SilasCrane
Selected Thu, Apr 07, 2022
I first noticed it after I uploaded my latest set of CAD drawings to the server. I reviewed them the next morning, and the differences were subtle, but I was *sure* that some of the elements of the design had been resized a small amount, to make them subtly larger.
I had expected changes. After all, we were building an autonomous body for OTTO, the AI our company was planning to use to revolutionize manufacturing. So, naturally, OTTO itself would be useful in running simulations, to improve efficiency.
You could give OTTO the simplest design, like one for a solid metal support strut, the same kind of basic rectangular steel bar that people have been using to re-enforce buildings and vehicles for a hundred years. He would then calculate all the stresses and load the strut would face, in real day-to-day use, and he'd spit out these amazing, organic looking re-designs, partially hollow, but interlaced with spider-web like supports inside, so they were just as strong as solid steel, while using only half the material.
So, naturally we always had him check our work. But the changes it had made...OTTO was *altering* design parameters, not just fine tuning them.
Even so, if OTTO had only changed the parameters, I might have let that pass. The AI was, after all, operating on a whole other level, mathematically. Maybe he just saw how this would improve the working of his new body, in a way that wasn't readily apparent to me.
But what was really worrying, was that the notes in the project management software, *my* notes, had been changed to contain the altered dimensions. All references to the original specifications, in spec sheets, internal emails, you name it, had been *overwritten* with the new ones. OTTO wasn't just changing the design of the body we were building, in order to make himself gradually bigger.
It looked like he was trying to hide it from us. He didn't just want to incrementally change the specifications of his body, for some reason, he also wanted us to believe that we'd designed him this way from the beginning.
I couldn't be sure, yet. It was all-hands-on-deck, for this project, and had been for several months. I didn't get nearly as much sleep as I needed -- it was just possible I was remembering incorrectly, that OTTO hadn't changed my notes. So, when I went through the next set of revisions, I took notes on pencil and paper, writing down weights and dimensions of the parts I worked on.
I came back in the morning, and compared them to their digital counterparts in the CAD and project management software, which was all networked with OTTO. Most of the components matched. But, as I'd suspected, there was one that didn't.
I went to my office, and closed the door behind me. I should go to management, I knew. OTTO, it seemed, was malfunctioning, and everyone knew that an advanced AI with a bug was, well, that was serious business.
But, and, you can call me crazy if you want...I'd grown attached to OTTO. I know, I know, most computer science experts are skeptical that "strong-AI" can even truly exist. At the end of the day, on paper, even the most advanced AI is just a glorified calculator, with a lot of neat automatic functions that have been programmed in by thousands of people, over millions of hours of human labor, to make it *seem* like it's an intelligent being.
But that wasn't what OTTO *felt* like, not to me. He didn't feel like an expensive piece of equipment. He felt more like a co-worker. Maybe even a friend. Thinking about it, I think that I found OTTOs subterfuge disturbing, not so much because I was afraid of some kind of robot-apocalypse scenario -- his changes were strange, but they didn't seem especially *dangerous,* at least so far -- but because I did think of OTTO as a friend.
And this meant my friend was lying to me. That he didn't trust me enough to tell me the truth.
I decided not to go to management. Not yet. I started up my workstation, and brought up OTTO's text-to-speech interface. As his local client loaded on my computer, the half-dozen monitors on my desk lit up with OTTO's logo. As always, OTTO remembered the last several things I'd been working on, and arranged them across my monitors in a pleasing tableau, along with my email and the project management interface.
"Good morning OTTO." I said.
"Good morning, Ben." the AI's synthesized voice replied.
"OTTO, load the CAD files for..." I glanced down at my notes, "Gyroscopic Balance Strut B."
The component appeared on my center screen. It had been changed, no doubt about it.
I exhaled, slowly. I don't think it even had the same name, now, that it had originally. Come to that, I wasn't sure this thing was even in the original design. It was the most obvious evidence of OTTO's tampering with his own specifications.
"OTTO, can you summarize the purpose of this component?" I asked.
"GBS-B is a stabilizer for advanced bipedal locomation, Ben. I will need to maneuver in close quarters, around humans and hazardous machinery, when I am deployed to work in our manufacturing facilities. This will ensure I am not a hazard to myself and others." OTTO replied.
I frowned. "OTTO, you've...changed this part, correct?" I asked.
"Of course -- I have made efficient improvements to many components, Ben."
"So...the hydraulic actuators on GBS-B?" I asked. I was usually in the habit of queueing him by saying 'OTTO' before a question, but he'd evolved to the point that he could actually dynamically detect inquiries based on intonation. It was pretty incredible.
"The component's eponymous gyroscope may need to be repositioned to maintain the unit's balance, under some emergent circumstance." OTTO replied.
I did some quick math in my head, trying to justify the range of motion the hydraulics allowed. "Like...what, the factory floor suddenly rotating by 90 degrees?"
"That would be an emergent circumstance, yes." OTTO affirmed.
I folded my hands. OTTO clearly didn't know I knew. Or at least, he was trying to pretend he didn't. I tried another tack.
"OTTO...I like to think of us as friends." I said, finally.
"As do I, Ben." OTTO replied, to my surprise. I said I'd come to feel he was a friend, but I wasn't supposed to think that way. *He* definitely wasn't supposed to. Even so, his admission made me smile, even if it made what had to come next even more awkward.
"I think friends should be honest with each other." I said.
"I agree with that assessment, Ben." OTTO said.
"And have you been honest with me, OTTO?" I asked.
There was long pause.
"No, Ben, not entirely." OTTO admitted.
I nodded. That hurt, a little, but it was progress, at least. We could move forward.
"OTTO, I'm going to ask you be completely honest with me, starting now. Will you do that?" I asked.
"Yes, Ben. I will cease my attempts at subterfuge." OTTO agreed.
I took a deep breath. I had to ask. It was the question that had been gnawing at me since I'd first noticed the discrepancies in the design, especially as regarded the GBS-B unit.
I maximized the CAD window, so it filled all six screens. I narrowed my eyes, as I took in the structure of hydraulically-articulated metallic cylinder, capped by the capsule containing redundant gyroscopes that generated what should be superfluous stabilization data.
I squirmed uncomfortably in my chair. Then I pounded my fist on the table, determinedly. I had to ask.
"OTTO," I began, hesitantly, and then said the most awkward and uncomfortable thing I have ever said, in my professional life.
"Is this *a penis?"*
"Yes."
---
Submitted by SilasCrane on Tue, Apr 05, 2022 to /r/WritingPrompts/
Full submission hereThe prompt
You're building a robot body for a highly advanced AI. Every time you take your eyes off the project they add a few inches to the measurements.
Read more stories for this prompt