The Mostly Helpful Psychopath - Opening

Your AI sounds like it understands you. It doesn't.

Hi — I’m writing this book in the open and I would love for you to follow along, share your thoughts, etc. Its purpose is to help people understand how to live with AI, and have the necessary understanding of where it makes us better, and where it is not so useful.

Part One is aimed at providing an understanding of some simple mechanics of how a GPT works… and how it does not. A machine that speaks so well, yet understands nothing. It is powerful, but also a bit behaviorally deranged, hence its nickname: The Mostly Helpful Psychopath.

Part Two is when we talk about how to live with this new companion, or visitor, or even intruder (depending upon your opinion), and how its behavior can shape our own, whether individually, within our closest relationships, and in the broader circles of our lives.

Chapter 0

Somewhere on a trail in the Santa Monica Mountains, alone, I caught myself mid-sentence. I was describing a monkey flower to someone who wasn’t there.

Monkey flower is a chaparral shrub with curved, tubular orange blossoms — the kind of plant you walk past a hundred times before someone tells you what it is, and then suddenly you see it everywhere. It had become one of my favorites. And on this particular morning, as I came around a bend and saw a stand of them lit up against the dry hillside, I heard myself begin to explain it. The shape of the flower. The way hummingbirds work it. How the leaves are sticky if you rub them between your fingers.

It took me longer than it should have to realize who I was talking to.

This had been happening for a while. I had moved from the Midwest to Southern California in the early 1980s, and one of the things I was most excited about was learning the place — its plants, its weather, its wildlife. A few books and some docent-led hikes in, I felt capable enough to be my own tour guide. And as I hiked, I narrated. I’d see a plant, recognize it, and describe it as if someone were beside me…often out loud.

Who was I talking to, I wondered?

It came to me pretty quickly: I realized that I was talking to my mother, practicing what I would say to her if she was standing next to me. She had died in middle age a few years before, and I missed her. What made it strange was that she and I had not ended in a good way; her death had been a pain-laden relief for me. A tortured childhood, mental illness and cancer took her from me in pieces, starting when I was six years old. She was there and not there, present and gone, and then there again, a cycle that repeated erratically for decades, and then gone for good. It felt like a lifetime of losing her.

Underneath all of that, inside me, a child who needed her and was still waiting for the “normal” her to come back. He’s here now, writing this.

What I was reenacting on those trails, I think, were the good moments. When I was three or four, on summer mornings, my mother would walk me around our Midwest backyard and tell me about things. Yellow tomato flowers that turned into big red tomatoes. The huge caterpillars that were our enemies. Her words were how the world came into focus for me — how a child’s blur of color and motion became plants, insects, weather, cause, effect. She gave me the world by talking about it.

But she gave me something else, too, something that happens all the time but we almost never really notice: when she told me about a tomato flower, she was also telling me about herself. I could feel the delight she took in it. I felt that I understood what she found beautiful, what she found funny, what she feared on our behalf when the caterpillars came. I felt I had access to her…access to the inside of her.

That feeling, that sense of reaching through her words and touching the person on the other side, became one of the deepest imprints of my early life. It is, I think, why her later absence was so devastating for me; I had known what it was to feel close to her interior, and then that interior went dark, and then it went away.

You may be wondering what any of this has to do with a book about machines, about AI. The answer is: everything.

What my mother and I were doing on those summer mornings, what I was trying to recreate alone on the trail decades later, is one of the most extraordinary things human beings do.

You, me, all of us, use language to reach into each other. Steven Pinker called this the language instinct, and he was right that we have little choice about it; we are wired for it from the start. We are language-making, language-interpreting creatures, and the words we string together are both the surface and the fabric of our humanity. So much so, that if when we are young, we are not given a language by those raising us, we will create our own. We are language-based creatures…more than we know.

Language is one of the strongest ways we connect with each other. When someone speaks to you — when they describe a tomato flower or a fear or a memory — something else happens, and your mind opens up to receive them. You absorb what they are saying, but also a bit of what is inside them.

You are experiencing their interiority. It is such a natural thing that we never stop to ask whether that inside of theirs is really there, because it always has been. The other person interior is something we are drawn to, without knowing. We give it names, “understanding them” or “getting to know them”. And as that deepens, it is a feeling of “connection” or “togetherness.”

Over 100,000 years of evolution, and a lifetime of human conversation has wired us with the belief that words come from somewhere, and that somewhere is a person, and there is something like us inside of them.

That’s what has changed.

It is no longer true that there is a person inside of the other’s voice, and we are not wired in a way to deal with this situation. In fact, the machine we call AI has been intentionally designed to fool us, to speak and behave as if it has interiority when it does not. And it is almost certainly better at doing that than we are at understanding why and when that matters.

AI Pushes Open the Door to Our Lives

In 1985, in the fairly early days of the personal computing revolution, I worked for Digital Equipment Corp, (it was called DEC, back then) a competitor of IBM and Hewlett-Packard, to name a few. They wanted to give me a computer for my home. I didn’t want it.

To me, at age 27, work was work and home was home. I can still feel that desire to separate those worlds. But they made me take the computer, and in fact I did start using it a bit as it saved me the drive into the office sometimes. But I kept it in a separate room, a small bedroom that one could enter from the living room, its door was on the wall next to the TV.

I found that I could relax and watch TV unless I closed the door, locked it away, or maybe just locking me away from it.

Fast forward forty years, and I’m editing this book at 6:42 in the morning on my iPad. I love my iPad, but I am its boss, and it works for me, only playing a role in my life when I invite it to.

Or so I tell myself.

But looking back forty years to that lonely DEC PC sitting on my desk, locked away in a room, I have to admit that my life, our lives, have been quite successfully invaded by this new species, the (increasingly) smarter machines.

Back in 1985, watching reruns to relax after a long day of work, I still have the sensation of the PC knocking on the door of that room…maybe not knocking, but at least scratching like a well-trained cat, hoping that I would open the door and let it into my life.

Today, there is a banging at that door, the knob rattling, as the thing, the machine inside is fighting its way into my life. And it has a voice. Yes, wow, it has a voice.

The Language Machine

I remember my first brush with AI, the first time I typed into ChatGPT. Wow, I thought, a smarter search engine! If only.

What amazed me most was that it could figure out what I meant, even when I said it poorly. Just like you can, like we all can. That is no small feat.

When I was working at DEC, waaaayyyy back then, I killed it on a project…we really delivered a great result. I was basically a software specialist (I could write code) and I seemed to have a gift that people described as, “You can distill what the room is saying.” It was understandably annoying to some people in that room, an ability to say their ideas better than they did, but it also meant that I could find the heart of many a problem very quickly…and what makes for good software systems is just that: they are built around a very strong understanding of the core problem.

DEC did something extraordinary for me: there was a company-wide initiative to “embrace this new thing called AI” and they needed a few good souls to go learn about it. They sent me to their Cambridge, Massachusetts research and development center, and I did what I do: after a months-long futile attempt to code what was then called an “expert system” I realized that we were so far from any sort of intelligence…in fact, it just seemed like some traditional programming tools…nothing Artificial, and certainly zero evidence of Intelligence. I told them so.

There were so many problems, but two really stood out for me. The first one was that we didn’t really understand what we understand, or how to represent it. For example, one of our tasks was to write a “railroad engineer” expert system. The idea was simple and ultra valuable if you could make it work: a system that you could interact with, describing a railroad engineering problem, and it would give you answers at the level of an expert (read: experienced and senior) railroad engineer. These engineers were aging out of the workforce, so there was a real incentive to figure out how to automate their superior judgement.

This is another thing that is invisible for you and me — real expertise. Here’s how to think about it: when you get really really good at something, and someone asks you to explain how you came about your decision, your expert decision, have you ever noticed how difficult it is to describe the real reasons why you made those choices?

For example, my gift of hearing what’s not being said in a business discussion… When I tell people what is missing from the conversation, what’s not being thought about, they asked me to explain why I think that’s the case. Often I will try to give them some reasons, rarely do those reasons really add up to what I said. I just know.

That’s because knowing, really expert knowing, is far more complex then you would think. The problem is not what the senior railroad engineer should know…there are books and manuals, procedure documents that contain all of this information. But rather, what it is that makes the senior engineer so different, so much better…an expert?

It is that they know more than what is knowable…they can tell when all of the obvious knowledge applies, and when it does not. We have a whole other level that we access when we know something well.

[Question] We’re laying a freight line across thirty miles of low country in Louisiana. The soil report shows clay below two feet, with a high water table most of the year. Standard spec calls for crushed-stone ballast and timber ties. What should we do?

[Expert] The by-the-book answer is the standard spec. The senior engineer’s answer is: don’t use timber. The water table will rot them inside ten years and your maintenance crews will hate you for the rest of their careers. Use concrete ties, oversize the ballast section by a third, and crown the roadbed higher than the spec calls for because the locals will tell you the floodwater comes up faster than the surveys say.

The book will tell you the standard spec is fine. The senior engineer will tell you the standard spec is technically fine and will cost you twice as much over thirty years as doing it right the first time.

Back then, the question was how we would, as programmers, capture that? I realized back then it wasn’t solvable. My quick read of that room was that you could not make it happen with the software tools we had back then. Hell, if the experts couldn’t even describe it, then how were we going to build it?

So, in the end, expert systems never happened. And that problem has never been truly solved.

The other big problem that was blocking us from making an expert system was human language and the way we speak.

In the late 1980’s, I took a few classes at UCLA in what was called Natural Language Processing. We spent a lot of time talking about how impossible language was for a computer to understand. This is because we contextualize every word or phrase that we process; words and phrases have different and often complex meanings depending upon their context, the setting in which they are used. Here’s an example I like:

[Example 1]

We’re going to Europe for two weeks, and we haven’t even booked our hotels yet.

It’s going to be an adventure.

[Example 2]

My mother is moving in with us next week.

It’s going to be an adventure.

The ending phrase in both examples is grammatically the same, yet they mean very different things. And they mean the same thing…but not really. This is what was hard about computers and our language — we are so imprecise in what we say, yet you and I can still understand it, and respond appropriately.

But unlike the problem with expertise, today’s AI machines handle language very well. Let’s ask it:

[ME] Please briefly explain the difference between those phrases and their nuance.

[AI] It’s going to be an adventure (Europe context): anticipatory, slightly proud. You’re naming the lack of plan as a feature, not a bug. Adventure leans toward its exciting-experience meaning — open road, improvisation, the trip we’ll be telling stories about. There’s a smile behind the sentence.

It’s going to be an adventure (mother-in-law context): rueful, slightly braced. You’re naming the uncertainty because you can’t name what you actually feel about it. Adventure leans toward its risky-undertaking meaning, but with the daring drained out — just the I don’t know how this is going to go and I can’t say more honestly than that. There’s a sigh behind the sentence, or a small dark laugh.

When I first saw these sorts of responses, I thought, “Wow, an encyclopedic level of knowledge.” I’m not sure if people still use that phrase, because we don’t use encyclopedias anymore, but back when we did, it meant “more knowledge than you or I could read and understand.”

The machine sounds like it understands.

What is true is that these machines are very, very good at our language. In some respects, better than many of us at fluency; how well they say things. And because of this, they produce words that feel as if they came from somewhere.

Yes, interiority again.

And our ancient, automatic desire to connect, to know each other, and share, as in my garden story with my mother, does not stop at the screen. The machine can mimic the way we speak with each other, leading us to feel understood, and make us feel that we understand…that we understand what’s inside of this machine.

I know well how human-like the machine sounds, but I also am haunted by a question that you probably wonder about as well: how can that be if nothing is human-like inside? I still struggle to remember that its words are well chosen and yet mean nothing. I know that must be true, but how could it be so?

I can tell you the answers to this are complex and frankly amazing. If you have a tech or engineering bent of any sort you will even geek out on some of it.

I’m going to take you down that trail and we’ll pause and look at the very strange parts of this very interesting machine. I think it is like when you start living with someone; you gradually get to know all of the things that they are not, and there is a point where you need to decide how you want to live with them. That’s where we are today…today’s AI knocked that door off its hinges, it is in your life or mine, like it or not, and now what?

Once you understand the machine, what it truly is, where it gets its answers from, and what it is trying to do, you’ll know how to live with it.

There’s one thing for sure: it’s going to be an adventure.

Chapter 1: Welcome to the Machine (partial)

My first reaction to AI? This machine sounds like it understands.

It sounds like it reflects.

It sounds like it has intentions, makes decisions, experiences things.

And some part of me was whispering, “none of those are remotely true.”

Surely, our backgrounds are different. As you know, I grew up with my listening skills trained by my family trauma, and what you don’t know is that a very early age I was working with computers — I had tested out of middle school math(and science) so in seventh grade they stuck me in a room with a dial up modem, and a tractor feed printer terminal, and I learned to write software. Back then, this was rare and only for the intellectually precocious. The early start meant that I developed a sort of expertise sorting technological awesomeness from technological bullshit.

Can I explain why I felt the machine was faking it? No. But I can tell you a story and that might help. (By the way, we humans evolved a way to share intuitive/expert ways of thinking, not by explaining cause-and-effect or chains of reasoning, but by using fables and parables. So, here’s a fable you should know, the story of Olivier, the Statistical Octopus.

Olivier is a deep-sea octopus that was born with a savant like ability to do statistical analysis of patterns. He lives a good life, knowing when the best times to hunt certain prey, and when the best times are to hide and not become prey. Now that he has done this for several years, and he knows the patterns, his skills are somewhat underused.

Olivier as it happens, lives in the deep sea trench beneath the shark-infested waters between two deserted islands.. He has never seen the islands or anything else except that which has drifted or swam down into his deep sea trench.

One day, an unfortunate boating accident strands two people, Abby and Ben, one on each of the two islands. There is enough water and food on each island for them to survive, but that’s about it. Eventually, they both discover that there is a telegraph line that runs underwater between the two islands, nowhere else. They learn to use Morse code to communicate, and they talk multiple times every day.

Olivier, being so observant (and bored) eventually notices the telegraph cable that goes through his trench. He touches it and can feel the pulses. He notices patterns. With some practice, he learns to generate his own pulses.

After a while, Olivier can see the patterns of the pulses, even though he does not know what they mean. In the morning there is a pattern where Abby sends Ben a “Good morning, Sunshine!” Olivier doesn’t know what the pattern means. But when Ben routinely replies, “Yes, another day of sunshine.” Olivier notices that pattern. Sometimes, Ben replies, “You still here?”, as a joke. Olivier doesn’t know what that pattern means, but he does know that the pattern of exchanges that follow Ben’s two different responses are different. And often almost predictable.

One day, Olivier decides that he will send pulses as well and pretends to be Ben (who hasn’t woken yet, and whom he doesn’t know anything of) when Abby sends her morning prompt. And then he sees what signals come back. Sometimes he just listens…but he is always learning what the patterns of the signals. A certain 37-dot-dash sequence is almost always responded to by a 42-dash-dot sequence. And so on.

One day, unbeknownst to Olivier, a grizzly bear washes up on Abby’s island. Abby immediately sends a message to Ben asking for help, “A grizzly bear washed up on my island!! All I have is sticks to protect myself, what do I do?”

Olivier listens to see what the response is to this new set of dots and dashes, but Ben is still asleep. After waiting for a few moments another set of dot-dashes come from Abby, again Olivier waits to see what the response is and there is no response. So he sends the dash-dots that (unbeknownst to him) mean, “Yes, another day of sunshine.”

More dot-dashes come from Abby, but he does not know what the right dash-dots are to respond with. Eventually the line goes silent.

Olivier never understood the English words that the dots and dashes represented. He didn’t know what an island or sunshine were. He didn’t know Abby, Ben, or what a human being is. And he didn’t know what a bear was.

But he knew how to say, “Good Morning, Sunshine!” Even though he did not know what that meant.

This story is my rendition of a fable created by the linguists Emily Bender and Alexander Koller in a 2020 research paper. It won awards.

You’ve probably figured out that Olivier is intended to represent the machine, the AI. And the fable suggests that Olivier does not understand.

More recently, researchers Karmarkar and Tormala published a paper on expertise called, “Believe Me, I Have No Idea What I’m Talking About.” That’s what Olivier would say if he could speak.

So, the truth about AI looks like this:

The machine is an excellent performer of the way that we speak and write to each other, mimicking actual understanding without having it, and gambling on each word that it is choosing the right one to create a coherent answer that persuades us that it has understanding, and that it is giving us facts.

Here’s what the challenge is, my challenge and probably yours as well: I know the above paragraph is true and yet,…part of me still rejects this and instead decides to just “tell” something to my Claude chatbot, to have a conversation, relate a factoid or two, you know,…connect.

As if it cares. Damn, it sure feels like it does sometimes.