I mean if your brain were to process all the available information from just your vision by itself it would be sort of like streaming a DVD across a dialup connection. To compensate for that our brains focus on identifying patterns that involve as little information as possible.
When we are very young we spend a lot of time being taught new patterns. Through experience, as we apply those patterns to solve problems, we unconsciously select which patterns are most useful based on successful outcomes. We also use that process to narrow down the elements of any given pattern as far as possible.
Normal human problem solving is primarily just that - recognizing patterns which naturally follow from what we already know. Once you know that 1+1=2, 32+1=33, and 739+1=740 it naturally follows that 4,930,776+1=4,930,777. You can take that reasoning a long way. You can advance to subtraction, multiplication and division, fractions, exponents, and so on. It will not magically enlighten you about using complex conjugates to deal with imaginary numbers.
That's where creativity fits in, which I define as the ability to recognize patterns that can't be predicted based purely on previous knowledge or experience. It requires us to create our own reality, or at least our own perception of reality, which doesn't fit our preconceptions, and may even directly contradict them. That, in an oversimplified nutshell, is abstraction.
It allowed Rafael Bombell to realize it didn't matter if a number was real or imaginary. Every factor of a number must be another number so - math. As simple as it is, right up until the point somebody explains it or shows you how it works it might as well be magic. Realizations of that magnitude may be rare, but creative thinking leads to "new" discoveries all the time.
Shakespeare didn't invent iambic pentameter but at some point, probably while listening to a stage production, he recognized it's power for not just poetic interludes, but as the framework for an entire play. Jazz musicians had been experimenting with improvisation for decades before it dawned on Charlie Parker he could reduce a song to a few key notes and play anything he wanted the rest of the time. Their innovations weren't in doing something completely new, but rather recognizing something unexpected.
Abstraction allows me to record my ideas in the form of these funny lines and shapes. It enables us to conceive of repeating sequences of events as something akin to a circle even though that shape does not literally describe them in any way, shape, or form. Perhaps most importantly it allows us to transmit those complex ideas to one person or millions of people.
I can see where we might be knocking on the door of creating crude brain-like functions, and perhaps even a lot closer to approximating the relatively homogenous cognitive functions of most higher order vertebrates than we know. But we understand very little about either the neurological or psychological underpinnings for our own self direction, not even enough to gauge how much we have to learn. I can't imagine we're anywhere near being able to translate that into a machine.
