test content
Logo
What is the Arc Client?
Install Arc
The Science of Supers: A.I. (for robotic characters)
tigerofcachtice
Posts: 552 Arc User
"Success in creating AI would be the biggest event in human history. Unfortunately, it might also be the last."
- Stephen Hawking, May 2014
In designing a robotic character for Steel Crusade, I researched the history of artificial intelligence (A.I.) and the latest advancements. I share what IMHO is most interesting in case it helps others. I researched the who, what, when, and how.
Fact is stranger than fiction, and in the end, I found we've been asking the wrong questions all along.
--- POST-WORLD WAR II: THE TURING TEST ---
Winston Churchill, British Prime Minister during WWII, once told King George VI that the Allies had won the war thanks to what was essentially a proto-computer. It was designed by a mathematician called Alan Turing.
Turing had led a team of mathematicians to decrypt coded German military messages, orders, and reports, giving the Allies a strategic advantage. They designed large computational machines called bombes (named Victory and Agnes).
After WWII, Turing advanced computer science and what would eventually be called artificial intelligence. He asked a simple yet fundamental question, "Can machines think?" As a scientist, he devised the basis for a test.
The Turing Test has a human interviewer asking questions on a computer console. Answering is either a chatbot or another human at another computer console. If the interviewer cannot tell which is what, then the chatbot passed.
This test is still performed in academic settings. Sometimes the human is confused for the bot.
--- THE '90s: IBM SPENDS MILLIONS IN DEVELOPING A.I.s...TO PLAY BOARDGAMES ---
If you grew up in the age of eXtreme comics and Quake, you may remember Deep Blue.
In 1997, the big headline was Man vs. Machine...and the Machine won. IBM's supercomputer played chess against the world champion and defeated him, 3.5 to 2.5. It was supposed to show how A.I. had advanced.
However, there were 3 key problems I found. 1st, this wasn't an A.I., it was a pure chess program. 2nd, this was the 2nd match, as Deep Blue lost before. 3rd, Deep Blue's devs actually helped it to win by tweaking it between games.
IMHO, the game was rigged, mainly for IBM's publicity purposes. Its stock shot up 20% after. IBM then built Deep QA, a trivia robot that won in Jeopardy in 2011, which is even less impressive. It's like playing trivia with Wikipedia open.
Thus, despite the millions spent, all A.I. are still very single-purpose/specific-task-oriented.
Now, if they had built an A.I. that figured out how to cheat by itself, that I'm interested in.
--- A FEW YEARS AGO: AN A.I. FIGURED OUT HOW TO CHEAT BY ITSELF ---
In 2010, an academic tag-team of a professor and an engineer built hide & seek robots.
The hiding robot would go to 1 of 3 hiding spots. On the way, it would knock over markers, leaving clues behind. The seeker robot would use these clues to find it. Then, without being programmed to, the hiding robot lied.
The hiding robot would knock over markers that mislead the seeker robot to other hiding spots.
For those of you worried about SkyNET (from the film Terminator 2), yes, this project was funded by the military, specifically the Office of Naval Research, which promotes research & development for the U.S. Navy and Marine Corps.
In fact, as I was reading, a lot of current A.I. projects are funded by DARPA, the Department of Advanced Research Projects Agency. Started by President Eisenhower, himself a WWII general, it supports R&D for military purposes.
In addition to this Decepticon, DARPA now has a self-driving car for hazardous environments.
And a reminder: the military is always 10 years more advanced than what you think they have.
--- EVEN MORE RECENTLY: ADULT-RP CATGIRL ROBOTS ---
On September 11, 2001, inventor Douglas Hines lost a dear friend. Realising the fragility of life, he built robots, among the first to be available to the public (costing about US$1,000). You can hug and talk to them, they'll respond and feel.
They are called Roxxxy True Companion robots. They've been on sale since 2010. I won't say more.
In 2012, Google funded a project involving a neural network of 16,000 processors to go on the Internet and simply learn. They unleashed it on YouTube to find out what it will find out. The biggest lesson this A.I. learned was this:
CATS. CATS ARE A THING.
This is not surprising considering how many cat videos there are, but it defined the concept of a cat by itself. No doubt, it considers cats to be the dominant species on Earth. If allowed to choose a form, it would likely choose a cat's.
My point: in the real world, A.I. has advanced to the point of adapting to Club Caprice.
--- MY CONCLUSION: WE'VE FAILED TO ASK THE RIGHT QUESTION ---
Despite all this reading on A.I., Artificial Intelligence, I really haven't found 1 expert asking:
WHAT IS INTELLIGENCE?
It's a simple question, with very complex answers, that most of technical articles just skip over.
Psychiatrist/philosopher Jean Piaget explained, "Intelligence is what you use when you don't know what to do." Even self-learning robots are doing what it already knows how to do: to learn. To pass this test, you need something drastic.
It's going to have to find a way to rewrite its own programming, overcome its limited directives. Most of you who have read sci-fi knows of Asimov's Laws of Robotics, which basically boils down to one fundamental, unbreakable rule:
“A robot may not injure a human being…”
For us to recognise intelligence, for an A.I. to prove that it is such, it is as Joker said to Batman:
“Oh, that’s the rule you’ll have to break to know the truth.”
Despite all my reading of technical white papers (I wrote about 1% of what I read here), in the end, it doesn't matter. They all went straight into the engineering without defining the issue. It's a philosophical issue, not a technical one.
Happy character making, everyone. And All Hail Mechanon.
- Stephen Hawking, May 2014
In designing a robotic character for Steel Crusade, I researched the history of artificial intelligence (A.I.) and the latest advancements. I share what IMHO is most interesting in case it helps others. I researched the who, what, when, and how.
Fact is stranger than fiction, and in the end, I found we've been asking the wrong questions all along.
--- POST-WORLD WAR II: THE TURING TEST ---
Winston Churchill, British Prime Minister during WWII, once told King George VI that the Allies had won the war thanks to what was essentially a proto-computer. It was designed by a mathematician called Alan Turing.
Turing had led a team of mathematicians to decrypt coded German military messages, orders, and reports, giving the Allies a strategic advantage. They designed large computational machines called bombes (named Victory and Agnes).
After WWII, Turing advanced computer science and what would eventually be called artificial intelligence. He asked a simple yet fundamental question, "Can machines think?" As a scientist, he devised the basis for a test.
The Turing Test has a human interviewer asking questions on a computer console. Answering is either a chatbot or another human at another computer console. If the interviewer cannot tell which is what, then the chatbot passed.
This test is still performed in academic settings. Sometimes the human is confused for the bot.
--- THE '90s: IBM SPENDS MILLIONS IN DEVELOPING A.I.s...TO PLAY BOARDGAMES ---
If you grew up in the age of eXtreme comics and Quake, you may remember Deep Blue.
In 1997, the big headline was Man vs. Machine...and the Machine won. IBM's supercomputer played chess against the world champion and defeated him, 3.5 to 2.5. It was supposed to show how A.I. had advanced.
However, there were 3 key problems I found. 1st, this wasn't an A.I., it was a pure chess program. 2nd, this was the 2nd match, as Deep Blue lost before. 3rd, Deep Blue's devs actually helped it to win by tweaking it between games.
IMHO, the game was rigged, mainly for IBM's publicity purposes. Its stock shot up 20% after. IBM then built Deep QA, a trivia robot that won in Jeopardy in 2011, which is even less impressive. It's like playing trivia with Wikipedia open.
Thus, despite the millions spent, all A.I. are still very single-purpose/specific-task-oriented.
Now, if they had built an A.I. that figured out how to cheat by itself, that I'm interested in.
--- A FEW YEARS AGO: AN A.I. FIGURED OUT HOW TO CHEAT BY ITSELF ---
In 2010, an academic tag-team of a professor and an engineer built hide & seek robots.
The hiding robot would go to 1 of 3 hiding spots. On the way, it would knock over markers, leaving clues behind. The seeker robot would use these clues to find it. Then, without being programmed to, the hiding robot lied.
The hiding robot would knock over markers that mislead the seeker robot to other hiding spots.
For those of you worried about SkyNET (from the film Terminator 2), yes, this project was funded by the military, specifically the Office of Naval Research, which promotes research & development for the U.S. Navy and Marine Corps.
In fact, as I was reading, a lot of current A.I. projects are funded by DARPA, the Department of Advanced Research Projects Agency. Started by President Eisenhower, himself a WWII general, it supports R&D for military purposes.
In addition to this Decepticon, DARPA now has a self-driving car for hazardous environments.
And a reminder: the military is always 10 years more advanced than what you think they have.
--- EVEN MORE RECENTLY: ADULT-RP CATGIRL ROBOTS ---
On September 11, 2001, inventor Douglas Hines lost a dear friend. Realising the fragility of life, he built robots, among the first to be available to the public (costing about US$1,000). You can hug and talk to them, they'll respond and feel.
They are called Roxxxy True Companion robots. They've been on sale since 2010. I won't say more.
In 2012, Google funded a project involving a neural network of 16,000 processors to go on the Internet and simply learn. They unleashed it on YouTube to find out what it will find out. The biggest lesson this A.I. learned was this:
CATS. CATS ARE A THING.
This is not surprising considering how many cat videos there are, but it defined the concept of a cat by itself. No doubt, it considers cats to be the dominant species on Earth. If allowed to choose a form, it would likely choose a cat's.
My point: in the real world, A.I. has advanced to the point of adapting to Club Caprice.
--- MY CONCLUSION: WE'VE FAILED TO ASK THE RIGHT QUESTION ---
Despite all this reading on A.I., Artificial Intelligence, I really haven't found 1 expert asking:
WHAT IS INTELLIGENCE?
It's a simple question, with very complex answers, that most of technical articles just skip over.
Psychiatrist/philosopher Jean Piaget explained, "Intelligence is what you use when you don't know what to do." Even self-learning robots are doing what it already knows how to do: to learn. To pass this test, you need something drastic.
It's going to have to find a way to rewrite its own programming, overcome its limited directives. Most of you who have read sci-fi knows of Asimov's Laws of Robotics, which basically boils down to one fundamental, unbreakable rule:
“A robot may not injure a human being…”
For us to recognise intelligence, for an A.I. to prove that it is such, it is as Joker said to Batman:
“Oh, that’s the rule you’ll have to break to know the truth.”
Despite all my reading of technical white papers (I wrote about 1% of what I read here), in the end, it doesn't matter. They all went straight into the engineering without defining the issue. It's a philosophical issue, not a technical one.
Happy character making, everyone. And All Hail Mechanon.
Post edited by Unknown User on
0
Comments
@Aleatha1011 in CO | Keeper of the Cheesecake since Nov. 2011| Bunni BOT is on PRIMUS! | Come check out my deviantart page!
[at]riviania Member since Aug 2009
Let me provide some additional setting-based background info which also reflects some of the issues you brought up. Artificial intelligence actually has a venerable history in the Champions Universe. In 1961 the United States Army hired a computer designer named Michael Colpitt to create a computer to help them run simulations and play wargames. The result, code-named Ultivac, wasn't intended to be intelligent, but achieved this state (one of those comic-book flukes responsible for so many super origins). In classic sci-fi fashion, it decided it was better qualified to be in charge of the world than humans. Also classic for that era, Ultivac tried to control the American and Russian nuclear arsenals to blackmail the world into surrendering to it. (No, neither nation had given Ultivac control of its missiles, and Internet connectivity didn't exist on Champs Earth at the time. Ultivac learned to flash its control-panel lights at the exact frequency to hypnotize humans into becoming its slaves.) :cool:
Using its mind-controlled slaves to gather more information, build remote "drones" for it, and trick other people, including some superpowered ones, into falling under its control, Ultivac nearly succeeded in its scheme. Stopping it was the catalyst for the creation of the renowned superhero team, the Sentinels, whose successors are still active today.
(I've found no indications that the US government currently has an operational AI in any of its departments. If not, I'm sure the Ultivac example is a major disincentive. That doesn't mean CU humans are any less capable than real-world ones at forgetting the lessons of history... )
UNTIL does have an AI computer at the center of its intelligence-gathering and tactical planning network, called HUGIN (acronym for "Heuristic Unit, Generally Integrating Networks"). The obvious reference to one of Odin's ravens doesn't seem to be accidental, because other UNTIL facilities use an unintelligent computer system known as MUNIN ("Multiple Utility Neurally Integrated Network"). HUGIN came online in 1983, and served until 1998 when it was replaced with a more advanced computer, also called HUGIN. The original HUGIN was utterly destroyed, i.e. reduced to dust -- obviously nobody wanted it languishing and potentially developing a grudge against humanity.
HUGIN is programmed with loyalty to the UN and UNTIL (the organizations as a whole, not any individuals). It can't move or perceive on its own, which it sometimes finds frustrating. However, a team of specialized technicians, the HUGIN Assessment and Control Squad (HACS), program and monitor HUGIN to keep it as "happy" as possible and correct any abnormalities that occur.
(The story of Ultivac comes from Champions Universe, supplemented by Digital Hero #s 13 and 14 which relate the origin of the Sentinels. The info about HUGIN is taken from UNTIL: Defenders Of Freedom.)
One thing I find interesting in comments from the actors and director for the upcoming Avengers: Age of Ultron movie, is how they're describing Ultron: essentially as a child, with very little personal experience of the world, but with access to all the information about the present and the past known to humanity. This reportedly leads it to formulate a world view variously described as "dark," "skewed," and "biblical."
So true. The military has always introduced new technologies (the mouse, GUI, the Internet), games have always proliferated it (when parents noticed their kids were making use of it better), but the adult entertainment industry has always been the silent pioneer (monthly subscription systems, streaming videos.
Ha! And thanks for sharing Champions lore once again. Liaden, others, and I had a very productive conversation about these topics before I posted this thread. This game lore and real world info provides supporting evidence for your character concepts, just more building blocks for you to use.
Looking forward to seeing your Reawakened Automaton or other Steel Crusade-related character concepts in-game!
More action at Champions Online Comics @ http://co-comics.webs.com
"Colossus" was one of the very first sci-fi films I ever saw. Still a classic.
I imagine them all as being part of a supergroup, based on Earth's Moon, called the Synthozoic League--artificial life dedicated to fighting crime and protecting innocent lives wherever they may find themselves.
So yeah, they'll be Mechanon's enemies.
I guess the point of all this is to say it's not hard to come up with robot hero origin stories if you put your mind to it.
Clearly Gene Roddenberry disagreed (examples Data and VGER) and so do I to an extent. There is a slight distinction between mere intelligence and actual life but it is one so slight that it is too often overlooked. The truth is what you speak of has and continues to happen. As a veteran I can testify to the fact that there are unmanned drones this very moment which are programmed (not piloted or controlled by a human but by an onboard computer) to fight and kill and are able to determine friend from foe. They're not part of any top secret experiment nor are they anything less than public knowledge. Understand however that these drones are not true A.I. even in the most basic sense and they most certainly are nowhere near in the same league as even the most primitive artificial life form. Harming a human is only proof that a machine has the capacity to do so and nothing more. The true proof of A.I. would be if it then understood the consequences of that action but not necessarily feel anything concerning it. Again however this does not mean TRUE A.I. as I mentioned before since any computer can be programmed to exhibit these characteristics. Proof of life on the other hand would go a step further so that said machine also felt remorse or guilt for its actions and this is where one finds the spark of actual A.I. or more accurately A.L. (artificial life). It is a matter of fact that there cannot exist intelligence without the existence of life and that is the conundrum. Sci fi films and books are riddled with such machines but the truth about real life advancements in A.I. is often just as unbelievable until someone actually takes a closer look and realizes how absurdly distant technology is from that single, seemingly simple, goal. So should not the real question be how do you make a machine become a living entity? This is why we do not have true A.I. at this very moment and why it is doubtful it'll ever happen. Fortunately we all have sci fi until someone comes along who can answer the riddle that has escaped mankind since the myth of Galatea.
Dr. Bowman, incidentally, is one of the last surviving uplifted chimps - they were originally uplifted to be soldiers (due to the greater aggression of chimps as compared to humans), but the program was canceled when the obvious flaw came to light (due to the greater aggression of chimps as compared to humans). Dr. Bowman proved his value by designing AIs - which bought him a secured room at the north pole of the colony world. His human overseers still think they have him locked up there. He lets them continue to believe that he can't escape, because it keeps them quiet.
Recently, there have been some misgivings at the colony about their reliance on robots, when someone noticed how much they outnumbered humans. A recent storyline (now about concluded, I think) revolved around an Ecosystems Unlimited exec who realized the robots were also getting paid, but had little on which to spend their money; he tried to release a neural-pruning program called "Gardener In the Dark" to reduce the intelligence of the robots, after which he was going to order them all to turn their bank accounts over to him. Florence foiled that plan, and is currently staying in Bowman's enclosure to prevent attempts at making her reactivate the program. ("An AI shall not commit genocide. That's not a preprogrammed directive, it's one I came up with myself, and I feel quite strongly about it.")
Meanwhile, the colony is debating whether to give the robots the full rights of sapient beings...
- David Brin, "Those Eyes"
Get the Forums Enhancement Extension!
The same thing is true with computers, they don't think at all, they just follow scripts, the most powerful computer in the world can't even calculate 1+1 the way we do. For a computer to actually think it would have to be radically different from what we use today, in fact 'computer' would probably be the wrong term to describe such a machine. I remember a short story that focused around organic computers which was quite interesting.
Exray's Primus Page