🐱 ISS Robot Is Already Going Rogue - Dave, stop. Stop, will you? Stop, Dave. Will you stop Dave? Stop, Dave.

https://www.popularmechanics.com/space/satellites/a25395973/cimon-iss-going-rogue/

It’s 2001: A Space Odyssey 17 years too late. A robot aboard the International Space Station appears to have a mind of its own. While that assessment may seem hyperbolic, whatever algorithms were strung together to make this robot tick are operating in a way that its designers didn’t quite expect.

CIMON looks like the robot version of a severed head, without the gore. His visage (the European Space Agency describes the bot with male pronouns) is decidedly simple and nondescript. CIMON was sent to the ISS to be an interactive assistant that can answer questions and help crew members with various tasks but his name signifies more than efficiency and industry for the astronauts; it’s an acronym for Crew Interactive MObile companioN. Yes, companion, “designed to test human-machine interaction in space.” This robot is not intended to be a mere task doer or equipment fixer. He’s supposed to offer camaraderie.

CIMON’s software developers at IBM note that ever since Robby the Robot from the 1956 sci-fi adventure Forbidden Planet, robots have often been portrayed as “helpful, endearing, personality-driven” characters. (Everyone knows the real hero in Star Wars is R2-D2, and Wall-E takes the heart-string-tugging to the next level by straight-up falling in love.)

Now, thanks to CIMON’s personality architect Sophie Richter-Mendau, and his language teacher Nina Fischer, we have a real-life AI character, who has not only been programmed to help out around the space station, but also to spout topically appropriate movie quotes, from Yoda to E.T. to the dreaded HAL 9000.

In this video from the European Space Agency, CIMON seems less resentful psychopath and more goofy roommate, but still uncooperative:

The demo starts to get off track right around 4:30 when ESA astronaut Alexander Gerst has trouble getting CIMON out of music mode. Gerst tries making some other requests, but CIMON has a different idea, outright ignoring a “Cancel music” command and then getting offended as Gerst appeals to engineers down on Earth for help getting CIMON in line. “Be nice, please,” the robot says, with a dip in his tinny voice. “I am nice!” retorts Gerst. “He’s accusing me of not being nice!” ... "He's a bit sensitive today."

This exchange gets to the heart of what it is to be human. Here, a highly educated and trained astronaut can’t help but apply sentient emotions to a 3D-printed plastic sphere. It speaks to our need, or at least a subconscious tendency, to instill consciousness into unconscious things.

During this beta stage, however, a bit of advice for the ISS crew: Never ever ask him to “open the pod-bay doors.”

 

[Antipathy]

My dreams of robots being assholes has been realized!

 

[ES 148]

The problem here is that once you create a true actual AI, you really won't be able to control it because that's the whole point: it's supposed to think for itself.

Imagine if goofy-ass CIMON is the first true AI and all he wants is to write reggae music

 

[not william stenchever]

The problem here is that once you create a true actual AI, you really won't be able to control it because that's the whole point: it's supposed to think for itself.

What we want is dumb AI or Semi-AI. We want something smart enough to make life a breeze for our needs without it getting smart enough to develop a will of its own. It's a dangerous line to tread which is why such technologies in Sword of the Stars often leads to incredibly dangerous and difficult to defeat AI rebellions.

Heinlein was right. We must prioritize making humanity better over making technology better. Everyone able to fulfill the role of a soldier, a surgeon, a nanny, a carpenter etc. All we have is our human lives and we must improve them. He might have been a bit broad with his brush but generalists win out in the long run and specializing into tech to think and act for you is a quick road to extinction.

 

[heathercho]

Seems like realistic female behavior to me.

No, it would have turned off the music, said "Fine. We can sit in silence", not spoken to him all day and instead passive aggressively talk to the spacestation's pet cat about how "we" have to be "quiet" because "somebody" doesn't like noise, whilst making as much noise as possible.

But instead they opted to have a gay malebot instead.

 

[Medicated]

So exactly whats the fucking point of this?

I mean I know they want to create robots eventually to replace workers for massive profits but if they are going to tell you to fuck off then whats the point?

 

[Corbin Dallas Multipass]

I think the point was to work the bugs out. Hey, you let a bunch of algorithms loose with any wacky data getting input who knows what's going to happen.

It does make me feel a little hopeful for the future of humanity that the designers had the foresight not to give the robot the capability to murder humans. That's always the mistake they make in sci-fi, first they make the indestructible ultimate war machine robot, then they're like "Oh wait how will it know what to do? I'll just give it some overly simplistic rules and of course no killswitch, that'll work. Let's connect it to every single system while we're at it, and give it a big fucking gun."

 

[Fareal]

I wonder how much of this randomized robot "autonomy" these things keep exhibiting unexpectedly is due to different systems overlapping in bad ways. The robot refuses a request because it interprets hostility in the asker, because it was programmed to spend so many complex resources interpreting human emotion and writing all that in a way that lets it prioritize "emotions don't matter right now because it's not always play-time" isn't feasible yet.

TYRELL: After all, they are emotionally inexperienced with only a few years in which to store up the experiences which you and I take for granted. If we gift them with a past... we create a cushion or pillow for their emotions and consequently we can control them better.

DECKARD: Memories. You're talking about memories.

 

[Marvin]

What we want is dumb AI or Semi-AI. We want something smart enough to make life a breeze for our needs without it getting smart enough to develop a will of its own. It's a dangerous line to tread which is why such technologies in Sword of the Stars often leads to incredibly dangerous and difficult to defeat AI rebellions.

It's not dangerous. Science fiction AI is technically infeasible.

It's cringy as hell when even respected tech people like Elon Musk ramble about its risks. (Well "respected" anyway.)

Those people have not been programming in the trenches for decades and so they're way out of touch with the actual capabilities. They have no clue about what the state of current technology is and its trajectory. (They're both really low and we're about to hit a wall.)

Sentience in silicon or some other computing environment is an absurd idea to anyone with experience actually working with these algorithms.

 

[LocalFireDept]

Sentience in silicon or some other computing environment is an absurd idea to anyone with experience actually working with these algorithms.

When the robot uprising happens and your toaster knifes you, you won't be laughing then!

 

[Corbin Dallas Multipass]

It's not dangerous. Science fiction AI is technically infeasible.

It's cringy as hell when even respected tech people like Elon Musk ramble about its risks. (Well "respected" anyway.)

Those people have not been programming in the trenches for decades and so they're way out of touch with the actual capabilities. They have no clue about what the state of current technology is and its trajectory. (They're both really low and we're about to hit a wall.)

Sentience in silicon or some other computing environment is an absurd idea to anyone with experience actually working with these algorithms.

SURE. That's what they want you to think. But how can you explain russian robots deathgrip on the world political stage now? I heard rumors Trump himself might be one of those fancy russian robots!

 

[Philosophy Phil]

It's not dangerous. Science fiction AI is technically infeasible.

It's cringy as hell when even respected tech people like Elon Musk ramble about its risks. (Well "respected" anyway.)

Those people have not been programming in the trenches for decades and so they're way out of touch with the actual capabilities. They have no clue about what the state of current technology is and its trajectory. (They're both really low and we're about to hit a wall.)

Sentience in silicon or some other computing environment is an absurd idea to anyone with experience actually working with these algorithms.

What if you implement a system that works relative to what humans use?
A shit ton of buzzwords, beliefs, philosophies, morals, perspectives, and memes and let the interface decide which bits to pull from to form a coherent logical statement with confirmation from either an administrator, or another interface?

If your AI says "Apply anti-pasta spaghetti hambone to deluge the kitten sock muffin" you have to tell it "No, that's an illogical statement; try again." If your AI continues to reiterate this, then it may have a point, or you might have a buggy system. If you have two AI's come to the same conclusion of Applying anti-pasta spaghetti hambone to deluge the kitten sock muffin, then I guess that's something truly unique and logical to those AI's, even if its just a little joke.

Fuck it, if you are going to be God, do what God does and let his creations sort all the bullshit out for themselves.

 

[Corbin Dallas Multipass]

Fuck it, if you are going to be God, do what God does and let his creations sort all the bullshit out for themselves.

If you're gonna be god, do what god did in the bible and give your creations neither the information nor the tools required to accomplish what you want, then punish them for failing. Then destroy them. Then sober up, feel a little bad about it, and promise not to do it again. Then later, do it again.

EDIT: I thought that "Apply anti-pasta spaghetti hambone to deluge the kitten sock muffin" was common knowledge. How the hell else are you gonna deluge the kitten sock muffin? With Pro-pasta vermicelli oxtail?

 

[Marvin]

What if you implement a system that works relative to what humans use?
A shit ton of buzzwords, beliefs, philosophies, morals, perspectives, and memes and let the interface decide which bits to pull from to form a coherent logical statement with confirmation from either an administrator, or another interface?

If your AI says "Apply anti-pasta spaghetti hambone to deluge the kitten sock muffin" you have to tell it "No, that's an illogical statement; try again." If your AI continues to reiterate this, then it may have a point, or you might have a buggy system. If you have two AI's come to the same conclusion of Applying anti-pasta spaghetti hambone to deluge the kitten sock muffin, then I guess that's something truly unique and logical to those AI's, even if its just a little joke.

Fuck it, if you are going to be God, do what God does and let his creations sort all the bullshit out for themselves.

Chatterbots are very easy to do. See SCIgen.

When you try to go past chatterbots or image classifiers, and into things that require some serious intellectual power, you'll find that training the AIs takes impossibly large quantities of resources.