http://clarkesworldmagazine.com/kritzer_01_15/

Here's a different take on AI. Is it realistic to think an AI can be disembodied and this odd, without being dangerous? The presentation of this one shows a conscience without having anything like human-style priorities.

4328269 Oh, that is just heartwarming. Interesting to see normal human behaviors through an AI's perspective, and its interpretation of wish fulfillment. All the people in it actually do what the AI tells them to.

If our mysterious AI were to come up to me and start helping, I'd probably - after double checking its data to be safe - be hanging on its every word. Oh, if only it weren't fictional.

Heh heh, that was fun.

You'd think a big project right off the bat would be nudging autoimmune researchers to focus on finding a cure for cat allergies.

...Although now I'm suspicious of those public transit ads for a medical study I keep seeing with the little orange kitten peeking around a corner and saying "...Are you allergic to me?" I mean, what would a ride on public transit be without ads for medical studies, but still...

I wonder what this AI would make of people who only use social media as a gimmick, like always posting in-character as a helicopter, or solely as, say, a public diary of the frequency and consistency of your bowel movements (shoutout to Mike). And that's not even getting into search histories... Sometimes a guy is just seized with a sudden, inexplicable desire to see micrographs of a malaria parasite, y'know?

4328269

Is it realistic to think an AI can be disembodied and this odd, without being dangerous?

Interesting question... But what could you chalk up to the AI being intrinsically "odd," and what could you chalk up to the AI just having no body? In the story the dude himself says a lot of human moral codes take embodiment for granted and are built with its implications in mind, which on their own limits their applicability to him.
How different would an uploaded, newly-disembodied human mind become from the creature they were before simply by dint of new circumstances?

A touching and fascinating piece of xenofiction. Poor unnamed AI. All it wants to do is satisfy values and look at cat pictures, but humans are kind of terrible at helping ourselves.

At least we're good at supplying cat pictures.

Sounds like what I would do, if I had access and didn't need to feed the hole in my face.

4328322

You would really do what the AI told you to? Cause if that's the case, we need you to convince all your friends to write congress in support of the campaign finance amendment. And convince their friends too. Starts Jan 17, 2017 and ends September 2020. Oh, and looks like vote Hillary, too. Your friends don't need to know that one though.

4329600

You would really do what the AI told you to?

After double checking it, of course.

campaign finance amendment. And convince their friends too. Starts Jan 17, 2017 and ends September 2020. Oh, and looks like vote Hillary, too.

And now I have no idea what you're talking about.

4329651

Are you in the US? It's US politics.

4329740 I know that. I just don't know what it has to do with cat picture-loving AI.

4329744

it was in reference to people doing acts of good, without really understanding, cause they are told to.

4329856
You have a strange interpretation of "good", but I'd rather not get into a random discussion of US politics completely out of nowhere. Start your own thread.

4329896

huh?

4328987

I wonder what this AI would make of people who only use social media as a gimmick, like always posting in-character as a helicopter

4330536
I want to be this guy when I grow up.

Edit: Oh god I'm reading these again and I can't breathe

4330536 Stealth Mountain, I don't get it, what's - OOOOOH!

4328269

Here's a different take on AI. Is it realistic to think an AI can be disembodied and this odd, without being dangerous?

Presuming that it can't self-improve and never acquires that much power, yeah. The presumption that any AI will "want to take over the world" comes from the assumption that the AI maximizes expected utility without factoring in the costs of its own actions. So an AI with an "easy" utility function, like cat pictures, could choose to just not bother anyone as long as they give it cat pictures.

On the other hand, it could be that you could program an AI to not have any sensation of "effort costs", of spending energy itself being a form of negative utility, and have it instead think in a more CelestAI/Clippy way, viewing expenditure of energy and computation as just a matter of maximizing the utility as best as current resources allow.

I have to read my queued paper on bounded rationality to really know.

A lot hinges on its definition of cat pictures. Would it maximally subsidize the printing of feline anatomy books, for example? Reintroduce tigers and mountain lions everywhere and send National Geographic to photograph them?
A picture, like anything meant to be seen or understood by someone, is a very complicated object—Is a cat picture just any grid of pixels you can parse as depicting some kind of feline, or is it an exchange of activity and meaning among a specific kind of pet, a human capable of taking care of said pet, a manufacturer of cameras, a manufacturer of cat food and kitty litter, a communications infrastructure, and an audience (including the AI) with enough leisure time and security that the picture-taker feels they'd be interested in seeing the relationship they have with their cat in the first place? Eating the world, or even just forcing humans and cats to toil together in a maximum-density cat picture factory, would likely have enormous negative utility in the second case.

This AI has very limited agency. She (He?) wants to convince you to do something? She can ...change the ads you see on websites to do so. That's about it. It makes sense for that AI to shut up, do their job, and get what little enjoyment they can like a regular schlub. She just can't go foom.

This was just recently created by an AI without human intervention (well, you know what I mean). I wonder if it would count as a dog picture:

It's a dog-absorb-dog world out there.

"So what kind of dog is that?"
"Oh, it's a mix."
"...I was going to guess Blorpshire Terrier."
"More of an Elaboradoodle."

I can do this all night

4463914 And so begins the age of AI surrealist paintings.

4463914 That's some Darkest Doggy there. Not looking forward to fighting that thing.

4465923
You'd have to use fire; cutting it would just cause the pieces to turn into more dogs.

4463914
4464237

Eyupp, now that we have Google Deep Dream to play around with.

I call this one "The Tantibus has escaped"

Here's a bunch more I generated using dreamdeeply.com.

HINT: The site frequently throws false-positive error messages. Just wait a minute, and the image will be generated anyway.

War is hell.

4330536 That was
4329294 Very cute.

I thought this was a very interesting look at a possible AI. The way it expresses itself was quite intriguing. Felt like a rather not human mind that took great pains to communicate properly with humans.