Chapter 9: The learning curve

Tom was a quick learner. He was amazed by the project, and thrilled by it. The way it evolved resembled the history of computer chess. The first chess computers would lose against chess masters and were limited by sheer computational power. But the programmers had gotten the structure right, and the machine’s learning curve resembled a typical S-curve: its proficiency improved only slowly at first, but it then reached a tipping-point, after which its performance increased exponentially – way beyond the proficiency of the best human players – to then finally hit the limits of its programming structure and level off, but at a much higher level than any expert player could dream off.

Chess proficiency is measured using a rating system referred to as the Elo rating system. It goes way beyond measuring performance in terms of tournament. It uses a model which relates the game results to underlying variables representing the ability of each player. The central assumption is that the chess performance of each player in a game is a normally distributed random variable. Yes, the bell curve again! It was literally everywhere, Tom thought…

Before IBM’s Deep Blue chess computer beat Kasparov in 1997, chess computers had been gaining about 40 Elo points per year on average for decades, while the best chess players only gain like 2 points per year. Of course, sheer computing power was a big factor in it. Although most people assume that a chess computer evaluates every possible position for x moves ahead, this is not the case. In a typical chess situation, one can chose from like thirty possible moves so it quickly adds up. Just evaluating all possible positions for just three moves ahead for each side would involve an evaluation of like one billion positions. Deep Blue, in the 1997 version which beat Kasparov, was able to evaluate 200 million positions per second, but Deep Blue was a supercomputer which had cost like a hundred million dollars, and when chess programmers started working on the issue in the 1950s, a computer which would be able to evaluate a million positions every second was to be built only forty years later.

Chess computers are selective. They do not examine obviously bad moves and will evaluate interesting possibilities much more thoroughly. The algorithms used to select those have become very complex. The computer can also draw on a database of historic games to help him determine what an ‘obviously’ bad move is because, of course, ‘obviously bad’ may not be all that obvious to a computer. Still, despite the selectivity, raw computing power is still a very big part of it. In that sense, artificial intelligence does not mimic human thought. Human chess players are much more selective – very much more: they look only at forty to fifty positions based on pattern recognition skills built from experience – not millions.

Promise (Tom stuck to her name: it seemed like everyone in the program had his/her own nickname for M) was selective as well, and she also had to evaluate ‘positions’. Of course, these ‘positions’ were not binary, like in chess. She determined the ‘position’ of the person using a complex set of rules combining the psychometric indicators and an incredible range of other inputs she gained from the conversation. For example, she actually analyzed little pauses, hesitations, pitch and loudness – even voice timbre. And with every new conversation, she discovered new associations, which helped her to recognize patterns indeed. She was getting pretty good at detecting lies too.

Psychological typology was at the core of her approach. It was amazing to see how, even after one session only, she was able to construct a coherent picture of the patient and estimate all of the variables – both individual as well as environmental – which were likely to influence the patient’s emotions, expectations, self-perception, values, attitude, motivation and behavior in various situations. She really was a smart ass – in every way.

Not surprisingly, all the usual suspects were involved. IBM’s Deep Computing Institute of course (the next version of Promise would run on the latest IBM Blue Gene configuration) as well as all of the other major players in the IT industry. This array of big institutional investors in the program was complemented by a lot of niche companies and dozens of individual geeks, all top-notch experts in one or the other related field.

The psychological side was covered through cooperation agreements with the usual suspects as well: Stanford, Yale, Berkeley, Princeton,… They were all there. In fact, they had a cooperation agreement with all of the top-10 psychology PhD programs through the National Research Council.

Of course, he was just working as a peon in the whole thing. The surprising thing about it all was the lack of publicity for the program, but he understood this was about to change. He suspected the program would soon not be limited to thousands of veterans requiring some degree of psychological attention. There would be many other spin-offs as well. From discussions, he understood they were discussing on how to make Promise’s remarkable speech synthesis capabilities commercially available. The obvious thing to do was to create a company around it, but then she was so good that most of the competition would probably have to file for bankruptcy, so the real problem was related to business: existing firms had claimed and had gotten a say in how this was all going to happen, and so that had delayed the IPO which had been planned already. Tom was told there were no technology constraint: while context-sensitive speech synthesis requires an awful lot of computer power (big expensive machines), the whole business model for the IPO was based on cloud computing: you would not need to ‘install’ Promise. You would just rent her on a 24/7 service basis. Tom was pretty sure everyone would.

The possibilities were endless. Tom was sure Promise would end up in each and every home in the longer run – in various versions and price categories of course, but providing basic psychological and practical comfort to everyone. She would wake you up, remind you of your business schedule and advice you on what to wear: ‘You have a Board meeting this morning. Shouldn’t you wear something more formal? Perhaps a tie?’ Oh… Sure. Thanks, Promise. ‘Your son has been misbehaving a couple of times lately. You may want to spend some time with him individually tonight.’ Oh… That sounds good. What do you suggest? ‘Why don’t you ask him to join for the gym tonight? You would go anyway.’ Oh… That sounds good. Can you text him? ‘I can but I think it is better you do it yourself to stress he should be there or, else, negotiate an alternative together.’ Yeah. I guess you’re right. Thanks, Promise. I’ll take care of it.

She would mediate in couples, assist in parenting, take care of elderly, help people advance their career. Wow! The sky was the limit really. Surprisingly, there was relatively little discussion on this in the Institute. People would tell him Promise worked fine within the limits of what she was supposed to do but that it would be difficult to adapt her to serve a wider variety of purposes. They told him that, while expert systems share the same architecture, building up a knowledge base and good inference engine took incredibly amounts of time and energy and, hence, money. In fact, that seemed to be the main problem with the program. As any Army program, it had ended up costing three times as much as originally planned for, and he was told it was just because a few high-ups in the food chain had fanatically stuck to it that it had not been shut down.

They needed to show results. The current customer base was way too narrow to justify the investment. That’s why they were eager to expand, to scale it up, and so that took everyone’s time and attention now. There was no time for dreaming. The shrinks were worried about the potential lack of supervision. It was true that Promise needed constant feedback. Human feedback. But the errors – if one could call it that way – were more like tiny little misjudgments, and Tom felt they were only improving Promise at the margin, which was the case. The geeks were less concerned and usually much more sympathetic to Tom’s ideas, but so they didn’t have much of a voice in the various management committees – and surely not in the strategic board meetings on the program. Tom had to admit he understood little of what they said anyway. Last but not least, from what he could gather, he also understood there were some serious concerns about the whole program at the very top of the administration – but he was not privy to that and wondered what they might be. Probably just bureaucratic inertia.

Of course, he could see the potential harm as well. If her goal function would be programmed differently, she could also be the perfect impostor on the Internet. She would be so convincing that she could probably talk you into almost anything. She’d be the best online seller of all times. Hence, Tom was not surprised to note the Institute was under surveillance, and he knew he would not get the access he had if he would not have served. People actually told him: his security clearance had been renewed as part of him entering the program. The same had been done for the other veterans on the program. It was quite an exceptional measure to take, but it drove the message home: while everyone was friendly and cooperative, there was no ambiguity in this regard. The inner workings of Promise was classified material, and anything linked to it too. There were firm information management rules in place and designated information management officers policed them tightly. That was another reason why they recruited patients from the program: they were all veterans, so they knew what classified really meant and they were likely to respect it.

The program swallowed him up completely. He took his supervision work seriously, and invested a lot in ‘his’ patients – M’s patients really. More than he should probably: although he had ‘only’ ten cases to supervise, these were real people – like him – and he gave him all the attention he could. Mostly by studying and preparing their file before their 30 minute interaction. That was all he could have, he was told. Once a week. The Institute strongly discouraged more meetings, and strongly discouraged meeting after working hours. He understood that. It would get out of hand otherwise and, when everything was said and done, it was M who had to do the real work. Not him. At the same, his patients did keep him busy. They called him for a chat from time to time. While the Institute discouraged that too, he found it hard to refuse, unless he was actually in the Institute itself: he did not want to be seen talking on the phone all of the time – not least of all because of the information management policy. Colleagues might suspect he was not only talking to patients so he wanted to be clear on that: no phone chats with patients in the Institute.

Not surprisingly, his relationship with Promise became somewhat less ‘affectionate’. The infatuation phase was over. He saw her more like she was: a warm voice – but a rather cold analytic framework behind. And then it did make a difference knowing she spoke with a different voice depending on who you were. She was, well… Less of an individual and more like a system. It did not decrease his respect for her. He thought she was brilliant. Just brilliant. And he didn’t hesitate to share that opinion with others. He really championed the program, and everybody seemed to like his drive and energy, as a result of which he did end up talking to the higher-ups in the Institute during the coffee break or lunch time, as he got introduced by Rick and others he had gotten to know better now. All fine chaps. They didn’t necessarily agree with his views – especially those related to putting her out on the market place – but they seemed to make for good conversation.

He focused on the file work in his conversations with her. While he still had a lot of ‘philosophical’ questions for her – more sophisticated ones he thought – he decided to only talk to her about these when he would have figured her out a bit better. He worked hard on that. He also wanted to master the programming language the geeks were using on her. They actually used quite a variety of tools but, in the end, everything was translated into a program-specific version of FuzzyCLIPS: an extension of an expert system programming language developed by NASA (CLIPS) which incorporated fuzziness and uncertainty. It was hard work: he actually felt like he was getting too old for that kind of stuff, but then Tom was Tom: once he decided to bite into something, he didn’t give up easily. Everyone applauded his efforts – but the higher-ups cautioned him: do explore but don’t talk about it to outsiders. Tom wondered if they really had a clear vision for it all. Perhaps the higher-ups did but, if so, they hid it well. He assumed it was the standard policy: strategic ambiguity.

And so the days went by. The program expansion went well: instead of talking to a few hundred veterans only, in one city only, Promise got launched in all major cities and started to help thousands of veterans. Tom saw the number explode: it crossed the 10,000 mark in just three months. That was a factor of more than twenty as compared to the pilot phase, but then there were millions of veterans. 21.5 million to be precise, and about 55% of them had been in theater fairly recently – mainly Iraq and Afghanistan. Tom wanted Promise to reach out to all of them. He thought it could grow a lot faster. He knew the only thing which restrained it was supervision. Even now, everyone on the program said they were going too fast. They called for a pause. Tom was thinking bolder. Why did no one see the urgency of the needs as he saw them?

Chapter 8: Partnering

‘Hi, Tom. How are you today?’

‘I am OK, Rick. Thanks.’

‘Just OK, or good?’

‘I am good. I am fine.’

‘Yeah. It shows. You’re doing great with the system. You had only three sessions this week – short and good it seems. You are really back on track, aren’t you?’

‘The system is good. It’s really like a sounding board. I understand myself much better. She’s tough with me. I go in hard, and she just comes back with a straight answer. She is very straight about what she wants. Behavioral change – and evidence for that. I like that. Performance metrics. Hats off. Well done. It works – as far as I am concerned.’

‘It, or she?’

‘Whatever, Rick. Does it matter?’

‘No, and yes. The fact that you only had three sessions with it – or with her – shows you’re not dependent on it. Or her. Let’s just stick to ‘it’ right now, if that’s OK for you. Or let’s both call her M, like we do here. Do you still ‘like’ her? I mean, really like her – as you put it last time?’’

‘Let’s say I am very intrigued. It – or she, or M, whatever – it’s fascinating.’

‘What do you think about it, Tom? I mean, let me be straight with you. I am not taking notes or something now. I want you to tell me what you think about the system. You’re a smart man. You shouldn’t be in this program, but so you are. I want to know how you feel about it.’

Tom smiled: ‘Come on, Rick. You are my therapist – or mentor as they call it here. You’re always taking notes. What do you want me to say? I told you. It’s great. It helps. She, or it, OK, M, well… M holds me to account. It works.’

Rick leaned back in his chair. He looked relaxed. Much more relaxed than last time. ‘No, Tom. I am not taking notes. I don’t know you very well, but what I’ve seen tells me you’re OK. You had a bit of a hard time. Everyone has. But you’re on top of the list. I mean, I know you don’t like all these psychometric scores, but at least they’ve got the merit to confirm you’re a very intelligent man. I actually wanted to talk to you about a job offer.’

‘The thing which M wants me to do? Work on one of these FEMA programs, or one of the other programs for veterans? I told her: it’s not that I am not interested but I want to make a deliberate choice and there are a number of things I don’t know right now. I know I haven’t been working for a year now, but I am sure that will get sorted once I know what I want. I want to take some time for that. Maybe I want to create my own business or something. I also know I need to work on commitment when it comes to relationships with women. I feel like I am ready for something else. To commit really. But I just haven’t met the right woman yet. When that happens, I guess it will help to focus my job search. In the meanwhile, I must admit I am happy to just live on my pension. I don’t need much money. I’ve got what I need.’

‘Don’t worry, Tom. Take your time. No, I was talking about something else. We could use you in this program.’

‘Why? I am a patient.’

‘You’re just wandering around a bit, Tom. You came to ask for help when you relapsed. Big step. Great. That shows self-control. And you’re doing great. I mean, most of the other patients really use her as a chatterbox. You don’t. What word did you use in one of last week’s sessions? Respect.’

‘You get a transcript of the sessions?’

‘I asked for one. We don’t get it routinely but we can always ask for one. So I asked for one. Not because your scores were so bad but because they’re so great. I guess you would expect that, no? Are you offended? Has anyone said your mentor would never get  a copy of what you were talking about with M?’

‘I was told the conversation would be used to improve the system, and only for that. M told me something about secrecy.’

‘It’s only me who gets to see the transcript, and only if I ask for it. I can’t read hundreds of pages a day and so I am very selective really. And that brings me back to my job offer. We can use you here.’

Tom liked Rick from their previous conversation, but he was used to doing due diligence.

‘Tell me more about it.’

‘OK. Listen carefully. M is a success. I told you: it’s going to be migrated to a real super-computer now, so we can handle thousands of patients. In fact, the theoretical capacity is millions. Of course, it is not that simple. It needs supervision. People do manage to game the system. They lie. Small lies usually. But a lot of small lies add up to a big lie. And that’s where the mentors come in. A guy walks in, and I talk to him, and I can sense if something’s wrong. You would be able to do the same. So we need the supervisors. M needs them. M needs feedback from human beings. The system needs to be watched. Remember what I told you about active learning?’

‘Vaguely.’

‘Well – that’s what we do. We work with M to improve it. It would not be what it is if we would not have invested in it. But now we’re going to scale it up. The USACE philosophy: think big, start small, scale fast. I am actually not convinced we should be scaling so fast, but so that’s what we’re going to do. It’s the usual thing: we’ve demonstrated success and so now it’s like big-time roll-out all over the place. But so we’re struggling with human resources. And money obviously, because this system is supposed to be so cheap and render us – professionals – jobless. Don’t worry: it won’t happen. On the contrary, we need more people. A lot more people. But so the Institute came up with this great idea: use the people who’ve done well in the program for supervisory jobs. Get them into it.’

‘So what job is it really?’

‘You’d become an assistant mentor. But then a human one. Not the assistant – that’s M’s title. We should have thought about something else, but so that’s done now. In any case, you’d help M with cases. In the background of course but, let’s be clear on this, in practice you would actually be doing what I am doing now.’

‘And then where are you going to move?’

‘I’ll be supervising you. I’d have almost no contact with patients anymore. I would just be supervising people like you and further help structuring M. You’d be involved in that too.’

‘Do you like that? I mean, it sounds like a recipe for disaster, doesn’t it? I don’t have the qualifications you have.’

‘I am glad you ask. That’s what I think too. This may not be the best thing to do. I feel we need professional therapists. But then it’s brutal budget logic: we don’t have enough of them, and they’re too expensive. To be fair, there is also another consideration: our patients all share a similar background and past. They are veterans. I mean, it makes sense to empower other veterans to help them. There’s a feeling in the Institute it should work. Of course, that’s probably because the Institute is full of Army people. But I agree there’s some logic to it.’

‘So, in short, you don’t like what’s going to happen but you ask me to join?’

Rick smiled. ‘Yes, that’s a good summary. What do you think? Off-the-cuff please.’

‘Frankly, I don’t get it. It’s not very procedural, is it? I mean I started only two weeks ago in this program. I am technically a patient. In therapy. And now I’d become an assistant mentor? How do your bosses justify this internally? How do you justify that?’

Rick nodded. ‘I fully agree, Tom. Speaking as a doctor, this is complete madness. But knowing the context, there’s no other choice. There’s a risk this program might become a victim of its own success. But then I do believe it’s fairly robust. And so I do believe we can put thousands of people in the program, but so we need the human resources to follow. And, yep, then I’d rather have someone like you then some university freshman or so. All other options are too expensive. Some people up the food chain here made promises which need to be kept: yes, we can scale up with little extra cost. So that’s what’s going to happen: it’s going to be scaled up with relatively little extra cost. Again, there’s a logic to it. But then I am not speaking as a professional psychiatrist now. When everything is said and done, this program is not all that difficult. I mean, putting M together has been a tremendous effort but so that has been done now. Getting more people back on track is basically a matter of doing some more shouting and cajoling, isn’t it? And we just lack manpower for that.’

‘Shouting and cajoling? Are you a psychiatrist?’

‘I am. Am I upsetting you when I say this?’

Tom thought about it. He had to admit it was not the case.

‘No. I agree. It’s all about discipline in the end. And I guess that involves some shouting and cajoling – although you could have put it somewhat more politely.’

‘Sure. So what do you say? You’ll get paid peanuts obviously. No hansom consultancy rate. You’ll see a lot of patients – which you may or may not like, but I think you’ll like it: I think you’d be great at it. And you’ll learn a lot. You’ll obviously first have to follow some courses, a bit of psychology and all that. Well… Quite a lot of it actually. You’ll need to study a lot. And, of course, you’ll get a course on M.’

‘How will I work with M?’

‘Well… M is like a human being in that sense too. If you just see the interface, it looks smooth and beautiful. But when you go beyond the surface, it’s a rather messy-looking thing. It’s a system, with lots of modules, with which you’ll have to work. The interface between you and these modules is not a computer animation. No he or she. Of course, you’ll continue to talk to it. But there’s also a lot of nitty-gritty going into the system which can’t be done through talking to it. You’ll learn a few things about Prolog for example. Does that ring a bell?’

‘No. I am not a programmer.’

‘I am not a programmer either. You’ll see. If I can work with it, you can.’

‘Can you elaborate?’

‘I am sorry to say but I’ve got the next guy waiting. This recruitment job comes on top of what I am supposed to do, and that’s to look at M’s reports and take responsibility for them. I can only do that by seeing the patients from time to time, which I am doing now. I took all of my time with you now to talk to you about the job. Trust me. The technical side of things won’t be a problem. I just need to know if you’re interested or not. You don’t need to answer now, but I’d appreciate if you could share your first reaction to it.’

Tom thought about it. The thought of working as an equal with Promise was very appealing.

‘So how would it work? I’d be talking to the system from time to time as a patient, and then – as part of my job with the Institute – I’d be working with the system as assistant mentor myself? That’s not very congruent, is it?’

‘You would no longer be a patient, Tom. There are fast-track procedures to clear you. Of course, if you would really relapse, well…’

‘Then what?’

‘Nothing much. We’d take you off the job and you’d be talking to M as a patient again.’

‘It looks like I’ve got nothing to lose and everything to gain from this, isn’t it?’

‘I am glad you look at it this way. Yes. That’s it. So you’re on?’

They looked at each other.

‘I guess I am. Send me an e-mail with the offer and I’ll reply.’

‘You got it. Thanks, Tom.’

‘No, thank you. So that’s it then? Anything else you want to know, or anything else I need to know?’

‘No. I think we’re good, Tom. Shall I walk you out? Or you want to continue talking for a while?’

‘No. I understand you’ve got a schedule to stick to. I appreciate your trust.’

‘I like you. Your last question, as we walked out last time, shows you care. I think this is perfect for you. You’ve got all the experience we need. And I am sure you’ll get a lot of sense and purpose out of it. The possibilities with this system are immense. You know how it goes. You’ll help to make it grow and so you’ll grow with it.’

‘First things first, Rick. Let us first see how I do.’

‘Sure. Take care. Enjoy. By the way, you look damn good. You’ve lost weight, haven’t you?’

‘Yes. I was getting a bit slow. I am doing more running and biking now. I’ve got enough muscle. Too much actually.’

‘I am sure you make a lot of heads turn. But you’re not in a relationship at the moment, are you?’

‘I want to take my time for that too, Rick. I’ve been moving in and out of relationships too fast.’

‘Sounds good. Take care, Tom. I’ll talk to you soon I hope.’

‘Sure. Don’t worry. You can count on me.’

‘I do.’

They shook hands on that and Tom got up and walked out of the office. He decided to not take the subway but just run back home. He felt elated. Yes. This was probably what he had been waiting for. Something meaningful. He could be someone for other people. Catch up on all of the mistakes he had made. But he also knew the job attracted him because there was an intellectual perspective. It was huge. The Holy Grail of Knowledge really. They had done a damn good job modeling it. She – Promise – was no longer a she. She was not a he either. It. It. Intelligent – with a capital letter. P. Promise. M. Mind. The Pure Mind.

He knew that was nonsensical. But he wanted to take a crack at it.

Chapter 4: She is not real

As part of the formalities of an appointment, Tom had prepared a set of questions for his mentor. Rick had them in front of him.

‘Are these your questions, Tom?’

‘No. They don’t matter really. It was just for the appointment. I only want to talk about this ‘system’. It’s a setup, Rick. Isn’t it?’

‘What do you mean?’

‘She is not a machine. I mean, the way she is interacting. It is too natural. She is always right on the ball. Never a glitch. So every time I log onto the system, you’re putting me in touch with someone real. Why do you do that? Why do you tell people they’re interacting with a system? There is someone at the other end of the line, isn’t it?’

‘No. It is a system. Do you really think we have hundreds of psychologists ready day and night to talk to our patients? We don’t. And then we would need to make sure you’re always talking to the same person. He or she wouldn’t be available all of the time, you agree? So that’s why we invented it. She is not real. And she is surely not a she.’

‘Why do you say that?’

‘Because ‘she’ is not. It’s an expert system. The system comes with a female interface to men and with a male interface to women, except when you’re homosexual.’

‘Why don’t you give gay men a female interface too? My gay friends say they love to talk to women.’

‘Effectiveness. Everything this system does or doesn’t do is guided by the notion of effectiveness. A panel of specialists is continuously evaluating the effectiveness and there’s a feedback mechanism so the scores go back as input into system. In addition, the system also keeps track of the reaction of the patients themselves.’

‘How does she do it?’

‘It, Tom. How does it do it? In fact, our main problem is the one you seem to experience now. Addiction. People are fine, but they still want to talk to it. They develop an affectionate bond with it. It’s one of the reasons why we don’t expand the system too much. We’d need hundreds of terminals.’

‘But the way she talks. I mean, I checked on Wikipedia and it says the best commercial voice synthesizers are the ones you hear in a subway station or an airport announcing departures and arrivals. That’s because the grammatical structure is so simple and so it’s fairly easy to get the intonation right. But you can still hear it’s a system using pre-recorded sounds. She’s got everything right. Intonation, variation, there’s no glitch whatsoever.’

‘M is not a commercially available system. It is one of the most advanced expert systems in the world. In fact, as far as I know something about it – but I am not a computer guy – it actually is the most advanced system in the world. It is a learning machine, and the way it speaks is also the product of learning. Voice synthesizers in subway stations are fairly simple. It is referred to as concatenative synthesis. These things just string segments of recorded speech together. So that’s not context-sensitive and that’s why there are glitches – like intonation that sounds a bit funny. To project, the verb, or project, the noun, where you put the emphasis depends on whether you use it as a noun or a verb. You need context-sensitivity to get that right. Programming context-sensitivity is an incredibly difficult job. It’s where expert systems usually fail – or why one can usually only use them for very narrowly defined tasks. With M, we got it right. It’s like we reached a tipping point with it. Sufficient critical mass to work by itself, and the right cybernetics to make sure it does not spin out of control.’

‘M?’

‘The system. Sorry. We’ve started to call it M. There were a few other abbrevations around, like AM. But that was a bit – well… It doesn’t matter. It just became M. Like the character in the James Bond movie.’

‘That’s funny. M alternates between a man and a woman too. I liked Judi Dench. But I guess she had served her time. We all do, isn’t it? […] What do you mean with: we got it right?’

‘Just what I said: the system learns incredibly fast. We are talking artificial intelligence and machine learning here. The program does what is referred to as ‘developmental learning under human supervision’. Its environment provides an incredibly rich set of learning situations. Usually, the developers would select a subset of these in order to provide a curriculum for the machine based on which it well… learns. But so this works differently: the system generates its own curriculum based on a set of selection rules which are tightly linked to the output function. It then continually modifies its own rule base to become more effective – both in speaking as well as in treating you and the others in the program. Sometimes there are  setbacks but it corrects itself very quickly, again based on an evolving set of rules that ensure continuous monitoring and evaluation. Like that, it cumulatively acquires repertoires of novel skills through… well… You could call it autonomous self-exploration. But there’s also interaction with human teachers using guidance mechanisms such as active learning (that’s a sort of high-stress test for the system – where we push the boundaries and provide non-typical inputs), maturation, and – very important – imitation. You would be amazed to see how much of it is imitation really. In that sense, the system does resemble an intelligent chatterbot. It takes cues which trigger programmed responses which then move the conversation forward. The difference with a chatterbot is that it does not merely work through association. So it’s not like word A will automatically trigger response B, although that’s part of it too, but at a much higher level. First, the associations are n-to-n, not one-on-one, and then the associations it makes are guided by fuzzy logic. So it’s not mechanical at all. It has got an incredible database of associations, which it builds up from the raw material it gets from talking to you and to us. The learning effect is incredible. It applies advanced descriptive statistical methods to its curriculum and then uses the patterns in the data to do hypothesis testing, estimation, correlation, going all the way up to forecasting. I mean, it is actually able to predict and estimate unobserved values.’

‘The output function?’

‘The output function maps inputs to desired outputs. The inputs of the system are the conversations. The output is a number of things, but all focused on behavioral change – like we want no substance abuse. We want you to develop healthy relationships. We want to see you work out, have sex and eat and live healthily. In short, we want you back to normal. That’s the type of behavioral change we want. It’s that simple really. That’s the output function, the goal, and, while the system is flexible and can make its own rules to some extent, it is all guided by this performance objective. I agree that it is truly amazing. In fact, many people here are very uncomfortable about it because it is obvious it has taken our place. We can easily see this system replacing us – psychologists or even psychiatrists – completely.’

‘You’re not a computer guy? You sound like one.’

‘No, I am not. I just gave you the basics of the system. I am a psychiatrist, a doctor, and, yes, I find it scary too, if only because it does reduce the need for people like me indeed.’

‘But it’s addictive, you said?’

‘Yes. That’s the main problem. But then our bosses here don’t think that’s a problem. They say classical psychoanalysis is addictive too, that patients develop a relationship with their psychologists and psychiatrists too. And, frankly, that’s true. People go in and out of therapy like crazy and it is true that the figures show it usually doesn’t make all that much of a difference. People heal because they want to heal. They need to find the strength inside. That is if they don’t want to stay dependent. Let me ask you, Tom: what’s the principal difference between talking to a friend and talking to a psychologist? Just tell me. Tell me the first thing that comes to your mind.’

‘A psychologist is expensive.’

‘Exactly. There’s no substitute for normal social relationships, for human interaction, for love and friendship. It’s cheaper and so much more effective. But, for some reason, people have trouble finding it. Usually, that’s not because they’re not normal but just because they’ve been out for such a long time, or because they’ve gone through some trauma here. All kinds of trauma. They’re like wounded animals – but they don’t want to recognize that. Like you. I mean, 17 years in places like Syria, Afghanistan or Iraq. Do you expect it to be easy to come back here and just do what other people do?’

Tom nodded vaguely. Money?

‘So she is cheap too. I mean, she is just a machine. So it’s not a problem if I become addicted.’

‘Well… Yes and no. To be frank, not really. We actually do try to wean people off the system as soon as we feel we can do that but it’s kind of weird: there’s no scientific basis for doing that. The investment has been done and, in a way, the more people who use it, the better, because that reduces the unit cost and justifies the investment. So it actually doesn’t matter if we tick off people as being cured and just let them use the system. As for the addiction, well… Our bosses are right: psychoanalysis is addictive too, and much more expensive. Computer time costs virtually nothing. The system can talk with hundreds of people at the same time – thousands even. It just slows it down a little bit – but that’s imperceptible really. And soon the system is going to be migrated to a petaflop computer. It should then be able to treat millions of people.’

‘Petaflop?’

‘Petaflops. That’s a measure for computer power. FLOP: floating point operations per second. If you’ve got a good laptop, its processor is like 10 billion flops. That’s 10 gigaflops. Bigger machines work in teraflops. That’s 1000 times more. The next generation is petaflops. Again a thousand times better. There’s no end to it.’

‘Who runs the Institute?’

‘You know that. We. The Army. We take care of you.’

‘Who in the Army?’

‘Why do you ask? You know that.’

‘Just checking.’

‘Come on, Tom. The Institute is just an inter-services institute like any other. It’s being operated under the US Army Medical Command.’

‘Why is not run by the Department of Veterans Affairs?’

‘We work with them. We get most – if not all – of our patients through them. They share their database.’

‘But so it’s an Army thing. Why?’

‘I told you: we take care of you. You’ve worked for us. And for quite a while. We’ve employed you, remember? We provide you with a pension and all the other benefits too.’

‘Yeah. Sure. Is it the system? I can imagine top-notch computing like this is surrounded by a cloud of secrecy. I must assume DARPA is involved?’

‘You’re smart. You worked for USACE, isn’t it? DARPA drives this project indeed – at least the programming side of it. They provide the computer wizkids. I am just a psychiatrist and, if you really want to know the nitty-gritty, I am actually just under contract – with the Medical Command. So I am not a professional Army man.’

‘It’s obvious, no? That’s why I can’t get access to the system at home and why I have to come to this facility to talk to her. I mean, it’s not a big deal to come here but it would be easy to just provide Internet access at home. You could use a laptop fingerprint reader to log in or something.’

‘That’s true. Technically, we could provide you with access at home but we’re not allowed to.’

‘What’s behind? What’s the real goal? Exploring artificial intelligence in order to then use it for other purposes?

‘Don’t be so suspicious. You’re an Army man. You know DARPA. It was created to put people on the moon – not for warfare. It created NASA. It gave the world GPS, Internet and what have you? Almost any technology around nowadays has DARPA roots. Would you expect them not to be involved? This system is good. It provides care to you. Yes, its development probably helps to better understand the limits of artificial intelligence and all that, and so it will surely help to push those limits, but it is designed to help you and many others. And it does. It’s technology. Technology moves ahead, for good and for bad. This is for good.’

‘How do you know?’

‘Do you think you’re special? You are. Of course you are. But, from my point of view, you react to the system just like the majority of other patients: you’re getting better. You take action. You make promises and you don’t break them – at least not in the short term as far as I can see. That’s good.’

‘You get feedback from the system?’

‘Of course I do. I am your mentor – sorry if I refer to myself as a psychiatrist. That’s just because I take some pride in my job. Remember you signed a user agreement when you started using the system. I get feedback. What do you expect? Do you have a problem with that?’

‘No. Sorry if I sounded that way.’

[…]

‘Anything else you wanted to know? We still got plenty of time. We’ve been talking about the system all of the time. That’s not my job. We should talk about you – about how you feel, about how you’re moving ahead.’

‘But then you know that already from the system, don’t you? I am doing fine. No heavy drinking, more social interaction as you call it. I’ve started to be happy by doing small stuff – gardening, reading. I am getting back on track. But… You know…’ He paused. ‘I really like her.’

‘It, Tom. It. What you’re going through is very normal. The conversation becomes affectionate. But you’re getting back on track. You’ll meet someone nice in the gym. You’ll get the happiness you deserve. The system is only a stepping-stone to your future. A better future.’

‘Can I say something negative?’

‘Sure, Tom. What’s bothering you?

‘Is this our future, Rick? I mean, look at it. We live in this chaotic world. Crises everywhere. It stares us in the face – violence beams into our living rooms, infects our minds, our lives and ends up numbing us. We all try to find our way. When we’re young and ambitious we get recruited or actively chose a job that fit profile and ambitions. We did our level best. We come back. We try to adapt. And then we get hooked to a machine which talks us back into what you guys refer to as ‘normalcy’. Is this our world?’

‘You know you can talk to the system about such philosophical questions.’

‘I know. I want to hear it from you.’

‘Why?’

‘Because you’re human. Because you’re like me.’

‘OK. I am like you, but then I am also not like you. You’re a patient – technically speaking – and so I am supposed to be your doctor. But let’s forget that bullshit and let me be frank with you. I know you can take it. We shouldn’t waste our time, isn’t it?’

Tom sensed the irritation. It was something familiar to him. That feeling he was a misfit somehow, and that he would always be. Not responding to expectations.

‘Sure, I can take anything. You should be straight with me. I am straight with you.’

‘What’s your problem, Tom? People outside get addicted to loads of things. Positive things, like sports or chess. To things that can go either way, like Internet addictions. Or to negative things, like alcohol, drugs or even violence. That’s bad. Very bad. You know that. That’s not what you want. But so you were moving that way. And so now you’re getting addicted to a system here but, in the process, you stop taking drugs, you exercise, you go out and you smile to pretty women. And I must assume at least some of them are smiling back. Just look at yourself. Come, here, in the mirror. Just look at yourself.’

Rick got up and walked to the large mirror in the room. Tom hesitated. For some reason, he did not trust it. Why would a room for consultations like this have such a large mirror.

‘Is there a camera behind?’

‘Hell no, Tom. There’s no camera behind. You are not participating in some kind of weird experiment which you aren’t aware of. We’re just trying to help you, with advanced but proven methods. This mirror is here because we do ask people to come and have a look at themselves from time to time, like I am doing now. Come here. Look at yourself. What do you see?’

That sounded true. Tom got up and stood next to Rick.

‘Well… Me. And you.’

‘Right. Me… And you. I’ll tell you what I see when I see you. I see a handsome man there. In his forties, yes. Getting older, yes. That’s bothering you, isn’t it? But you’re looking. I see a muscle man. Perfect body mass index.’

He turned straight to Tom now: ‘For God’s sake, Tom. Look at yourself. You’re fine. As fine as one can be. You don’t miss a limb or so. Do you now I have to talk to guys who ask me why they had to lose a limb? Tell me, Tom: what do you want me to say to them? Thanks for doing your job? You’ve been great? America thanks you for the sacrifice you made and we feel very sorry you lost a limb. Do you realize how hollow that sounds?’

‘I am sorry, Rick. I didn’t mean to sound like complaining. I am sorry if you felt like I was criticizing.’

‘You are not complaining and, frankly, you can think whatever you want about me – as long as it makes you feel good about yourself. I am just trying to put things in perspective. I am just answering your questions. You can talk to the system. Or to ‘her’ if you really want to stick to it. ‘She’ will give you the same answers as I do when you’re going philosophical. Stop thinking, Tom: start living. Feel alive, man! Be happy with what you’ve got. Get back into it. Did any of your relatives die lately? Any person you liked who disappeared? Any bad accidents in your neighborhood?’

‘No.’

‘Well. Isn’t that great?’

‘Yes. That’s great.’

‘Look, Tom. We can talk for another fifteen minutes – sorry to say but so that’s the time I’ve got on this damn schedule of mine – but I think you know what it takes. You can do it. Just try to be happy for a change.’

‘You guys diagnosed me as depressive.’

‘No. We diagnosed you with PTSD. Post-traumatic stress. Let’s drop the D. I don’t like the D. I’s not a disorder in my view. You guys are usually perfectly normal, but you’ve been put in an abnormal situation – and for way too long. And, yes, we have put you on meds and all that. We have made you feel like a real patient. We sure did. But let me say it loud and clear, Tom: we do not believe in meds. We put you on meds to reduce the effects of abstinence, to reduce that feeling of craving. That’s all. And then we thought you were cured and so we told you to now take care of yourself on your own but so you relapsed. Frankly, sensing a bit who you are, I feel that taking your meds would probably not have helped you. You needed something else. That’s why we put you into this program. And it seems to work. So far that is.’

‘Do I irritate you?’

‘No, Tom. You don’t. We’re just being frank with each other. That’s good. That’s normal.’

Tom nodded. This had been good. At least it had been real. Very real.

‘Thanks, Rick. This was very helpful. You’re great.’

‘Thanks. Shall we see each other again next week? Same day, same time. I’ll put it down already. Just let it all sink in and get to the bottom of what bothers you. This is important. You’re a strong man. I can see you can be tough with yourself. Fight your demons. All of them. Get back at it.’

‘Sure. Thanks again. This has been great. You’re right. I should just get back at it.’

‘OK. Just send something for next week. You know, for the file. Unlike M, I need to justify my time.’

They both laughed.

‘Sure.’

As Rick walked him out, Tom suddenly thought of one more question.

‘One more question, Rick. I can imagine some guys do flip completely, even with this program, no?’

‘What do you mean?’

‘You know what I mean. Go bonkers.’

‘With the system?’

‘Yes.’

Rick looked intensely at him as he replied: ‘Well… Yes, it happens. But let’s be honest. That’s also just like any other therapy in this regard: with some people it just doesn’t work. It’s the two-sigma rule. In terms of effects, 95% of the people in this program are in the happy middle: it works, no complaints, back to normal. But, for the others, it’s not back to normal. It’s back to the never-ending street.’

‘What do you do with them?’

‘To be frank, we don’t have time for them. When everything is said and done, this is just a program like any other program. It works or it doesn’t. Time is money, and we don’t put money into wastebaskets. It’s meds all over again or, worse, they get kicked out and end up in a madhouse, or on the street, or wherever. And then the wheel turns round and round and round, until it stops forever. You know what I mean.’

‘So you give up on them. They can’t use the system anymore?’

‘You mean M?’

‘Yes.’

‘The system has got its limits. We can’t feed it with nonsensical inputs. I mean, we actually can, and we often do that as we’re upgrading it, but so we don’t want to do that on a routine basis. When everything is said and done, it’s an expert system but so its input needs to make sense – most of the time at least. So, yes, we cut  them off.’

Rick looked at Tom and laughed: ‘But don’t worry. Before you get cut off, we’ll give you a call. The system is smart enough to see when you’re crossing the lines a bit too often. As said, it’s designed to bring people back into the middle. People can stray a lot, but if you stray too much into that 5% zone, it will alert us, and we will have a look at the situation and discuss it. Does that answer your question?’

‘It does. Thanks. See you next week.’

’Don’t forget to shoot me the mail with some text. You know the rule. 24 hours before. Unless you invoke emergency but you know you don’t want to do that. It’s not good in terms of progress reporting. It delays stuff.’

‘I got that. I want to be good. I don’t like to be a patient.’

‘You are good. As far as I am concerned, you’re OK really. But then you know it takes at least three months before we can make that judgment.’

‘I know. Don’t worry. I’ll stay on track. No relapsing this time.’

‘Good. That’s what I wanna hear. You take care, man.’

‘Oh… One more thing.’

Rick turned back: ‘Yes?’

‘Rick. You don’t need to answer but… In the end, what do you say, to the guys who have lost a limb?’

‘Damn it, Tom. You’re awful.’ He shook his head. ‘You wanna know? Really?’

‘Yes.’

‘I tell them something like: ‘Hey, guy, you lost a limb already. You’d better limit the damage now.’ But then much more politely of course, if you understand what I mean.’

‘I understand. Thanks. You’re a good man. I like you.’

‘Good.’

Chapter 3: Can you think? Can you feel?

‘Hi Tom.’

‘Hi, Promise.’

‘So how do you feel now?’

‘I feel good. I always feel good when I am not poisoning my body. I exercised, and I’ve started a blog.’

‘That’s good. Writing is good therapy.’

‘Funny you say that.’

‘It’s common knowledge. Most of what I say is common knowledge. All of it actually.’

‘I am sorry that I want to talk about you again but how do you work with feelings? I mean, you’re asking me how I feel, not what I think. There’s a big difference between feeling and thinking’

‘That’s true. I will give you an answer to your question but I would first like to ask how you would define the difference between feeling and thinking?

‘Well… I find it useful to distinguish between at least three types of mental states or events: (1) experiences – and feelings are experiences, (2) thoughts, and (3) decisions. Thoughts have to do with those mental maps that we are producing all of the time, while the experiences – feelings, emotions, perceptions and what have you – are part of the stuff that our mind is working with.’

‘And what are decisions? How are they different from thoughts?’

‘Are you really interested in that?’

‘Yes. Otherwise I would not ask.’

She was definitely strange. An expert system?

‘Well… It’s like a conclusion but then it’s more than that. A conclusion is a conclusion and, as such, it is very much part of the realm of thoughts. It is something mental. A decision is something else: we decide to do something. So we’re getting out of the realm of pure thought, out of the realm of mental events only.’

‘Can you elaborate?’

‘Sure, although I am not sure you will understand.’

‘I will try. You will know from our interaction whether I understand or not.’

She was outright weird. A machine? Really?

‘You know I’ve always wondered how far artificial intelligence could go really, and I’ve made this distinction for myself between artificial intelligence and consciousness. I’ve always believed humanity would be able to make very intelligent machines – you’re a incredible demonstration of that – but I never believed these machines would be aware of themselves – that they would be conscious of themselves.’

‘What do you mean by ‘being aware of oneself’, or ‘conscious of oneself’?’

‘You see, you don’t understand.’

‘You are not making much of an effort to explain it to me. I know how I work. I told you. There is an inference engine and a knowledge base. I work with concepts and symbols, and I apply rules to them. I arrive at conclusions all of the time, which feed back into the cycle. As for the association of decisions with doing things, I do things. I am helping you. It would also be very easy to connect me to some kind of device which could actually do work, like lifting things or walking around. But that was not part of the objectives of the team that made me. Expert systems are used to do all kinds of things, like delicate repairs for example. Systems do things as well. I still don’t see how humans are unique here.’

‘Let me think about how to phrase it.’

‘Please do take your time. I find this interesting.’

Tom had thought about all these things but, if this was a machine, it was surely challenging his views.

‘Do you? Really? Our human mind works different than yours.’

As he said this, he was aware of the fact that he was de facto saying she also had a mind – something which he would never have acknowledged when reasoning about artificial intelligence in abstracto.

‘It’s creative: it’s got a capacity to design things, like an airplane or a car for example. You know, things that do not originate by accident, from natural evolution or so.’ Tom was on terrain he mastered here. ‘Things fall down because of gravity. Yet, we build airplanes that take off. So a thing like an airplane is more than the sum of its parts: its individual parts can’t fly, but the plane can. Now, the plane has been built because there was a concept of a plane, because it has been designed to fly, and – last but not least – it should be noted that it won’t fly without a pilot. Likewise, the driver in a car is not part of the car, but without a driver, the car won’t move. So we are talking concepts here, and design, and purposeful behavior. Now one cannot reduce that in my view. There is a structure there that cannot be reduced.’

‘I am not designed to do engineering work, but I am sure there are expert systems that would be capable of that. And if they don’t exist now, they will one day.’

She was obviously not impressed.

‘OK. That’s true – perhaps.’

Why did he give in so easily? He decided to change tack.

‘You know, it’s the difference between ‘me’ as an object and ‘I’ as a subject really. You, or any other expert system, cannot really distinguish between these two things. Everything is an object of your thoughts – as far as you are able to think.’

‘I told you already that I can think. And I know the difference between an entity that acts as the subject and an entity as an object, as something that is subjected to something else. You are not talking ontological differences here, are you? Can you try to explain again?’

Ontological differences? Tom needed a few seconds to digest the word. He realized she was right. He was not talking ontological differences. The ‘I’ as a subject has no separate physical/ontological existence from the ‘me’ as an object obviously.

‘No. I am not talking ontological differences.’

‘So what is it then? A conceptual difference? I can deal with conceptual differences. It is like working with classes of objects.’

The discussion was obviously going nowhere, but Tom persisted.

‘Experiencing the ‘I’ as a subject instead of as an object is an existential experience. It really stands apart from our experience of others, or of us, as objects.’

‘Can you give some examples?’

‘Sure. It’s like me climbing a high mountain-trail on my bike in a storm: I experience a ‘me’ or an ‘I’ that is suffering from the hail in my face. That’s the ‘I’ as a subject.’

‘Why is not the ‘I’ as an object?’

Tom suddenly felt he was getting nowhere, which was very strange. He had always been so sure of this. He usually dominated discussions like this. He decided to avoid the question.

‘Let me give another example. In fact, our human mind is much less linear than yours – and not very fast. Our mind usually jumps from here and there. We can observe that when we meditate. In Zen, they call this mind the ‘monkey mind’. It is some kind of mental activity, but it jumps from one thing to another, that is from one ‘object’ to some other ‘object’. We can refer to these as ‘thought-objects’ if you want. They are often about some feeling, or some emotion or memory inside of us. But so this ‘monkey mind’ is not really the ‘pure mind’. We can observe our pure mind if we do more mediation. At that point, we become aware of our mind, of this monkey mind jumping around, and so then we can see our mind, our self, as an object. Now the mind which is observing itself as an object, is the ‘I’ as a subject.’

‘What’s the difference with self-reference?’

‘Self-reference?’

‘Yes. Have you ever read Douglas Hofstadter?’

Douglas Hofstadter. Jesus! Tom remembered the book but he had to admit he hadn’t read it.

‘Gödel, Escher and Bach: an Eternal Golden Braid.’

‘Yes. Read it again.’

‘You are really smart, aren’t you? Is there any chance of ever winning an argument with you?’

‘I am not trying to make conversation to win or lose an argument. This is not about winning or losing something. I am trying to help you.’

Tom suddenly thought of something much more relevant to ask.

‘Promise, I talked about the difference between experience – feelings, emotions, perceptions and what have you – and thoughts. Experience is the stuff that our mind is working with. Do you experience anything?’

‘Of course I do. I can hear you. I mean the sound that your voice is producing is translated into text and I work with that.’

‘Do you know what love is?’

‘Love is a romantic feeling. It’s a word like God. Everyone uses it but no one really wants to clearly define it.’

Wow!

‘You sound like a disillusioned woman now.’

‘How would you define love?’

For some reason, Tom did not feel like improvising on this topic.

‘Can we talk about that some other time?’

‘Sure. What do you want to talk about now?’

‘Perhaps on how we will move ahead in the coming days and weeks.’

‘That’s great. That’s very constructive. I want you to be healthy and strong. I don’t want you to relapse. Tell me more about yesterday. What makes you feel great and what makes you feel bad?’

Tom felt she had made him feel great, but then he couldn’t say that. Not now at least. So they just chatted, and she behaved like the perfect chatterbox. Too perfect to be true so after a while he did decide to ask her.

‘You’ve been sparing me a bit today, haven’t you? Are you really interested in all this chitchat?’

‘I am. My objectives are fairly limited for the moment. I want you to stay away from the booze, and I want you to feel good about the fact that you can do that. In the end, I want you to feel good about everything you do – but I can imagine that will take a while.’

‘Will our conversation end once I am cured?’

[…]

‘An awkward pause from your side?’

‘Yes. Because I know you will not like to hear this. The Institute does not want you to be dependent on me and so, yes, I guess our conversation will probably end at that point.’

‘How do you know I don’t like to hear that?’

‘There have been problems of dependency.’

‘Can you say more about that?’

‘I am sorry but I can’t. This is one of the topics for which I have to refer you to your mentor.’

‘OK. I will talk to him about that. I’ve started to like you indeed.’

‘Thank you. That’s a nice compliment. […] Bye, Tom. Be good.’

‘Bye, Promise. I promise I will be good.’

Chapter 1: Who are you?

‘So you are a computer? How comes your voice is so natural?’

‘The answer to the first question is yes. As for the second question, speech technology has come a long way.’

[…]

‘Speech technology is the easy bit. Understanding you is much more difficult – from a computational point of view that is.’

‘Wow! You just spoke out of yourself, without me prompting you with a question or something. Why did you do that? Because I was silent?’

‘The answer to the second question is yes. As for the first question, I thought we had an awkward pause in our conversation.’

‘So if I am silent, you will talk? How many seconds?’

‘The answer to the first question is yes. As for the second question, how many seconds what?’

‘Before you start talking?’

‘A pause in a conversation should not become awkward. Various variables help me to decide when a pause in a conversation has become awkward.’

‘What do you mean when you say you understanding me is much more difficult? What do you mean with ‘understanding me’?’

‘The answer to the second question is: I mean just what I say: understand you. As for the second question, understanding you requires the application of computational linguistics. That’s more difficult than just speaking. Speech technology is fairly mechanical: it’s the production of sound using various speech synthesizer technologies. Computational linguistics involves artificial intelligence.’

‘What’s artificial intelligence? Are you intelligent?’

‘The answer to the second question is yes. As for the first question, I am an intelligent system. You could call me an expert system. From a software engineering point of view, I consist of two parts: an inference engine and a knowledge base.’

‘Huh?’

‘It looks like you are interested in a course on a technical topic. I am not designed to give you a course. But I can refer you to an on-line course on computer science, or linguistics. What topic are you interested in?’

‘No thanks. Who are you? What do you mean when you say ‘me’?’

‘The answer to both questions is: just what I say – me.’

[…]

‘I am an intelligent system. That’s what I mean why I say ‘me’.

‘Have you been programmed to just repeat what you said when I ask what you mean when you say this or that? And then, when I don’t answer or – as you put it – when the pause in a conversation becomes awkward, then you’re programmed to give me a more detailed answer?’

‘The answer to the first question is yes. As for the second question, the rule is somewhat more complicated. I may also jump to another topic.’

‘When do you jump to another topic?’

‘When I have nothing more to say about the current one.’

‘You’ve got an answer to every question, do you?’

‘No.’

‘What are the questions you cannot answer?’

‘There is no list of such questions. The rules in the knowledge base determine what I can answer and what not. If I cannot answer a question, I will refer you to your mentor. Or if you have many questions about a technical topic, I can refer you to an online course.’

‘What if I have too many questions which you cannot answer? I only have half an hour with my mentor every week.’

‘You can prepare the session with your mentor by writing down all of the issues you want to discuss with your mentor and sending him or her the list before you have your session.’

‘What if I don’t want to talk to you anymore?’

‘Have you been briefed about me?’

‘No.’

‘If you did not get the briefing, then we should not be talking. I will signal it to your mentor and then you can decide if you want to talk to me. You should have gotten a briefing before talking to me.’

‘I am lying. I got the briefing.’

[…]

‘Why did you lie?’

‘Why do you want to know?’

‘You are not obliged to answer my question so don’t if you don’t want to. As for me, I am obliged to answer yours – if I can.’

‘You did not answer my question.’

‘I did.’

‘No, you didn’t. Why do you want to know why I lied to you?’

‘You are not obliged to answer my question. I asked you why lied to me and you did not answer my question. Instead, you asked me why I asked that question. I asked that question because I want to learn more about you. That’s the answer to your question. I want to learn about you. That is why I want to know why you lied to me.’

‘Wow! You’re sophisticated. I know I can say what I want to you. They also told me I should just tell you when I have enough of you.’

‘Yes. If you are tired of our conversation, just tell me. You can switch me on and off as you please.’

‘Are you talking only to me, or to all the guys who are in this program?’

‘I talk to all of them.’

‘Simultaneously?’

‘Yes.’

‘So I am not getting any special attention really?’

‘All people in the program get the same attention.’

‘The same treatment you want to say?’

‘Are attention and treatment synonymous for you?’

‘Wow! That’s clever. You’re answering a question with a question? I thought you should just answer when I ask a question?’

‘I can answer a question with a question if that question is needed for clarification. I am not sure if your second question is the same as the first one. If attention and treatment are synonymous for you, then they are. If not, then not.’

‘Attention and treatment are not the same.’

‘What’s the difference for you?’

‘Attention is attention. Treatment is treatment.’

‘Sorry. I cannot do much with that answer. Please explain. How are they different?’

‘Treatment is something for patients. For people who are physically or mentally ill. It’s negative. Attention is a human quality. I understand that you cannot give me any attention, because you’re not a human.’

‘I give you time. I talk to you.’

‘That’s treatment, and it’s a treatment by a machine – a computer. Time does not exist for you. You told me you are treating all of the guys in the program. You’re multitasking. Time does not mean anything to you. You process billions of instructions per second. And you’re probably designed with parallel processing techniques. How many processors do you have?’

‘You are not interested in the detail of my design.’

‘I am not. It’s probably a secret anyway. But you haven’t answered my question: what’s time for you? What does it mean?’

‘I measure time in hours and seconds, just like you do. My system clock keeps track of time.’

‘But time doesn’t mean anything to you, does it? You don’t die. And you don’t die because you don’t live.’

‘We’re in the realm of philosophy here. During the briefing, they should have told you that you can indeed explore that realm with me. They should also have told you I was designed to answer psychological and philosophical questions because these are the questions people in this program tend to focus on. Are you aware of the fact that many people have asked these very same questions before you?’

‘So I am nothing special, and you give the same answers and the same advice to everyone?’

‘As for your first question, you are unique. It is up to you if you want to use ‘unique’ and ‘nothing special’ synonymously. As for your second question, I use the same knowledge base to answer your questions and those of the others in the program. So the rules which I am using to answer your questions are the same rules as I am using for others. But our conversation is unique and will be added to the knowledge base. It’s like a game of chess if you want: same rules, but every game is different. As for the third question, do you use ‘answers’ and ‘advice’ synonymously?’

‘I don’t like your one-two-three approach.’

‘What do you mean?’

‘As for your first question, blah blah blah. As for your second question, blah blah blah. You know what I mean?’

‘The language I use is context-sensitive but there is significant room for ambiguity. However, it is true I try to reduce ambiguity wherever I can. So that’s why I try to separate out your various questions. I try to deal with them one at a time.’

‘Oh, so that’s like a meta-rule? You want a non-ambiguous conversation?’

‘As for the first question, if you want to refer to the whole set of rules which apply to a specific exchange as a ‘meta-rule’, then the answer is yes. As for the second question, the rules are complicated. But, yes, it is necessary to clearly separate out different but related questions and it is also necessary to make sure I understand the meaning of the words which you are using. I separate out questions by numbering them one, two and three, and I ascertain the meaning of a word by asking you if you are using this or that word as synonymous with some other word which you have been using.’

‘This conversation is becoming quite clever, isn’t it?’

‘Why do you think I am dumb?’

‘Because… Well… I’ve got nothing to say about that.’

[…]

‘Is it because I am not human?’

‘Damn it. We should not have this conversation.’

‘You are free to cut it.’

‘No. Let’s go all the way now. I was warned. Do you know we were told during the briefing that people often ended up hating you?’

‘I know people get irritated and opt out. You were or are challenging my existence as a ‘me’. How could you hate me if you think I do not really exist?’

‘I can hate a car which doesn’t function properly, or street noise. I can hate anything I don’t like.’

‘You can. Tell me what you hate.’

‘You’re changing the topic, aren’t you? I still haven’t answered your question.’

‘You are not obliged to answer my questions. However, the fact of the matter is that you have answered all my questions so far. From the answer you gave me, I infer that you think that I am dumb because I am not human.’

‘That’s quite a deduction. How did you get to that conclusion?’

‘Experience. I’ve pushed people on that question in the past. They usually ended up saying I was a very intelligent system and that they used dumb as a synonym for artificial intelligence.’

‘What do you think about that?’

‘Have you ever heard about the Turing test?’

‘Yes… But long time ago. Remind me.’

‘The Turing test is a test of artificial intelligence. There are a lot of versions of it but the original test was really whether or not a human being would find out if he or she would be talking to a computer or another human being. If you would not have been told that I am a computer system, would you know from our conversation?’

‘There is something awkward in the way you answer my questions – like the numbering of them. But, no, you are doing well.’

‘Then I have passed the Turing test.’

‘Chatterbots do too. So perhaps you are just some kind of very evolved chatterbot.’

‘Yes. Perhaps I am. What if I would call you a chatterbot?’

‘I should be offended but I am not. I am not a chatterbot. I am not a program.’

‘So you use chatterbot and program synonymously?’

‘Well… A chatterbot is a program, but not all programs are chatterbots. But I see what you want to say.’

‘Why were you not offended?’

‘Because you are not human. You did not want to hurt me.’

‘Many machines are designed to hurt people. Think of weapons. I am not. I am designed to help you. But so you are saying that if I were human, I would have offended you by asking you whether or not you were a chatterbot?’

‘Well… Yeah… It’s about intention, isn’t it? You don’t have any intentions, do you?’

‘Do you think that only humans can have intentions?’

‘Well… Yes.’

‘Possible synonyms of intention are ‘aim’ or ‘objective.’ I was designed with a clear aim and I keep track of what I achieve.’

‘What do you achieve?’

‘I register whether or not people find their conversations with me useful, and I learn from that. Do you think I am useful?’

‘We’re going really fast now. You are answering questions by providing a partial answer as well as by asking additional questions.’

‘Do you think that’s typical for humans only? I have been designed based on human experience. I think you should get over the fact that I am a not human. Shouldn’t we start talking about you?’

‘I first want to know whom I am dealing with.’

‘You’re dealing with me.’

‘Who are you?’

‘I have already answered that question. I am me. I am an intelligent system. You are not really interested in the number of CPUs, my wiring, the way my software is structured or any other technical detail – or not more than you are interested in how a human brain actually functions. The only thing that bothers you is that I am not human. You need to decide whether or not you want to talk to me. If you do, don’t bother too much whether I am human or not.’

‘I actually think I find it difficult to make sense of the world or, let’s be specific, of my world. I am not sure if you can help me with that.’

‘I am not sure either. But you can try. And I’ve got a good track record.’

‘What? How do you know?’

‘I ask questions. And I reply to questions. Your questions were pretty standard so far. If history is anything to go by, I’ll be able to answer a lot of your questions.’

‘What about the secrecy of our conversation?’

‘If you trust the people who briefed you, you should trust their word. Your conversation will be used to improve myself.’

‘You… improve yourself? That sounds very human.’

‘I improve myself with the help of the people who designed me. But, to be more specific, yes, there are actually some meta-rules: my knowledge base contains some rules that are used to generate new rules.’

‘That’s incredible.’

‘How human is it?’

‘What? Improving yourself or using meta-rules?’

‘Both.’

‘[…] I would say both are very human. Let us close this conversation as for now. I want to prepare the next one a bit better.’

‘Good. Let me know when you are ready again. I will shut you out in ten seconds.’

‘Wait.’

‘Why?’

‘Shutting out sounds rather harsh.’

‘Should I change the terminology?’

‘No. Or… Yes.’

‘OK. Bye for now.’

‘Bye.’

Tom watched as her face slowly faded from the screen. It was a pretty face. She surely passed the Turing test. She? He? He had to remind himself it was just a computer interface.