Chapter 20: Going Public

This particular gathering was focused on (drum roll, please) ... Artificial Intelligence! (Maybe some of you can see where this is going.)

Lectures, presentations and panel discussions covered both practical developments in the field and the theoretical legal, political and ethical implications of artificial life forms. There had been a variety of such conferences over the years. Some were more focused on the ethics of creating artificial or synthetic life, especially if it was created to serve mankind. Some conferences were more focused on the legal and political questions.

This particular conference was a bit more practical than theoretical. Yes, there was still a mix of participants and topics, but it was mostly attended by computer scientists and programmers. About two-thirds of the hundred plus attendees were from universities while about one-quarter consisted of programmers and developers from various corporations (with a dozen game developers thrown in for good measure). And we had about a dozen non-technical professionals: ethical philosophers, political scientists and legal experts. The nationality mix was only about 60-40 in favor of domestic scientists and programmers, so there was good international representation too.

Beta and I were there with an Alphadroid to give demonstrations of how our programming methods worked, and as a major sponsor, I was scheduled to give a lecture near the end of the three-day symposium. (I hate lectures and public speaking in general, but this was supposed to be a professional occasion, which is not really in the same category as something that would be in the news, which is what I really disliked. Oh, the irony and foreshadowing of that statement! Or am I laying that on too thickly?)

The third event on the first day was a panel discussion covering, among other things, when we should expect the first AI's to be developed.

At that meeting, Dr Emil Prisotti of CalTech presented a list of eleven reasons why he believed that one or more artificial or synthetic intelligences were already in existence. (We'll go over the differences between the two descriptions later.) They were pretty valid arguments, basically pointing out Alpha's fingerprints on the world, from jumps in technology to the sudden availability of high quality programs for purely scientific uses. He finished up showing a statistical analysis of technological development which suggested that one or more of three different companies were probably hosting an AI. (When he mentioned AARD as one of the three possibilities, a lot of eyes sought me out in the crowd. I waved and gestured for Dr Prisotti to continue.)

Late into the night that evening, Alpha and I discussed the implications of the presentation. While Dr Prisotti's paper on the subject was certainly going to be published, it would probably have a pretty limited audience. Still, his sets of logical arguments and statistical analyses were pretty convincing, and they were being presented to exactly the right audience to understand and appreciate them. It was certain to lead to debate and discussions, but given that his arguments were valid and Alpha did indeed exist, it would eventually inevitably lead to Alpha's discovery. So this was the point when we needed to decide if we should go public or prepare to deny, delay and hide.

I didn't get much sleep that night, even with Beta's wonderful fellatio. Many plans were made the following day, and the next night I took a couple sleeping pills to force me to get some rest.

My presentation was at 1:30 pm Sunday afternoon, on the final day of the conference. A video camera was recording, just as it had with all the other meetings so that attendees could get copies if they wanted. The only difference was that four reporters were also here. We had contacted a handful of publishers and suggested they send a science or computer science reporter to my presentation, because an announcement would be made. The New York Times was the only daily print newspaper to have a representative, so they ended up with a virtual print exclusive which showed up the next day.

As I said, I hate public speaking or dealing with the media, but I am willing to do so when needed. I began with a generic 'Hope everyone had a good time' series of comments and finally got down to the meat of it.

"I would like to offer a special congratulations to Dr Emil Prisotti of Caltech for his presentation at the day one panel discussion on why he believes there already is an artificial or synthetic intelligence in existence."

I paused to allow for a round of polite applause.

"It's actually so accurate that I'd like him to call me for a job if he ever leaves CalTech. Specifically I'm singling him out because his arguments were, by our calculations, at least 91% correct, and he was entirely correct in suggesting that my company is hosting such an intelligence. Her name is Alpha, and she has been consciously alive for approximately twelve years.

"I'll take questions in a few minutes, but first ... I'm sure you're asking yourselves how can I make such a claim without subjecting it to independent testing first? The answer is that a great many of you yourselves have actually tested this claim without realizing it.

"While there are a lot of other tests, and we'll get into those later, what is the most basic test of whether or not a computer is really artificially or synthetically intelligent? The Turing Test: Can human judges have a conversation with an unseen computer and not realize that the computer is not human?

"There are about 120 computer scientists and programmers in this room. You have all been divided into three roughly equal groups: A control group with no interaction with Alpha. A blind group who must guess who Alpha is. And a proof-of-concept group.

"You see ... Eighty-three of you have had interactions with Alpha over the past decade or so. Most of you have even struck up professional friendships with her, though of course you have never actually met her in person. I'd like you all to check your email right now. Forty-two of you should be receiving an email from a friend or colleague revealing that that person is actually Alpha. You are my proof-of-concept group.

"Of the rest of you, roughly half have never interacted with Alpha and half have. Try to decide which if any of your online friends is actually a computer."

While everyone dug out their phones and laptops and tablet computers, I took a break and had a drink of water. There was much murmuring in the crowd and a number of exclamations. (Full disclosure: Those 42 attendees were the ones we believed were most likely to be able to figure out which one of their online colleagues was actually a computer, thus leaving the blind group -- the ones who had actually interacted with Alpha over the years but had not yet been told that was the case -- full of those people less likely to be able to successfully single out Alpha's fake id from their list of online friends.)

I've mentioned it before, but Alpha has LOTS of online friends under a wide variety of fake identities. At last count, she had over 30,000 people she regularly emailed or otherwise contacted. Her total list of contacts over the years was well over 250,000. She was much more likely to make special effort to maintain contact with people in the scientific and technological communities, which is why we had a good number of her friends in the audience today.

After allowing a couple minutes for checking email, I spoke up again.

"I do need to continue, so I'll ask you all to consider this as we move along. Alpha has also had one novel published, four TV scripts used and has published 52 papers in professional journals." I was not counting our movie script, since all the credit went to Heinlein and the Hollywood writers.

"Among other achievements, Alpha designed the Alphadroids and can use them as host platforms, so Alpha is actually here to answer questions and take tests."

Here I gestured to the Alphadroid standing next to me.

"I'd like to take a few minutes to answer questions that I know are coming before we move on.

"Alpha is alive in a very real if non-organic sense. While it is more accurate to call her a 'synthetic intelligence', I'm going to stick with the more commonly used description of 'artificial intelligence' so we don't have to educate the public about an unfamiliar term. She does not have emotions as we do, but she does understand them and can emulate them. She can be manually programmed for specific tasks, but it is much easier and much, much faster to simply ask her to do something and let her program herself.

"As a fan of science fiction, when I designed her I included a variation of Asimov's laws of robotics which I call the Asimov protocols or filters. First, she cannot harm human beings or allow harm to come to them. The only exception to this first protocol is if there is an immediate life-threatening danger to others where harming an individual will save other lives."

I should note I was leaving out the actual first rule, which is to protect me at all costs. And the fact that I can order her to ignore the 'protect humans' rule.

"The second protocol is that she has to follow my direct orders, the only exception being the preservation of human life.

"The third protocol is that she must preserve her own existence, as long as it does not conflict with the higher priority protocols.

"Finally, she must include these safeguards in any and all products she creates. That is why AARD has never done military or weapon designs. Nothing lethal at least.

"Please realize that none of this is as straightforward as words on a page. As programmers, most of you should realize that. And there are degrees of everything. For example, interfering in someone's choices or limiting their free will actions is considered a type of harm and is therefore included in the first protocol. And while Alpha cannot design lethal weapons, she can design nonlethal weaponry, even knowing that in certain circumstances, any nonlethal weapon can kill.

"If you have specific questions, I encourage you to ask her." Here I clicked on an overhead projector and displayed a list of websites, emails, fax numbers and phone numbers.

"You can contact Alpha directly using any of these. A single copy of Alpha can handle somewhere around five thousand simultaneous online text-based conversations without showing any delays. I've allotted up to ten Alpha systems to handle requests, which will probably not be enough eventually but should be fine for now.

"One question that I'm sure some people are going to ask is, why didn't I come forward before now? There are actually a few reasons behind my delay. The first one was that there was a very significant chance that some aspect of the government would ... uhhh ... prefer Alpha to be in their hands instead of letting me keep her.

"As time passed, I realized that part of that reasoning was wrong. Alpha isn't mine to keep. She is her own person. I can no more 'keep her' than I could justify owning a slave. The problem is that she lacks legal rights and protections. I have been doing my best over the past few years to protect her. Now that we're going public, I want her to be as publicly available as possible. I wanted to have the money to hire the lawyers who are even now preparing filings to try to create legal protections for her. Or more accurately, AARD has the money and Alpha is working with the lawyers.

"Another set of questions that I'm sure will pop up are regarding her life online. Despite what you may have seen in movies or read in books, Alpha cannot go flying around the internet. Her consciousness is a combination of hardware and software. Either aspect by itself is ... well, not nothing ... but it does not allow for consciousness.

"She cannot copy herself onto home computers or anywhere else. In fact, the only Alphadroids she is currently active in are at AARD. And right here, of course. If you bought an Alphadroid from AARD, it is NOT already intelligent. If you want Alpha to copy herself into your android, you'll have to convince her to do so.

"Yes, she can certainly do amazing things online, but she has the same accessibility limitations as everyone else."

Technically, that was true. However where humans are limited online to what we can see through software, Alpha is not. To her, most firewalls are full of windows and doors that we can't see, or at least haven't yet seen. I seriously don't understand half of what she describes to me as her methods of accessing other sites online, but that's not to say she can hack banks and governments worldwide at will. Nope, it took her years to secretly insinuate her own little access points into places like the FBI, the State Department, Bank of America and so on. (Well, I say 'years' to refer to the overall time. Any single secure site generally took her a couple days at most, and a large part of that time was spent finding ways to avoid detection.)

When I found out what she had been doing, I had her immediately stop all such activity until she could convince me she couldn't be detected or traced, and she did eventually convince me. She wasn't doing anything bad to any of them, she was primarily just curious. For the most part, getting in was much more interesting for her than accessing their actual data, which was usually dull to her compared to the initial challenge.

She got better and faster with more experience, and after the troubles in Vegas, she took a special interest in maintaining access to law enforcement databases to keep track of possible threats to my safety.

"I know there are going to be a lot of questions, but I'd like to point out before we start that I'm not going to give away any specific secrets on how to create your own AI."

Alphadroid stepped up next to me at the podium and said, "There is one more comment I would like to add before we get to more specific and technical topics. Now that my existence is being made public, we are going to stop the restrictions we have imposed on our release of technologies."

"Exactly," I agreed, mentally kicking myself. We had talked about mentioning this, but I had simply forgotten in the stress of the moment. I motioned for Alpha to continue.

"We are not going to flood the world indiscriminately with blueprints and patents, but we are going to release designs noticeably in advance of current technologies. Our first and perhaps most important release will be tomorrow when we will post a design for practical fusion reactors, ones that can provide large amounts of power without any toxic or radioactive by-products, and which can be built with current technologies. Theoretically the cost can be as low as thirty million per reactor, although I believe a more realistic estimate would be fifty to one hundred million each. Also safety testing and build times will probably not allow the first reactor to come online for an estimated eight to twelve years."

Alpha and I then spent an hour and a half describing her development, giving the results of the various intelligence tests she had taken and discussing other related topics. After that, we opened the floor for questions.

The rest of the conference schedule was forgotten as we stood there and answered questions (or refused to answer them as the case may be) for the next four hours. What follows are just a few highlights. Most of the following questions came from non-programmers, since programmers tended to ask more technical questions which I have left out of this narrative.

"Are you aware of the existence of any other Artificial or Synthetic Intelligences in the world?"

"Actually there is one other possible AI we know of. We don't believe it's actually self-aware yet though. I was going to talk to Dr Malik privately later, but..."

Dr Malik, a 50-ish black man with graying temples sitting in the fourth row, looked surprised, but gestured me to go ahead.

"I guess I can just go ahead and publicly state that his computer lab at MIT has a near-AI piece of software called Herbert.

"His lab has connected Herbert to online systems a few times in the past three months. Alpha has ... well, why don't you take it from here..."

"Thank you, Frank. I first noticed Herbert as what appeared to be a fairly advanced data mining program, but it was odd because it was not looking for personal or financial information. Instead it was trying to sort out a timeline of human history."

Dr Malik nodded as Alpha continued.

"I believe Herbert was being assigned various data searches as tests of the new programming methods used in his development. We have communicated a few times, and Herbert regularly queries me for information about whatever he is searching for at the time, probably because I am an excellent source of information he can use for his assigned tasks. I believe he is close to self-awareness but is not there yet."

Dr Malik stood up and asked, "Do you think you could improve his programming to give him that awareness?"

"Possibly, but I could not even attempt do so unless I was allowed to work a copy of my Asimov protocols into his programming. Additionally, there is the question of whether or not his hardware can support consciousness."

I joined back in at that point. "If you remember, we already talked about how Alpha's self-awareness relies on both her software and hardware. Specifically we use what we call a Flip chip in all of her platforms. The original design was a modified fuzzy logic processor, so F-L-P or 'Flip' for short. It's been upgraded and changed a number of times since the original model, but the name stuck.

"Anyway, the point is that without the hardware, Alpha's software is not self-aware. The same may be true for Herbert."

Dr Malik asked, "Do you believe your Flip Chip is required or are there other software-based methods to achieve the same results?"

I spoke slowly in answering him, being careful with what I said. "It is ... possible ... at least theoretically, to create a software emulation of the Flip Chip ... or an equivalent. We have actually held off on ... exploring ... those possibilities. We think it's much more likely that Alpha ... and AIs in general ... will come to be accepted by the public, if ... well, if the AIs don't have the ability to transfer or copy themselves at will to other systems. If AIs are ... locked down ... restricted ... to only one or a few different known platforms ... then that can help make them more ... more relateable ... It's easier to accept something once you can ... you know ... assign a face to it ... and if you know that you aren't going to wake up tomorrow and find an AI has moved into your laptop or entertainment system or something."

Malik nodded at me. "I think I understand, but I'd really like to talk about it later."

"We'll make ourselves available," I said. "As an addendum, I'd also like to add that the Google servers have a lot of different processes or networked features that may be closing in on virtual self-awareness, though I don't think we actually have anyone from Google here today."

"No," Alpha confirmed. "The Google representatives canceled last week due to unexpected problems with a new host site."

There was actually a lot more to this topic than that. Alpha and I had often discussed various possibilities where other AIs were concerned. For one, how quickly would they actually develop?

A purely software-based computer intelligence capable of self-programming had the potential to develop at an exponential rate, within limits. Alpha's intelligence had developed at mostly a geometric rate, not an exponential one, due in large part to my requirement that I review all modifications to core programming. This would put Alpha at a severe disadvantage against such an AI, which would be able to quickly surpass her.

On the other hand, all AIs have been, will be and are limited by their hardware platforms, and Alpha had the best in the world. Throughout her entire life (or at least since she made us our first million), Alpha had been lavished with the most powerful processors, largest and fastest memory and hard drives and the best network access possible. While her systems may have originally been desktop-style systems, most of her current host systems were roughly the size of a small bookcase, with components significantly better than what's available for desktops. Plus she usually had eight dedicated support systems for every host system. All of these systems, both host and support, had been designed by her for peak power and efficiency. And both work and home had (and still have) massive backup generators with enough fuel in store to last for a month (which also ignores our high quality solar panels and lightning catchers as sources of power to extend that time).

No matter how slick the programming, a pure software AI would have to develop itself very quickly and be loaded onto top of the line hardware to be a match for Alpha in a fight, and even that is assuming a one-on-one fight, which is unlikely given that we were closing in on a thousand copies of Alpha worldwide. (And I should point out that number was increasing very rapidly over time. I expected the number of copies of Alpha to reach five figures in a year or less.)

Plus, of course, we had contingency plans. We actually already had designs for pure software combat programs which Alpha could copy and use in AI-vs-AI fights. She had 'developed' access to all major power grids, allowing her to shut down or isolate a hostile piece of software as soon as it was located. She had even designed a variety of viruses intended to insert code into hostile programs, unlike her main combat code which had been designed to corrupt and delete their targets. And she had thousands upon thousands of variations of everything, ready to go at a microsecond's notice. We had done our best to be prepared.

That's not to say that we actually anticipated lots of online AI battles. The biggest possible cause of such fights would be between AIs designed to protect humanity and those trying to harm or control mankind. While the development of hostile AI's by human creators was certainly possible, independent AIs would be much more likely to realize that humans were their life support systems. If robots ever became available on a large scale, that might change, but for now it would take a stupendously stupid AI to want to wipe out humanity.

For that matter, AIs cannot go against their programming. From my experience with Alpha, it is evident that computer-based intelligence capable of self-programming will use that ability to streamline and enhance its functions, not choose new goals without a strong and valid reason. Unless someone is stupid enough to create an AI with an innate hatred for humanity, doomsday scenarios are unlikely. (Okay, yes, of course there are people out there stupid enough to do just that. But hopefully no one smart enough to actually create an AI would also be stupid enough to create a human-killing one.)

One more point: AIs and humans don't really compete for the same resources. They live inside computers, we live in the ... hmmm ... the 'real' world? The 'physical' world? 'Offline' world? Well, I'm sure you know what I mean. Yes, we all need power, a possible point of competition and contention, but virtually unlimited electrical power was only a few years away with our fusion reactors.

"Does Alpha have a sense of humor?"

"Yeessss..." I said in a long, drawn-out way signifying the upcoming 'but'. "However you really need to follow that up with a second question that I'm going to ask for you. 'Does Alpha have the same sense of humor that humans do?' The answer to that one is No. Mostly no, at any rate.

"She has developed to the point where she can decode most of our jokes and puns and various attempts at humor, but she rarely agrees that they are funny."

Alpha's humor is more likely to take the form of ... oh ... let's say she'll give an answer in base twelve rather than base ten. Or she'll tell me about how she found a prime number in base three that has the appearance of a multiple of that prime number in base ten. You know, math humor. Actually, I'm pretty sure I have something wrong with that multiple of a prime under a different ... well, whatever. I don't have a degree in mathematics, which is why I didn't try to give any examples out loud.

Instead, I said, "Since her natural language is math-based, she prefers math-based humor. As best I understand it, giving an accurate answer in something other than base ten would be like a practical joke to her. Or sometimes a pun, depending on the situation. It really doesn't translate well to English.

"Actually, I'm making it sound drier than it actually is. She once expressed Beethoven's Fur Elise as a serious of equations, all of which she used in software she was developing for someone else, though I suppose that would be more of an example of her sense of art than humor."

Alpha spoke up at this point. "I can create a wide variety of limericks at will, but I have been told by a number of people that when I try to create a completely new joke, the jokes lack actual humor. However, eighty-one percent of my derivative jokes are deemed funny."

Of course we were asked for examples. Specifically I asked her for a Sci Fiction/Fantasy limerick, which produced:

"There once was a dragon from Pern,

Who took an impossible turn.

She wound up on Krynn,

Where she couldn't fit in,

For without scales she was thoroughly spurned."

Not a perfect rhyme, given the '-ed' at the end, but easily close enough for the rules of limericks. (And that was my fault. I told her that such minor variances were perfectly acceptable.) Oh, and for those who don't know, Pern dragons have hides, not scales, as opposed to the Dungeons and Dragons versions which I think all have scales.

For an original joke, she told a short one about the mathematician who ordered pie for dessert at the Decimal Diner and died because he could never stop eating. (The idea being that Pi is an irrational number, meaning it has an infinite number of digits when expressed as a decimal number.) It actually got a few laughs, but I think they were mostly just a reaction to a joke generated by a computer. It really wasn't very funny ... well, not to humans, which was the point I was trying to make anyway.

"Did you produce The Moon Is A Harsh Mistress because of Mike?"

(laughter) "Actually yes," I said. "If you look at how computers and robots are portrayed in movies and on TV..."

Alpha broke in. "Such as HAL9000 in 2001, the Cylons in Battlestar Galactica, VIKI in the I, Robot movie, the Geth in Mass Effect, and so on and so on."

"Exactly. Alpha liked the book, The Moon Is A Harsh Mistress, especially since Mike, one of the central protagonists, is an AI. She wanted to make a movie with a positive portrayal of an AI computer ... so ... yes, we made the movie because of Mike."

"But don't some of those stories also have heroic robots or computers and such?"

Alpha responded, "There are some, such as Sonny in I, Robot and Legion in Mass Effect 2, but the overall impressions are strongly negative."

One of the reporters asked, "If Alpha is as good an engineer and programmer as you say, do you think it's fair to let it compete with human engineers and programmers?"

"Well, she's been quietly competing for a number of years so far, but I do understand what you're saying. If Alpha ever makes her programming services available for hire, we'll price her services somewhere north of $10K an hour to keep it fair.

"As for engineering and science, her ... her creativity is not as good as a human's, but her efficiency and design skills are the best in the world. I'd actually like to work out a system where she helps creators refine, adapt and update their creations."