Would you like to play a game... of global war... against AI?

[lightbulb27]

Military is doing so now... and didn't go well. They take little issues and blow them up to global war. Lets hope they can tune the AI's so they learn how to de-escalate as a priority.

Don't fire those Presidents yet...

We humans are so awsome. We've built things to model our own brains, but have not addressed the underlying issues of fear, security and trust.

Find a loving God with a great bailout plan yet?

That be fun as long as not connected to real armies

No

other

[GuyAnswersGirls123]

The AI would easily win if it understands the conditions and goal. We stand no chance.

We're not just talking about something that can easily defeat the world's greatest video gamers without breaking a sweat. We're talking about something that can defeat all the world's great video gamers simultaneously while playing against them all at once.

On the video covering escalation and first-strike tactics, I suspect the AI involved there wasn't trained with peace as its longest-term goal. Cooperative but retaliatory strategies tend to win out in the longest run in repeated games where cooperation.

https://www.youtube.com/embed/BOvAbjfJ0x0

A problem humans tend to make (as well as AI without proper training) is to fail to see life as analogous to a repeated set of games, not one game. Selfish and retaliatory tactics tend to excel in a single game, but not repeated games against the widest variety of opponents (this is when cooperative tactics tend to excel but also ones still capable of retaliating only to forgive if the opponent resumes cooperating).

[GuyAnswersGirls123]

[*] We're talking about something that can defeat all the world's [greatest] video gamers simultaneously while playing against them all at once.

Too many people tend to focus on just the capabilities at the level of processing one human being's worth of sensory input. An AI is that it can easily process hundreds, thousands, millions of human being's worth of sensory input at once.

[GuyAnswersGirls123]

[*] An AI is [so much more in] that it can easily process hundreds, thousands, millions of human being's worth of sensory input at once [while understanding all of it without dispersing the knowledge across millions of human beings with major collaborative problems and knowledge gaps between them].

[GuyAnswersGirls123]

So if we want AI to discover the value of peace and cooperation, it needs to be trained against repeated battles and wars against various nations and optimize for the longest-term success that way. Same if we want to train a human to appreciate the value of peace and cooperation.

We want a formidable person like a gentleman, able to punch back if punched, but never throwing the first punch. But also like a gentleman, favor kindness and courtesy when possible and reciprocated. That is the optimal strategy in life both at the level of the individual and that of nations.

[GuyAnswersGirls123]

Therefore, if AI is trained against optimizing for the longest-term prosperity of a nation against all possible future wars, then it will converge to something like tit-for-tat. It will favor cooperation and peace above all else, and if nations were governed by such AI, we could only expect everlasting world peace in the same way we can only expect a room full of nothing but gentleman who never throw the first punch to never get into a fistfight.

If, however, we train the AI in a way such as to optimize for a single war, and not all possible wars, and our nations were all governed by such AI, heaven help us as this will very likely extinct the human species because the optimal strategy in that type of singular and isolated case is always going to be a never-ending series of retaliations with no forgiveness and pre-emptive strikes.

[lightbulb27]

Exactly!

[goaded]

They should teach them tic-tac-toe. (Spoiler, but you've had 40 years to watch the film!)

https://www.youtube.com/embed/F7qOV8xonfY

[lightbulb27]

Good memory. Impressive old machine for the era.
yea, that was one conclusion of the test... there's no winning all out.

[FreyaRed]

Fearmongering. I assure you a top secret simulation results will never be published, it's just usual propaganda stuff for Pentagon to demands more money. Buhu, buhu the evil AI is upon you!!!

[lightbulb27]

when the humans begin to submit to the AI it's trouble... e. g. "The AI says there is no other play than to...". China's got AI, Russia will have it, all of Europe will have it. Some rogue hackers will get their own.
maybe in the end we are more secure since there is no answer... and it's just like the old west... guns pointed at each other ensures our safety?

[IslaTheWitch]

I would love to try my skills as a strategist in the war room if this was just a game

[Dongie]

My other is go hot against the traitors that have besieged the White House and not to mention, my peace of mind.

Eeh. . I'm waitibg