Are you seriously claiming that computers with neural networks human brains and have the same capabilities?
What about Iron man suits? Why settle with walking in this fantasy future of yours?
If airplanes had the same risks and reliability as the Wright brothers' first planes they wouldn't be very popular. That's where end to end machine learning using computer vision is today, and also likely in "a not too distant future". Autocomplete and driving have different requirements.
Um.
Look:
We run on neural networks. We are not, except by emulation, a von Neumann architecture. We're massively parallel processing neural networks running on (for a processing unit) dead slow chemical processes, put together by chance and Nature's bloody tooth and claw into something reasonably efficient. How neurons actually
work was figured out in, well, my lifetime, at least. Research on how nerves, brain cells, and the like continue, of course, but the bloody building blocks are known.
And, once How Neurons Work was figured out, bright people said, more or less, "Well, this works for living things. How's about we take a look at the kind of problems it can solve?"
And, just like practically everything else in engineering and mathematics, it turns out that approaching certain problems with neural networks results in
much faster solutions. Image recognition. Object recognition. In their own weird way, neural networks act almost like analog computers, with fixed and ridiculously fast input-to-output solution propagation delays. In fact, from reading the popular engineering literature, it's clear that the people investigating what NN's were capable of, and how, were starting with problems that the human brain was known to solve efficiently, but, at the start, nobody had a clear idea of
how.
We now have a
much better idea of how NN's work in living things. And those ideas have clearly been transferred to how humans can
make NN's that have desired processing characteristics. When one listens to hardware researchers talking about what NN's are doing in their hardware, it's clear that there's a foundation of biological methods to their madness. And, probably the reverse, with biological NN researchers getting insight from the hardware types.
As far as integrating NNs into hardware: As these things go, as fancy a job as being currently done goes, these are the Early Days.
As far as Iron Man suits go: They run on unobtainium. Real, live,
exoskeletons are commercially available. Extending one of the latter with a driving computer off of a Tesla, like what they're using with the Optimus robots over a Tesla, is hardly a stretch.
As far as the Wright Brothers and early fliers go: The math of the time was barely up to getting a machine up into the air. As you may recall, the Wright Brothers approached the problem with one heck of a lot of guided cut-and-try and the math of the day. Even with all that, they had to invent their own ICE that was light and powerful enough to do the job. And, even after their first successful flight, it was still dangerous as all get-out. Research project, yes. Not for the general public, you betcha. But everybody involved (well, all except the Old Geezers who'd start things with, "Back in My Day...") saw the capabilities and were All For It. It took from 1903 until some time after WW1 before commercial flights began.
We have some notable advantages over the Wrights, not the least of which is a heck of a lot more computational ability with much better math and simulation.
So, what's your point?
