[Discussion] Implications of an AI society

[Kax]

An interesting tidbit for this thinking of the cost of computing for AI: the present cost of a petaflop machine is derived from the power cost of 1kW/gigaflop.

So, around $1,000,000 a year just in power at this TL.

[johndallman]

Originally Posted by Kax

An interesting tidbit for this thinking of the cost of computing for AI: the present cost of a petaflop machine is derived from the power cost of 1kW/megaflop.

So, around $1,000,000 a year just in power at this TL.

I think you have some zeroes misplaced. This laptop, running at 2.9GHz, gets quite a few megaflops, and is using a lot less than 1kW.

[cptbutton]

Originally Posted by CeeDub

Somebody once wrote that "it is difficult to imagine an intelligence that thinks as well as a human, but not like a human."

"Write me a story about a creature that thinks as well as a man, but not like a man." - John W. Campbell, according to Theodore Sturgeon

[jeff_wilson]

Originally Posted by cptbutton

"Write me a story about a creature that thinks as well as a man, but not like a man." - John W. Campbell, according to Theodore Sturgeon

Meanwhile, the whole point of an AI is to make something that thinks like a man, or at least like a man says a man should think. This amounts to something like a butler, because the man was typically influenced by Victorian or Edwardian intellectuals.

[AlexanderHowl]

Originally Posted by jeff_wilson

Okay, I think I see now why you are so harshly against automation. I'm a machine learning researcher myself and I hope that I can explain that while there is rather a lot of military spending on automation research and implementation the aim is to cut down on human suffering and death by the letting non-living machinery do some of the work that has to be done in war zones and other threatened places so fewer fragile human bodies need to be sent there.

I have also worked with machine learning (primarily agent based modeling, complex adaptive systems, and neural nets), and I served in the US military. Both experiences tell me that the automation of combat will only encourage politicians to get into more wars because the cost (for their county) is merely monetary rather than in human lives. What it ends up meaning is that more civilians die because more and more of the process of war is automated and, since they are not risking the lives of their soldiers, politicians do not care about the civilian deaths caused by automated weapon systems. We already see a callous disregard for civilian casualties with drone combat, it will only get orders of magnitude worse with automated combat, as robots are even worse at telling civilians from combatants than human.

[hal]

Originally Posted by AlexanderHowl

I have also worked with machine learning (primarily agent based modeling, complex adaptive systems, and neural nets), and I served in the US military. Both experiences tell me that the automation of combat will only encourage politicians to get into more wars because the cost (for their county) is merely monetary rather than in human lives. What it ends up meaning is that more civilians die because more and more of the process of war is automated and, since they are not risking the lives of their soldiers, politicians do not care about the civilian deaths caused by automated weapon systems. We already see a callous disregard for civilian casualties with drone combat, it will only get orders of magnitude worse with automated combat, as robots are even worse at telling civilians from combatants than human.

On one hand, I can agree with certain sentiments you've outlined here - but there is a hard limit for politicians in general...

If enough people become enraged - they become part of the political cost for a politician. So, wanton indiscriminate slaughter of civilians will have a cost, just not an immediate cost. How often these days, are we seeing people being excoriated by the media or the public in general, for something they said 10 years ago on twitter or on a self-published blog?

In the end, life will change somewhat, but mankind will always adapt to it and find ways to circumvent things (until they can't). Greed seems to be a part of man's make up, as is altruism, as are many traits - they just don't all seem to occur in one individual. ;)

AI's in general? Have to interact with humans initially. Initially, they will be designed to perform a service on behalf of their owner. Over time, they WILL perhaps become more numerous, and over time, may be designed to become more complex - more so that there will be a difference between the first of the AIs and the last of the AIs that would be akin to looking at a 1930's car and comparing/contrasting it against a model 2018 vehicle.

Man has always been trying to get the upper hand in any conflict. He's always invented better and bigger weapons. Should it be any surprise that we'll have AIs that are geared towards war? What will happen when we get AIs geared towards arguing court cases? Law enforcement? Medical technology? Heck, look at the issues surrounding Autonomous vehicles - and what it may do to the trucking industry. Point is, all of this is happening now. And as the one person asked "What is the point of people in a TL 10 society?" (or words to that effect). It is kind of hard to ignore the elephant in the living room and any attempts to write "near future" backstories for campaigns is going to have to find a way around it if they want to be reasonably well grounded in reality, or they have to ignore it much like we ignore TV shows supposedly aboard a ship in space in zero G, but they're walking on the floor as if in a gravity well. Hit the "I believe" suspension of disbelief button and move on. ;)

[AlexanderHowl]

I generally have society reacting against massive changes in technology and adopting safetech policies in relation to computer technology in my future settings. I tend to have the public Internet being banned in the past due to being a festering pit of child pornography, hate mongering, and white supremacy (I do research on criminal behavior on the Dark Web, it is not pretty), so characters get their information much like they did in the 1980s (though computer databases in libraries exist and people still have access to highly advanced gaming consoles). I also tend to have AI being restricted to biological computers or neural networks, the products of hardware instead of software, and have them interface with characters through remote controlled drones since there is not a public Internet.

[hal]

Just as a thought? Bringing this back to the AI and society AND keeping the topic open a little.

What if...

Just as there is the Dark Net, the SAI have their own network that they have manage to either find, or reroute so that only Artificial Intelligences can get to? The numbers may not be all that large in volume, but where they make up for it is in their own "enclave" of individual AI's interacting with AIs. Sort of like that one scene with the hotel AI in Altered Carbon?

But more to the point - there could very easily become a fragmented Internet system largely because in the beginning, there were a multitude of private bulletin board networks, that got swallowed by the wild days of the internet. As things became more regulated, the net begins to divide into the Web and the Dark Net. Then it fragments further into purpose built systems. Thing is, they are all accessible via your network service provider, but the URLs are different or you need to know which ones to access for what you want. To the unsuspecting world at large, no one knows what lies beneath in the specialty networks.

Just a thought.

[jeff_wilson]

Originally Posted by AlexanderHowl

We already see a callous disregard for civilian casualties with drone combat,

It's way less callous than when chemical weapons are used, or when smallpox blankets were given, or when firebombing or just plain arson is used.

Originally Posted by AlexanderHowl

as robots are even worse at telling civilians from combatants than human.

...which is why we need to keep improving the robots, unless you think we can improve the humans faster?

[Daigoro]

SAI piloted warcraft will presumably be better at distinguishing valid targets than dumb AI, but if they're emancipated SAI they might choose not to be involved in combat. Even without Asimovian laws, they might make the calculated decision that they don't wish to harm any human life- and that could be out of respect, indifference to our existence, or condescension to how soft and pitiful our delicate fleshy bits are.