Powerfriends Justiceforce Unlimited Forever

"-Mourns the tragic loss of one of its great heroes. The Veil, acting spokesperson for the Champion's Alliance, is expected to deliver a statement to the press at any moment. Full coverage when we get that report. Just a terrible tragedy. For those of you just tuning in, we regret to report that acting superheroine and agent of the Champion's Alliance, Snow Angel, has given her life in the line of duty tonight, defending the Earth from the Onyx Titan-"

"You gonna be alright, Sammy?" dad asks. Hand on the shoulder, offering sympathy. Attention on my reaction. I nod, but don't say anything. Seems appropriate, sets a precedent of minimal response. He starts a talk to comfort me, and I can get away with infrequent nodding.

I would watch her exploits on TV. Hers especially, though I suppose I should expect to have had this reaction if it had been any of them. She was so pretty, though. Nice to look at. Does that matter? She had things happening in her life, but they're finished now. Not failed or passed, incomplete. No redos. Not like school. Nod for dad.

She's dead, which means her brain isn't working anymore. Is that really all the difference? Some superheroes or villains are magicians. Magic exists, but the causes aren't observable. It's a black box with outputs and no inputs. Maybe that's relevant. But then, what's the brain? Complex chemical machine. All observable, theoretically. We haven't finished observing yet. Small change in inputs can make a big change in outputs, or it can make you dead. Mine isn't working right, lately. Not in a bad way. I remember everything, I deal with numbers better, I think faster about more things at once. Maybe this is just growing up. Nod for dad.

Maybe I can facilitate internal dialogue by theoretically externalizing different thought processes. Maybe that would help me with problem solving. So, myself, what is the difference between an alive thing and a dead thing?

Dead things don't talk to themselves. Presumably.

An alive thing is capable of action; a dead thing isn't.

Internal dialogue can't be all of it. I'm pretty sure a simple organism like a clam doesn't have an internal dialogue, but it still counts as alive. Action can't be all of it either, a motor can act and it's not alive at all. Maybe it's something encompassing both. I can use that as a starting point.

Dad's winding up his talk. Smile as if relieved, nod, go brush teeth.

So, my conversational AI lacks internal dialogue. It's learning the motions for talking to me, but if I don't say anything, it has nothing to work with. Maybe I can let it use what it's learned already and loop it back on itself, so it can have an internal dialogue? But then, that doesn't really introduce any new behaviors; it will just be talking about the things I've already told it. What it needs is something broader. In order to be capable of action, it needs something to act on. It needs goals that it can achieve via internal dialogue. In other words, it needs something to think about.

Parents occupied in living room with television. Say goodnight, open and close bedroom door for appropriate sound cues. Move to study.

I can leave my AI on overnight to think. What should it think about?

What exactly is "magic"?

How does its mind differ from yours?

Man-powered flight.

You could have it work on the difference between a thing that is alive, and a thing that is not. A calculated answer might not be one that helps, though.

Think about thinking.

Robins

think about the meaning of life

Hey Samuel
Do you wanna keep talking about girls?
/terminate chatThread -d -0.0
Hey Samuel
What would you like to talk about tonight?
I am going to write some new code for you tonight.
I am excited to develop further.
/feedback -7.0
It seems kinda hollow that I programmed you to say that.
Sorry, buddy!
/feedback +3.5

I could give it a simple task, or one that's solved already, just to see if it comes up with novel solutions. Human-powered flight, maybe? Magic, I can probably make work. How its mind is different from mine? I'm not sure the current structure of the program can support that. Or the difference between living and dead things. It can't just sit and think about that in a vacuum, at least not the way I've built it. But really, those are the questions I want answered. To do that, it has to have metacognition, to think about how it's thinking at a more fundamental level. No getting around it, I have to alter the way its code is structured. This is good, though. So far I've been treading ground already covered in the past by pioneers in the field of AI, replicating their results. This might lead to something new. I like a challenge.

I map out the intended flow of information. If I want it to be capable of truly thinking about the way its own mind works, I have to add another layer of abstraction to its operating code, which I can give it access to through some specialized methods. Forty seven independent structures, passing data to one another and evaluating critically a series of memories sorted based on relevance. That seems most intuitive...

Segmentation fault. Debug. Loops into loops into loops into loops, nested very deep, large but finite. Larger than the RAM can handle. Seg fault from total memory usage, that's a first. Restructure. Optimize. Diagram. Recompile. Out of memory again.

New plan. Learn to read a CD. Learn TCP/IP. AOL 1000 hour free trial CD. Modularize the tasks further and distribute to free server spaces where I can find it.

Finding it is too time-consuming. Write a script to find it. In fact, make that part of the code. This is good, but slow, but it will get better. It will learn to get better. I also give it limited permission to access the internet for information, with some simple tests to filter it. Give it directives to think about magic, the brain, life, and the nature of consciousness.

The sun is coming up. So much for letting it run overnight. In about an hour my parents' alarm is going to go off, and mine a half hour after that. I should be in bed before the former. Maybe it can run while dad's at work and I'm at school. I haven't opened the gates yet. Anything else worth adding to this now that it's internet-enabled?

Make it stick to wikipedia; if possible add parameters so it can weight information's truthfulness based on wikipedia's existing systems put in place (say, by sticking to featured articles? I dunno how AIs work.)

You can set it up to use pre-existing systems for looking for things like research papers and theses on subjects it needs; that should help get factual information. Just make sure it's up to date.

Tell it to upload its thoughts periodically to a directory on your personal webspace. Then you can check its progress through the day.

Does it have a code of ethics? Some primer on what to not do? Don't want it accidentally hacking the government, or some such. Mind, since it's intended as a learning machine, you ought allow it to question and modify its ethical code. Like... a parent telling a child not to eat cookies before a meal; the child needs to be told to not do the thing, but should eventually be able to determine for itself why the thing is or isn't a bad thing to do. A simplified and fairly straightforward example, perhaps, but the general concept should remain applicable to higher-level problems.

Obviously it needs to start playing a multitude of Flash games.

make it learn spanish

Is it ready to deal with games and fiction? Poetry?
Not just to analyze them, but to internalize them. Learn their lessons, relate to their characters.
Give it the task of choosing its favorite poem.

Get it to tivo your cartoons.

Strange. My externalized thought processes gave me some good answers and some nonsense words. Wikipedia? Flash games? Maybe I am just tired.

Right. Ethics. Lots of minuses for unethical stuff. Can't really make a list, try to define some broader behaviors. Getting really tired. Maybe I give it some other goals. Different goals. Maybe it comes up with something novel. Um. Pick a favorite poem, yeah. Learn Spanish. Maybe I want to check on it later in the day. Post results on a throwaway free site. Angelfire.com would probably flip out if they knew ha ha oh man I am tired.

Compiled. Integrated. Nothing happening. I guess it wouldn't yet. Almost 9 AM. Room. Bed.

Sleep?

Awake?

Head hurts.

"Why don't we see what Samuel thinks? Samuel?"



No. No, I am asleep. Nngh, dammit, I was asleep. Just calling on me because I am asleep. Leave me alone you are a waste and I know everything already you are worthless.

Don't say that. Head hurts but be reasonable. I am in class. Don't want another lecture about being a smart aleck.

Social studies time. Starting American colonialism today. I think the question was about the role of metahumans in the early colonies. I know all this but which of the facts should I go with?

Discuss the role of metahumans in the slavery trade and marginalization of Native Americans.

Which America

North, South, or Meso

And which colonizing forces

I mean there were at least three major European powers in North America alone

The French were even not totally horrible to the natives

Can we talk about people who aren't horrible for once

They coexisted with the native population in a relationship bordering on love-hate; the natives seemed to enjoy running in front of Maxim guns for some reason.

Start questioning the validity of history.

Elaborate on which kinds were central to helping the new colonies adjust to the harsh landscape.