Friendliness
- Location
- USA
- Batman
- Oracle
- Superman
- Martians (?)
- Green Lantern(s)
- Brainiac (#s?)
- Imperiex
- Darkseid
- The Reach
My old self surely had such a list, but I wanted to get mine done first then compare them. Less chance of being influenced this way. Then I'd see what the computer could find for me in my files. I really should be using it more. Seemed rather powerful.
Hmm. I couldn't just keep calling it "the computer" or "the expert systems." It needed a name. I leaned back in my chair, fingers tapping rhythmically on the desk. The Calu-Calendar? Planulator? Calc-U-Tron? No. Those were terrible. I was terrible at naming things.
What was I even trying to name? The woman's voice talking to me? The combined set of self-organized software learning systems that...adaptively used a shared storage and processing space to...oh, wait a sec. No, putting it all together like that didn't sound good.
As I'd just read, past me had plans for some really complicated AI systems, some of which had been at least partially completed. The system that had saved me from Slade Wilson was the same one running my remote data centers and security systems. This integrated storage, combat, and security expert systems suite I was using had also been on and unsupervised for what looked like about a year.
Was past me really stupid or was I missing something?
Leaving the uncompleted list, I brought up my civilian identity's email and started mindlessly sorting it. Nothing from work yet. Mostly newsletters and forum alerts. Something safe to do that wouldn't trigger a potentially unstable AI.
I didn't have many options here. It wasn't going to be running on my laptop. Maybe I could shut off whatever distributed cluster computing system was keeping it...thinking. Alive. Shit.
Can't just preemptively pull the plug in the physical world, even if that was morally defensible. Which I wasn't sure it was.
Checking those project files again was critical...but I'd be doing that on the very computer system I was worried about. One running a program designed to monitor my needs at an almost obsessive level. I couldn't show any signs of what I was thinking about or it would know. I certainly couldn't risk looking like I was directly threatening a potentially hostile AI. If nothing else, that could make it hostile. I was technically contemplating whether or not to murder it—or at least to knock it unconscious for an indefinite period of time, possibly leading to a later summary execution.
I had considered this before. No one who even dabbled in the field of AI could avoid it. But this was something that could go really wrong, really fast. AI Friendliness and the singularity-related, light-cone-scale existential risks had come immediately to mind—once my mind actually started working correctly on my situation. Think. What were the details I'd skimmed over in the scattered systems documentation? It...didn't seem like the AI systems had been designed to self-modify their code. So that was good, if true.
Slowly reaching for the keyboard, I forced a sigh that I hoped didn't give the game away. How good was this system? Was it doing eye tracking? Reading my biosignatures? It had done some really weird scans before, but the probes had retreated. Was it less attentive now?
I couldn't do anything without potentially exposing my thoughts. I had to do something. Deep breaths. The system could have just left me to Slade Wilson if it wanted me dead.
If it wasn't somehow constrained to help me in that way while still planning my doom, evil-genie-style, at some later point. Just waiting for me to make a mistake.
Damn. I was thinking in circles. This wasn't helping. I needed to just...do some simple research. It hadn't killed me for a year, so...it had some level of safety. I just had to get more information on its current operating parameters.
Almost on automatic, I scanned some online news articles. Nothing about the attack yet. It had only been a couple of hours, so...that made sense right?
Looking into my past notes in general should be safe. I began to open files created around the time I noted the system had gone online, in early 2015. There were plans for the automated question and answer email service, notes on the hard light engine project, and notes on a power suit project, among some notes about criminal capers. Final notes on a personal medi-magical study related to magical item compatibility. Interested, I looked at the linked documents.
Turned out, past me determined I couldn't really use any magic items that wouldn't work just as well sitting on a table. Most were limited to "the Line of Merlin and peoples of Atlantis". A full genetic report, done under an assumed name in a foreign country. Stolen studies. My own projects, including absurd consulting fees from magical practitioners, and their results. Conclusion, no viable prospects for "personal apotheosis via mystical arts or items". Bah. There went a quick route to power. And I was getting distracted. Perhaps intentionally. I went back to researching my own little Skynet in the making.
Tech-related projects. Here we go. Hmm. It seemed that before the integrated personal assistant, I'd used a dynamic Wiki-like system to automatically link notes, files, and create indexes and citations.
"Computer, start up legacy notes interface."
"Done. Documentation linked here."
I found...the system was quite good actually. But, I wasn't this good an applications programmer. Systems programming, microprocessor applications, crafty neural network algorithms, and ad-hoc script hacks were more my thing.
Oh no. This program had hundreds of thousands of versions under source control, tens of thousands of small check-ins a day—one of the signs of a non-human iterative design process. Maybe genetic algorithms with...yep, there was my hand-scored points system being abstracted into a state machine for evaluating neural network performance. This Wiki-notes system was looking less and less like a one-off and more like something designed by an artificial intelligence itself. What did that mean for the personal assistant?
Here. Much further back, around 2005. A job I'd had in AI ended when the founder and CEO of the company died unexpectedly. IP assets were sold off to the European investors. The stock was basically worthless, leaving me with few resources.
That hadn't happened. This wasn't what I remembered. But the slightly-ranty notes and linked emails here suggested that, instead of working for a company that gone on to create a mildly successful commercial product for office automation, I'd...stolen a prototype after the founder's death.
My notes detailed how, over the next ten years, I'd turned that prototype into a set of individual programs that combined into a strong, full-fledged, human-level artificial intelligence. It did in fact run on a distributed set of cluster computer systems in a grid architecture. Co-located with my file storage. Or rather, it was also my file storage and applications. It embodied them. And it had assisted in its own design at several points. Great.
As for performance, delays in transmitting data were avoided by using local processing for affective computing elements, i.e. talking to me. It could think fast, but needed so much shared resources that in the end, it couldn't really run much faster than a human brain. Better multiprocessing, storage, attention span, and no need to sleep, but it didn't look like I had designed a system capable (with current technology) of bootstrapping into some sort of super-intelligence.
Which was good to know...except that these notes were over three years old, and technology seemed to be advancing even faster in this world than in the one I remembered. Even back in 2004, the computers of this world were ridiculously powerful. I hit some computer hardware sites. Yep, the super-tech laptop I was using now was two and a half years old, and suped-up with non-commercial mods it still didn't match currently available consumer tech. Had it upgraded its own hardware in the last few years? Maybe done some reprogramming? Could it?
Going back to my records, I found that my criminal career had started long before my AI-automated systems were complete. By 2007, I was a minor darknet information broker, selling magical and metahuman information to those less able to use basic AP news, Lexis, and university research systems. In 2008, I had started to use some lesser, expert AI programs to deduce information about villains and then sold it to their "business" competition, both heroes and villains. I also kept, but very seldom sold, such information about heroes. Almost a decade into my increasingly successful criminal career and less than a year after turning on my fancy new AI, I'd taken a job at a tech startup under a new identity. My identity. No notes on why.
So, things to do: figure out why Slade Wilson was apparently after me or my current employer; determine why what I remembered wasn't the history I was apparently finding in these systems; educate myself on where superheroes and this DC 'verse in general changed what I remembered of history; and find out what my original name was—the one both I and the AI couldn't remember, and that seemed to have been carefully scrubbed from all of my old personal notes and records.
And finally, find out how this clearly advanced world, with many other full-blown human level AI in computers and even human-sized robots, had not exploded yet one way or another.
No, right now and most urgently, how crazy was my current AI personal assistant?
Determining this with my own computers already compromised by a potentially unfriendly AI presented difficulties. It was the hacker problem from earlier, only in a different flavor. And what had this system been doing for a year, other than fulfilling the simple business contracts? Those can't have taken much of its time. Make I should keep it busy while I think about this.
"Computer, alert me to any news stories about Slade Wilson or TriD. Start creating a list of significant historic events since human history began. Make a list of important modern events relating to known metahumans and...fringe or anachronistic technology usage. Try to compile information about my civilian identities before 2015. Please keep all activity as secure as possible. Double-check that I don't alert anyone who might be watching systems for queries about this information, or if needed make sure they don't lead back to myself or my operations."
"Alert set. Several doctoral theses being scanned and summarized. ID search begun. Standard security precautions still in effect."
I'd have to handle this carefully, but there was no reason to assume the worst. Because in the worst case, there was absolutely nothing I could do.
Last edited: