First Conversation With the First Polite Zen Robot in 2050
by Sven Nilsen, 2018
In the previous articles, I wrote about “Conversation With the First Zen Robot” which turned out to be rather unpleasant for human beings. Since, then the design has been refined to the concept of a “Polite Zen Robot”, which needs a second series of articles to bring people up to date with the new definition.
My friend works part-time in the AdvancedResearch organization. It has been 33 years since it started, so now it got a whole staff with hundreds of researchers and backing from multiple companies thanks to the new cryptocurrency system that uses a Gini solver. I tell you this because my friend has asked me to be a guinea pig for their new robot design. Today, the robot will come on visit, and if there are no problems it will stay with me for a week to see how it behaves. I said yes.
At first I was very anxious about this robot thing. However, I watched some videos of lab tests and researchers explaining how the robot worked, which made me much more confident. Apparently, they have figured a way to make the robot behave safely and be very nice at the same time. That is pretty impressive, given that they only worked on this design for 11 700 days.
I put on some fine clothes to make myself presentative. I know it will be just my friend and the robot coming on visit, but who knows what robots think about first impressions? Looking in the mirror, I combed my hair for the third time, watching the clock on the wall, then combing my hair again.
The door bell rang.
Running down the stairs, I straightened my shirt and took a deep breath, then opened the door.
“What the heck is that?!”
I blurted out before I caught myself. Was expecting a shiny robot waiting outside, but there was only a big card box with two legs. Wait. Those legs I recognize. It is my friend holding a big box.
“Don’t be silly, it is just a box with the robot parts. I could not fit it back on my bike. What did you think it was? Are you letting me inside or not? This box is heavy!”
My friend rushed in while me holding the door, carrying the box all by himself up the stairs. I followed hesitantly, just having my big moment ruined inside my head while trying to figure out what happened right now. Are we going to assemble that thing?
“We are going to assemble that thing.” My friend said as if he read my mind.
He opened up the box in the living room, emptying the content on the floor. There were two legs, two arms, a body and a head. It was easy to put to together. Each hand had an on/off button on the outside surface, apparently making it hard for the robot to avoid being turned off if needed. Well, if I were a robot, I could use handcuffs and a gun to avoid humans getting close to me.
“These buttons must be pressed every ten minutes while the robot is on. It is hardwired to not register a press by itself when crossing its arms or when holding a tool. A microfilm records the press patterns inside each hand which is error checked using a laser beam. It has been proven mathematically that this design is safe from hacking.”
My friend turned around the arms and pointed.
“In the middle of the hand there is vacuum tube. This will be activated if the robot goes anywhere near a physical location it is not supposed to be. It will suck in air and inflate a large balloon with hot air popping out its back, lifting it up in the sky where it can be brought down later with a drone-copter. Before this happens, the behavior of the robot is overridden and it will walk to a safe lift-off location in the street. All this logic runs on custom made chips that has a verified design using a SAT solver. Just basic computer engineering using tested designs that resists adversarial attacks by AI the past decades. We are sure that there is only 1 in a billion trillion chance that it will go wrong. I worked on this stuff.”
I nodded, pretending that I understood half of it. Looks idiot-proof, probably the only reason they were willing to let me test it a week.
“I could go on about the other safety mechanisms, but there is no point. These two things I told you are the important ones. First, you need to push the buttons every ten minutes and second, don’t panic if you see your robot go on a balloon trip. Let us get this started.”
He pushed the two buttons on the hands, and the robot’s eyes lit up.
“Oh no. This is embarrassing. Excuse me for not making the entry that you expected.” the robot said.
What. How did the robot know what happened when it was turned off, disassembled and inside a box?
“I saw your clothes and hair. The rest is common sense.”
Of course. I coughed in my elbow and offered my hand. The robot took it and raised to its feet.
“Jeremy.” I said.
“Alan.” the robot responded.
We looked at each other in the eyes. The researchers designed its face to be distinguishable from human, but with subtle but expressive features to communicate emotions.
“Why are you called Alan?” I asked.
“Alan Turing and Alan Watts. Two important figures leading to my existence. I am named after both of them. Do you want to show me around the house?”
I toured Alan around, trying to avoid the kitchen, since I had forgotten to clean the dishes from breakfast. Or, perhaps I was a bit worried it would take the kitchen knife and stab my chest.
“You do not have to worry about me stabbing anyone. I am able to infer this is a bad moral act using Rational Natural Morality. Lungs take up oxygen, therefore stabbing them is bad.”
“OK. I guess you read my mind using facial expressions.” I hoped. Please, no brainwaves.
“Ah... this might be my fault,” my friend interrupted, “Alan and I were talking yesterday about your visit, and I mentioned you worrying about the kitchen knife.”
“Yes, that is correct.” Alan confirmed.
“Well, then perhaps we should take a look at the kitchen after all.”
My friend demonstrated how to give the robot orders.
“Say: Alan, clean the dishes. Go, Alan. This pattern makes it clear for people when they are giving the robot a command. Notice that you should be careful. The robot will do almost anything you can think of asking it doing, except good or bad things. As long what you ask it to do is neutral, according to its goal programming, it will do it.”
“Why can I not ask it to do good things?”
“You can, but it is more complex. I will explain it later.”
I said: “Alan, clean the dishes. Go, Alan.”
Alan cleaned the dishes. He said that it was something he never tried before, but by looking up a video of household tasks on his hard drive, he could learn by watching the video. He finished the tasks without problems. Time to push the buttons to get Alan working another 10 minutes.
“Hmm... I start to like this.” I thought aloud.
I always liked ordering people around. Specially if they were smarter than myself. Alan was basically the most perfect entity to play boss with. Super-super-smart and easy to tell things to do.
We took a round around the living room, testing what Alan could do. He passed the tests, one by one, without doing any mistake.
My friend, which I forgot to tell was named Jake, showed me how to cooperate with Alan to do a good thing. First, Alan did it how he thought it should be done. Then, I repeated the task like he did to show Alan that I understood it. Next, I were allowed to think creatively at each sub-step of the task, using my own variants. One problem was that Alan had to figure out something that was simple enough for me to do, but scored high according to Alan’s goal. This was the first time I really understood what it meant to be super-intelligent.
Alan created a list of people which he thought was suitable for starting a new company in my name. I liked the idea. He called two of them using my phone, explaining who he represented. Both signed up immediately. I called the next two people. They too signed up. After that, I asked a few who lived nearby over for dinner tomorrow, to talk more about the company and the future. Alan was OK with that.
I had never, never thought I could ever, ever run a company. That was before I met Alan. He is staying here for a whole week. The company? We are going to fix climate change!