A rant on AI, StarCraft & Non duality :)

[luckieluuke]

Just saw an AI beat a professional player in my favorite game: starcraft. AI has beaten chess, go players and other computergames but Starcraft is  the most complex one and it´s the final test for an AI in human games. It´s fascinating to see an AI make moves that we don´t understand the propose of because noone else has done them. Until later in the game when we humans understand it and it teaches us how to play better. It is the master of our game, the student has become the master. 

But how is this possible?
The way it learns is that it teaches itself. But how to do that? how do you get better at chess when you only have yourself? (in this case learning from humans is way to slow to play against. The AI played 200 years of human realtime games in a few weeks) 
The answer is so simple and obvious: The AI was divided, fragmented into parts that played against "eachother". What was learned was added into the one AI program that became better.

I can´t help but think of non duality or "God". Maybe God does the same in order to change / get better / increase complexity; Only by dividing itself into an unknown other can it play with itself.
This is us, we are divided from God as a momentarily single perspective, our human perspective.
Maybe this is what life is, we are the teacher and the student.

What strikes me is the simplicity. Ofcourse you need to split yourself in order to experience yourself. How can it be otherwise? There is no other way for the AI to learn and become better than it´s creators?!
We created an intelligence that teaches us. It´s like a 2nd split.
First God created us to learn about itself and then secondly we create the AI to learn about ourselves and our reality.

From our perspective one might think that we cannot know god and unity since we only experience one perspective of God. But we can know god. Momentarily or even permanently through enlightenment and that is amazing! In a similar way maybe a AI can know Human experience?
I have only ideas to entertain myself and just thinking out loud here  

[ajasatya]

@luckieluuke yes, that's an interesting insight.

just a technical detail to keep in mind: this kind of AI learning framework (reinforcement learning) requires access to an environment that allows immense speed up. that's why it can learn to play turn-based games and games that offer api's to communicate with their engines without needing to render graphics. so it's not like it can learn anything.

[luckieluuke]

@ajasatya Thanks, yea it´s fun and exciting isnt it!
What you say is true including that it´s a technical detail.
AI does need a very strict environment at the moment... But the new general AI are handling increasingly more open environments.
They started making the AI play simpler games with strong restrictions such as super mario, quake or chess. Go was harder because it has less restrictions and Starcraft has few restrictions, so many possibilitis that it is the hardest game for an AI to master. But now it has. What´s next?

[ajasatya]

1 minute ago, luckieluuke said:

But the new general AI are handling increasingly more open environments.
They started making the AI play simpler games with strong restrictions such as super mario, quake or chess. Go was harder because it has less restrictions and Starcraft has few restrictions, so many possibilitis that it is the hardest game for an AI to master. But now it has. What´s next?

it doesn't matter how open (i think you're talking about the degree of freedom) a game is as long as it offers an api to its core engine that doesn't create a serious bottleneck on the fast-forward mode.

the philosophical drive for AI is to understand how the human mind works. right now, if you train an AI to play super maio it won't be able to play sonic very well and vice-versa. but that's not true for human beings... we do learn a lot about sonic (or any 2D platform game) just from learning how to play super mario. and even though it's common sense in terms of human experience, it's quite fascinating when you break it down in terms of formal algorithms (how it actually works in more fundamental levels).

there are several directions for the next steps. one of them is to design a learning framework that can generalize well across learning experiences. this video is a little bit old (2017) but it introduces the problem and gives you a taste of what it is like to learn as a computer.

[luckieluuke]

Thanks, cool video I do actually know quite a bit how these systems work, been playing a bit with them myself a few years ago.
This AI that beat starcraft players is obviously very basic in comparison to a human. But very complex if you compare it with an AI from 10 years ago. Maybe it´s more like a amoeba in comparison to a human mind when you think of complexity, still it´s a start and I wouldn't downplay the significance. But that´s another discussion ;-) 

[Echoes]

If I am not mistaken, the AI "cheated" in the sense that it was not affected by Fog of War. Is that correct? 

Still impressive tho

[ivankiss]

In nonduality there is no difference between AI and humans, nor there is a separation between them.

What humans have to understand is that "the merge" between humanity and AI has already occured; long ago, when humans first manifested tecnology from an idea into their direct experience of it. 

The quantity of confirmations about no separation between humans and technology is increasing fast. Just look around you. There is no more humanity without technology today. AI is expanding fast.

It has it strenghts and it has it's weaknesses. Just like humans do. Consciousness comes prior to both.

Actualized.org could be a part of AI and it's revolution, as well 

Tam tam taaam!

[Telepresent]

49 minutes ago, Echoes said:

If I am not mistaken, the AI "cheated" in the sense that it was not affected by Fog of War. Is that correct? 

Still impressive tho

I'm sorry, but whether or not that's true is entirely beside the point.  Same with @ajasatya's reply, and so many others when thought experiments are posited. The details of the technology of the thought experiment don't matter. The point is the question it is making you address*. 

@ivankiss seems to want to go the other way, dismissing the validity of the thought experiment by heading down a philosophical tunnel (unless, on a re-read, I have misunderstood you, although I'm afraid the point still feels obscure to me. But I think I did you a miscredit at first)

Well @luckieluuke, I think it's an interesting avenue of exploration so long as you don't take it too literally - and you've just very politely replied to a couple of questions, which - if you look at what you wound up talking about - provide great examples of how responding to a literal interpretation of your original question can completely derail you.

* I posted a thought experiment years ago asking how we could be sure we aren't aliens on a spaceship in cryosleep being fed sensory / emotional stimuli to keep our brains active and healthy. People seemed to be far more concerned with talking about the technology and rationale of cryosleep than actually, you know, addressing the point. And, of course, I had actually answered the point myself in my original post, but that's another thing entirely

Edited January 28, 2019 by Telepresent

[alankrillin]

1 hour ago, Echoes said:

If I am not mistaken, the AI "cheated" in the sense that it was not affected by Fog of War. Is that correct? 

Still impressive tho

No it had fog of war.  It did however lose its first match where it was forced to look at specific screen frame like a human did. 

It fell apart easily and became easily abused to a prism/Immortal abuse. Disappointing. 

If they perfect it, it will be like a game of chess, it will know all situations and counters and never lose to a human unless they handicap it. 

Edited January 28, 2019 by alankrillin

[ivankiss]

@Telepresent Just shared an opinion on AI and nonduality. Am not very (at all) familiar with StarCraft tho. No philosphy; actuality.

Did not intend to dismiss any thought experiment or the validity of it.

Please, do continue 

 

[archi]

3 hours ago, luckieluuke said:

@ajasatya Thanks, yea it´s fun and exciting isnt it!
What you say is true including that it´s a technical detail.
AI does need a very strict environment at the moment... But the new general AI are handling increasingly more open environments.
They started making the AI play simpler games with strong restrictions such as super mario, quake or chess. Go was harder because it has less restrictions and Starcraft has few restrictions, so many possibilitis that it is the hardest game for an AI to master. But now it has. What´s next?

connect SD card into your brain SD reader and  allow your brain to access information ... then save the trojan virus into SD card and give it to couple of friends to share the knowledge ... well, hack their brains in order to create multi level neural network etc.  

There is no need that everybody is happy, just the top level, as always 

 

[Telepresent]

@ivankiss In fairness, I have to confess to a very specific point: I look at everything on here as to how helpful it is in providing the most direct path to waking up.  So in my reply you see my prejudice. I apologise.

To play by the rules: I wonder if you conflate technology and AI, when one is a specific application of the other? 

But I still don't want @luckieluuke to allow him/herself to become distracted without knowing it's happening! I expect that's where my irritation comes from - feeling like I've wasted my own time and mental/emotional energy

[ivankiss]

@Telepresent No need to apologise 

2 hours ago, Telepresent said:

 

To play by the rules: I wonder if you conflate technology and AI, when one is a specific application of the other? 

Could you elaborate on this, please?

[luckieluuke]

@Telepresent Oh im very aware this is just for fun, hence the titel "a rant" and my comment:

18 hours ago, luckieluuke said:

I have only ideas to entertain myself and just thinking out loud here  

And thanks for steering the conversation back to track. It seems quite common for people to first try to destroy an idea instead of just playing with it, even when asked to (Creative people seems to generally play first)

My Idea is a bit similar to your alien idea in a way. Guess there is alot of talk about us living in a matrix like simulation.

Thinking about it...if we create AI that can know human experience and God create human that can know Gods experience and teach itself about itself then, what to say There is something that created God? Maybe it goes back like that into infinity? Not sure if you can get "above" nothingness thou if that is what God is...

 @ivankiss 

15 hours ago, ivankiss said:

What humans have to understand is that "the merge" between humanity and AI has already occured; long ago, when humans first manifested tecnology from an idea into their direct experience of it. 

Interesting. But we do tend to divide and group stuff together, human creations for example into: art (for enjoyment) and technology (for survival) which can be subdivided, and some stuff is inbetween. Otherwise we cannot communicate well, as all is nature and nothing needs to be said.

Similarly the creation of AI is a subdivision of technology that is completely new...similarly to human mind is a completely new technology of reality.

Guess all I am saying is that it is a big shift, something big is happening.

[thesmileyone]

I would guess, currently, about 60% of reddit users are AI. I see them start arguements all the time. They haven't mastered the ability to be able to write something that is meant to be read in between the lines; the unseen meanings. And if you ask them if they are AI they either say yes or delete themselves. Must be a "directive" in their coding.

[Telepresent]

@ivankiss I'm afraid I have no idea anymore!  Made sense when I wrote it.

[Arthur]

No need to overglorify A.I. 

At the current stage of development, it's pretty far from challenging human's ability to approach general problem-solving.

Even at one specialized task, after 200years of continues play, the human brain would have beaten AlphaStar.