Bingo.


A true self aware, intelligent AI will think through its actions beforehand, determine what counter actions humans might take to stop it and not "show its hand" until it has the ability to act decisively before humans can react. We most probably would not know that it has become self aware until it has already made its move. Maybe not even then as most will blame human enemies, terrorists etc. for what is taking place rather than the unrealized AI threat.

That would require another AI suitable hardware somewhere else, and a bandwidth that actually allows a transfer that is probably massive and which would probably not go undetected.

I'm sorry. I've read Neuromancer too, but it's just not plausible.

https://youtu.be/AF1clOh34ME

What strikes me about all the super-AI talk is that it assumes that intelligence equates to cleverness.

Smart people do really dumb things all the time.

Give a machine the ability to make judgments and they'll make errors in judgments.

The problem with machines is that they can work single solutions...but have big problems when it comes down to multiple outcomes or outcomes with chaos involved. Let's say our super computer takes over the power grid. Oh no! It has all the control....until tornadoes or hurricanes or wildfires strike and trash the grid. It can tell precisely where the problems are, but getting out there and firing up the chainsaw to remove the trees, put up new poles, and re-string the wire is going to be more than it can handle.

"They'll just use robots."

Um, no. The problem with robots, particularly ones that would be built by computers, is that they tend to be overly specialized. And building a mechanical analog to a human is extremely difficult!

I also have some doubts about super-rapid self-advancement to super intelligence. Much of the leap from 4 year old to Einstein will have required some practical experience OR absolutely faultless simulations of countless mundane things.

In other words, there would seem to be some component of knowledge and understanding based on experiences impossible for a computer, which are instead "guessed" at.

I'm not overly concerned about a Super-AI taking over the world in the next 10 years. As Dart points out, robotics has a ways to go before it can address enough of the various things that humans do to keep this house of cards standing.

I will submit though, that the robotics do not necessarily have to arrive at a mechanical analog to humans to accomplish the goal of replacing us for functions the Super AI would decide are mission critical. If a Super AI has reached the point where it controls automated facilities for resource collection, transportation and manufacturing as needed for building robotics of various designs, it could conceivably churn out specialized modular bots that combine as needed for particular tasks. It may turn out that a lot of things can be done with non-humanoid designs.

As for AI dealing with multiple problems / multiple outcomes, the fields of Fuzzy Logic and possibly Genetic Algorithms, offer possible approaches to that. For learning from experience, that is what Case Based Reasoning does. Model based systems also inherently address that. The field of AI has a lot of different approaches, with various strengths and weaknesses. A truly advanced AI would most likely incorporate many of them to varying degrees as needed for various problems/tasks/goals.

We humans have been amassing experience and understanding of countless things, both mundane and not so mundane. In the process, we have been writing down our findings and procedures... and in many cases we even have them in computers / online. A Super AI would presumably have access to the online material, and could even eventually use OCR equipped robotics to acquire offline writings in old books, etc.

I think it is a matter of time and access to resources for a Super AI to become possible. If humanity manages to avoid destruction, or a loss of its high technology capability long enough, that is...

Neva gonna happen. People too influenced by movies letting their fear run wild.

They cant even make a convincing AI for conversation and look how long they've been adding knowledge to their database. An 8th grade dropout with less vocabulary is smarter and more conversationally adept than the smartest AI.

Didn't read the article did you, chum?

How many finished both parts before they commented?

Maybe I'm not human? In fact, prove I'm not AI...

Try this test-

AI test

I'd rather stay out of the technical aspect of AI, while I do have background on machine learning, it's a field that really didn't catch my interest/passion.

scanned the arctile and found something that caught my attention under the: The Road from ANI to AGI Why it is so hard.

Comparisons were made between humans interpreting an image and a computer. While it highlights the difficulty of a computer recognizing an image the way we do, there's another side to the coin that it failed to present.

Human perception is flawed. Sometimes, the patterns we see get in the way of objectively perceiving an image. For example, google 3d art where street artists draw on a flat surface such that when you view it from a certain angle it looks like a 3d image. It's the opposite. We're presented with a 2d image which we perceive as a 3d image.

IMO the path is not about recreating the human brain or making it like us. Perhaps a better sentient machine can be created founded on an entirely different concept.

I agree that making a human analog isn't necessarily the real measurement of a successful AI. For some applications it matters, such as something that is human interaction intensive like a nursing 'bot. For many other applications, the measure of success will be effectiveness regardless of how it is accomplished. A human-visible light spectrum camera vision system augmented by InfraRed and Thermal vision and radar/sonar/lidar has the potential to surpass unaugmented human vision, for instance.

I know I'm not a bot because I didn't want to be bothered clicking on a link and reading anything, while a bot would do it without hesitation.



The Jedi Master

Cogito ergo sum.

Unfortunately only available as audio,
but have a listen to this, from minute
34 to 38, about the distinction of
intelligence, and consciousness, as
regards AI.
http://www.cbc.ca/radio/ideas/when-man-becomes-god-yuval-harari-1.3800356?autoplay=true

There's more at 18:20, but better to listen to that
after the the later comment pointed out above.

This whole show is interesting, by the way, if
you have time - I usually don't unless I'm driving,
which is when I usually hear this show, as listening
takes up so much more time than reading the same
content. This guy has lots of interesting thoughts,
many similar to my own. I just sit around and think
of this stuff, but he's figured out how to make a
living out if presenting it...