The Evolution of Creatures with Steve Grand Games Jessica | 11/11/2013|| 10|
An article from Steve Grand all about the evolution of Creatures and Grandroids, with plenty of interesting ideas!
Submitted by Steve Grand for the CCSF 2013
Hi everybody! Happy CCSF 2013!
Jessica asked if I would write something for you about Norn genetics and brain lobes and suchlike, but the scary truth is, I can't remember much about it any more! It's twenty years ago, darn it. My head is now filled with a new kind of genetics, a new kind of biochemistry, and a very, very complex and rapidly developing theory of the brain. All the mental space that used to house Norn biology has been cleared out, repainted and decorated with fresh code, ready for my new "babies." You guys would have to remind me how it all works!
So instead of talking about Norn evolution I thought I'd tell you a bit about the evolution of Creatures itself. You see, for me Creatures was just a step on a very long journey; a sort of half-way point, where I had a moment to rest on my laurels and enjoy the view for a while before setting off again.
To understand this journey I have to take you back into the dim and murky past, probably before any of you were born. Back to the year in which Armstrong and Aldrin first set foot on the Moon, in fact, and a skinny ginger-haired boy, with little screwed up eyes that looked (and felt) equal parts puzzled and annoyed, turned ten years old. That year, my dad made me an "electricity and magnetism kit" for my birthday. I don't really remember what was in it now, but he'd scavenged various magnets and metals and light bulbs and wire, and written a little booklet with some experiments in it, which I probably never followed. The reason we have to go back to the Swinging Sixties to begin is because of what I learned from that box of wonders.
My favorite toy at the time, you won't be surprised to hear, was Lego. So I already knew all about building blocks. I knew how the interface (as we'd call it now) of lumps and holes on the faces of Lego bricks meant that they could be plugged together to make a huge range of different things. But my Lego models just sat there and didn't really do anything. On the other hand, the objects inside my electricity and magnetism set, and the other cool gizmos that came soon after it (Dad was an electronics engineer at the time), were dynamic building blocks. They, too, could be plugged together in lots of different ways, but instead of making different structures, they made different kinds of behavior. Sometimes this was very odd and startling behavior, and what's more it was rarely the behavior I intended.
This was pretty thrilling and perplexing stuff! And remember we're talking about quite ordinary, everyday-looking objects, not the billion-transistor, inscrutable microchips of today. In those days a capacitor was not a tiny black dot but two strips of kitchen foil rolled up either side of a sheet of waxed paper. A resistor was a thin wire wrapped around a ceramic tube, and the olden-day equivalent of a transistor was a small glass bottle containing a light bulb filament and some metal gauze. But it worked. Some kitchen foil, a coil of wire, some coal and a very, very long antenna (mum's clothes line was perfect) were enough to make a radio that could pick up signals from Russia. If I'm ever stranded on a desert island I'll be fine, as long as I have some kitchen foil...
Anyway, I shall spare you my teens, much as I wish I could have avoided them myself. I just needed to start with one of the first things that generated a life-long interest in the building blocks of life and the mind, taught me a lot about complexity and emergence, and made Creatures what it is. The way I think is rather different from the way most programmers think, I've found, and I owe much of it to that box of delights.
By the time I was nineteen I had left the tinfoil behind and was training to be a teacher, but the truth is I was more interested in the component parts of children's minds than I was in lion taming. Before I gave up my fledgling career, though, I happened upon a book in the college library that was about a strange new idea called artificial intelligence. As soon as I started reading it, my little screwed-up eyes became even more puzzled and annoyed than usual. At the time, AI was all about symbolic representations. The human brain, it was argued, somehow converted sensory signals from the eyes into logical propositions, such as in(fridge, eggs), and our behavior was supposedly encoded in statements like "if(in(container, target)) then begin: open(container)..." My eyes were deeply puzzled, because I hadn't realized that apparently educated people could believe anything quite so stupid, and I was also rather annoyed that they could get away with writing books about it. So I decided to write a program or two of my own and see if there was a different way to go about it.
This was around 1977 to '78, when the "home computer" was just starting to be an achievable reality, albeit a completely pointless and exasperating one. Programs in those days looked something like, "3C 2F 22 4F 1D 00," and although programming in machine code like this was definitely an exercise in the elegant use of tiny building blocks that I'm very glad I had the chance to grow up with, it did hamper one's ambitions somewhat! Nevertheless, I did some experiments with networks of simulated nerve cells, and wrote a program for evolving different kinds of carnivorous plants. If they hadn't been screwed up so tight, you'd probably have seen a little glimmer of Creatures in my eyes around then, but I didn't know it myself yet.
Unfortunately, even carnivorous plants can't keep the wolf from the door, so when I gave up trying to be a teacher I instead started writing software for schools, a few of which were just beginning to experiment with using computers in the classroom. At the time, most educational software was trying to teach facts, often under the feeble disguise of a video game, but I was more interested in making open-ended tools for children to use. I didn't want the computer to be yet another kind of teacher, or worse still an examiner; I wanted it to be an enabler. So at first I wrote word processors and art packages and math tools that young children could use to explore information and express their own ideas. But as soon as I saw the first ever adventure game, "Colossal Cave," I knew what it was that I really wanted to do.
By then, thanks to watching a house fly washing its face on my TV screen (long story!), I had some new ideas about how to plug together large numbers of tiny software building blocks to make text-based virtual worlds and simulated people to inhabit them, in a way that would be much more robust and open-ended than the traditional branching story of an adventure game. So I set to work making some little worlds for children to explore. I forget what the first one was about, but it quickly dawned on me that I could write a strange kind of computer language for making more of them. So I did. I used this language to create a world starring my mum, when she was a little girl during the War. I created a little boy called Kenhirkhopeshef (Shef, for short), who lived in the Valley of the Kings in ancient Egypt, and some other children who'd lived in the village where I'd gone to school, but 70 years before I was born. I had big plans to make things like a working nuclear power station for children to manage, and a virtual chemistry lab where they could do experiments. I saw simulation as a big future for education, and I still do.
During this time I’d managed to persuade a company for which I'd written a port of someone else's word processor that these open-ended simulations had some value, even though I couldn't tell anyone what the children would actually choose to do with them. This company (Logotron) licensed the programs to the BBC, to go with a children's television series called Landmarks, and my wife and I produced a few more of them over the next year or so. But then Logotron suddenly split up into an educational software company and a games company, and I kind of fell down the gap in the middle. I can't quite remember the timing any more but I remember proposing an idea about a simulation of little alien creatures, inspired by a wonderful book called The Planiverse, but this was to a director of the education side of the company and it didn't fly. He still has my copy of the book, but that's as far as it went. I was really desperate for work, and so I was willing to go with whichever side wanted a cheap programmer. I knew nothing about the games industry and I didn't play computer games myself, but here I was with a simulation engine going begging. I plucked up the nerve to show it to one of the people at the new games company, and he took one look at it and said, "That looks boring. Not interested."
And that's how I became a games programmer!
What have we got so far? Bottom-up, brick-based, biologically inspired artificial intelligence? Check. Artificial life? Check, sort of. An open-ended approach to simulation, in which the user makes up her own storyline? Check. Graphics? Oh. No, just text. Sorry. Do you want graphics? I can do graphics. Please let me extend my simulation engine to use graphics...
And thus, after a brief foray into converting other people's games to run on the newly invented PC, Robin Hood was born. Or rather Cowboys and Injuns, which became Robin Hood about half-way through the project because a related movie was about to come out. The game consisted of an isometric world made from little cubic tiles, and a series of tiny (16-pixel high) people, each of who had their own needs and interests, and likely ways of reacting to events. The user could cast various "magic spells" but they couldn't really control anything directly, only influence it. Come to that, nor could I! All I could do was set up short scenes and exchanges that I hoped would occasionally link together into something approaching a plot. I hoped that everyone would find their own way to influence things, so that the Merry Men would triumph, Robin and Marian would marry, and all would be well in Sherwood.
Happily it went down quite well with reviewers, which was quite a lucky break for a game involving a sixteen-pixel romance and no photon torpedoes. My bosses somehow managed to draw the attention of Maxis, who were doing very well with Sim City at the time. So I wrote another game using the same engine but this time based in ancient Rome, and Maxis became the US publisher. For the last couple of weeks of the project I went to California to work at their offices, and met Will Wright, the creator of Sim City. Will showed me a project he was just starting to experiment with, called Dolls House. This is what, years later, became The Sims, and it was clear that Will's interests and mine had a lot in common. So after I finished up Rome and came home, Maxis asked whether we had any new ideas.
What? New ideas? But I've only just... Never mind. Sure, yes, we have LOTS of ideas. Um, could you just hold the line for a moment?
My boss Ian was over in the States at the time, and suggested off-the-cuff that we had a great idea for some kind of game involving little animals. Happily, at the very same moment, I was hurriedly putting together a scenario that I called "Little Computer Ewoks," the title of which was a reference not just to Star Wars but also to a wonderful early Macintosh game called Little Computer People, which Ian had introduced me to. As the "Ewoks" idea developed further, I began to call it "Small Furry Creatures" instead, which was another reference, this time to the "small furry creatures from Alpha Centauri," from Douglas Adams's book, the Hitchhiker's Guide to the Galaxy. I didn't know Douglas personally at the time but I was a big fan. It was sheer coincidence that this game would eventually bring the two of us together, because by then it had yet another name. Luckily the creatures were still small and furry.
Anyway, "Small Furry Creatures" gradually got shortened to "Creatures," and the rest, as they say, is history.
It was just another nine-month project, at first, and originally I'd suggested including some kind of simple neural network in order to make the creatures' behavior a bit more interesting and unpredictable. It was just the kind of thing I was used to doing from my experiments back in the 1970's, so I didn't think much of it at the time. But the neural stuff quickly became more complex, and I needed a way to provide all the feedback that the Norns' brains would use for learning and for modulating their behavior, so eventually I added a simple simulation of biochemistry as well. And then I realized that I had a really neat toolkit of dynamical building blocks but no easy way to define how they should be plugged together into structures. Plus I kept changing my mind and so I needed a more flexible way to reconfigure them. So I added genetics to provide a kind of programming language for defining structures made from chemicals and neurons, which in turn were a kind of programming language for creating something that could basically program itself. And somewhere during the middle of all this increasingly biological stuff I started to develop a really strong conviction, from which there was no going back: I decided it was futile and wrong to try to fool people into thinking they were playing with real living creatures when they weren't, so I would just have to go ahead and do my honest best to create life.
And there you have it: an innocent need to add some complexity to the behavior of the little creatures in just another ordinary nine-month video game project, for which no tinfoil was even really required, turned into a five-year intense personal odyssey about what it means to be alive...
That odyssey continues today. I'm not done with this yet. One of the things you guys have taught me, and for which I shall be eternally grateful, is that this stuff matters. The marketing people at Maxis told me to "hide the science, because people don't want to know about science." But they were wrong. Boy, were they wrong. Even biologists working on the human genome project wanted to know about the science, never mind all the kids who later went on to pursue scientific or medical careers, in part because they once studied Norns. Not only do people want to know about the science, they also want to think about the ethical and moral questions. They want to think about what artificial life says about us. They want to tell stories and care, to breed and study, to mod the game and build a community. Authors have written novels. Sociologists have written papers. Feminists have argued about identity. Educators have used the game to teach biology or help autistic children. Religious people have debated evolution...
So I'm still working on it. The biggest disappointment for me (and you might want to close your eyes for the next few sentences) was that Norns don't really think. They learn, they react, they make choices, but they don't actually think. Not in the way that you might think about where to go on vacation next summer. Not in the way you might worry about something, or fear the worst, or wonder why someone did what they did. This is important, because many of the things we care about, when it comes to other living beings, depend on whether they have this ability. When we wash our hands we don't feel particularly sorry for the millions of bacteria we kill, largely because we don't think they can feel sorry for themselves either. We haven't dashed their hopes or scared them about what's about to happen to them. They just stop metabolizing.
I want to know what has to be true for this special kind of existence to happen. Obviously it has something to do with imagination. If you've ever tried mindfulness meditation you'll have a small sense what it might be like not to imagine - to just BE. But even while meditating we still have a sense of the future and the past. We can't get rid of it. "Just being" may be restful now and again, but it's not really much of an existence. It's like reading a story one letter at a time and not ever being able to remember any of the previous letters, only the one you're currently looking at. Right now it's just an "e". All on its own. Nothing came before it and it'll be lost by the time anything comes after it. You don't have an opinion on it or think anything about it; it's just an "e." Whereas, whatever normal waking consciousness is, it lives inside the rich narrative of our imagination, filled with possible futures and past remembrance. We create a virtual world inside our heads - not just one letter at a time but a continuous story about what has just been, what we think is coming next, and often about things that haven't even happened and probably never will. This is where "we" live.
But not Norns. They just are. They feel pain and respond to it. They try to avoid it. But they don't "feel" and "try" in the way that you and I do. It's more like the way a plant tries to avoid shade. So what do Norns lack that we have? Which other creatures on earth have it besides us? How does imagination work? Does having one MAKE you conscious, or just provide some of the ingredients for it? These are not just big philosophical and moral questions, they matter even if all you want to do is play around with some cute artificial pets. You all know how stupid Norns are, bless them. They can't make plans or understand threats. They feel no relief or remorse or worry or embarrassment. They don't respond to what might happen, only to what has just happened. This ability to live in a virtual world inside your head, which can disconnect from reality and fast-forward or rewind over experience at will, is a really big deal. Dogs clearly have a moderate amount of it but snakes don't seem to have much, if any, and that's a big part of the reason dogs are much more fun to know. But how does it work? How could we make an artificial creature that has it?
I think I know some of the answers to this now, thanks in part to a scruffy little robot called Lucy. I don't know all of them yet by any means, but certainly the important ones that will one day make the others possible. It's taken me 15 years to figure out, and it's all about dynamical building blocks, of course. But they're bigger, more complex ones with some very interesting features. It seems to me that there is a really clever module inside our brains, repeated many times over. It seems to do very different things in different places - one of them might move our eyes while another can guess whether someone is telling the truth. One keeps us upright when we're walking on a slope while another can plan the route home. But like Lego bricks, at some level they're all the same thing. They can be plugged together in lots of different ways. And one of the most remarkable things they can do is switch from making things happen to pretending that they've already happened. In fact the two processes act in concert, and together they create the glimmerings of thoughts and imagining. This is all really, really, really, REALLY complicated, trust me. When my next game comes out, poking around inside the brain and making sense of what the scanners and electrodes are telling you is going to be quite a challenge. But it's worth the complexity, because even though they only have a little bit of their brains working yet, the new creatures are already turning out to be far more alive and interesting and fun to be around than their predecessors. It won't be long before I start feeling really bad about deleting them.
If neuroscience isn't your thing, I have to tell you the chemical building blocks are now much more complex too. You know as well as I do how limited (or perhaps the word should be 'unlimited') Norn biochemistry is. So this time I've made a "proper" chemistry. An enzyme with the structure ABXCD can only ever turn chemical ABCD into AB and CD. It gives out energy in the process, but not as much as you'd get from breaking down ABBC, while creating a big molecule (like ABXCD itself) from smaller ones consumes energy instead of liberating it. The rules are fairly simple but it's much more realistic. If your creature is dying because of the buildup of a certain toxin, you can probably figure out what drugs might cure it, but either you have to find a source in nature or you have to figure out how to make some from other ingredients. Immortal creatures and those that don't need to eat MAY still be theoretically possible, it's hard to know, but they're at least not going to result from a single mutation.
Which brings me to genes. These are much more complex too, but in a way they're easier to understand because they're just like the ones we all learned about in school. In no time people are going to be discussing whether their creature is homozygous dominant for trait X, or find themselves busy researching various polymorphisms. There's far too much to tell you about, but the upshot for the genetically incurious is that breeding is more realistic, more surprising and more varied. I have no idea what you guys will do to the gene pool with your breeding experiments, nor even what is possible, but I'm looking forward to finding out.
And of course the new game is 3D, the world is much more complex and rich, the extensibility possibilities are far greater, and the creatures have to learn far, far more than just "press food --> less hunger." They have to learn about their territory, how to get home, where they left their car keys, how to walk properly, what things are good for, how to pick them up, and so on. They’ll know you exist but whether they'll know THEY exist is anybody’s guess. But in any case we're talking about a major leap forward in artificial life, I think, and some big new questions to answer about everything from consciousness to how to care for these complex creatures. I'm exhausted from working and thinking so hard over the past few years, and there's still a long way to go yet, but I'm also pretty excited and I hope you will be too.
But anyway, this is not meant to be an advertisement for my next product. I just wanted to show that there is a kind of logic to it all and the idea has evolved over time. I've been doing this kind of thing, one way or another, for more than 35 years now, and Creatures was a big and heartwarming step on that journey, during which time I was able to get to know some of you. I don't write games, I create life. You write the games. It blows my mind what you all did with Creatures, and I'm so looking forward to seeing what you (or maybe in some cases your kids) will do with the next generation. As long as my head doesn't explode in the meantime, I'll catch you next year and maybe then I'll get you up to speed on some new genetics while I can still remember how it all works!
Wow! How many gaming communities can boast about something like this? I really enjoyed this article, it was an insightful and exciting read. A big thank you to Steve Grand!
Steve was awesome and put this together pretty quickly. I was expecting a paragraph or two, but I was very pleasantly surprised! This is a really unique piece that ties into his ideas. Some might complain that it takes ages to make a game when one thinks like this... But I'm all for waiting for Grandroids, knowing that Steve has so much going into it!
Is it weird that the most exciting thing about this for me was the bit about diploid creatures? Heck yes, sign me up.
I haven't visited the Grandroids backer site in ages... I should go see what all's new.
Sweet mother of marzipan. Wonderful article, I must say! (Got me thinking before noon on a holiday, that's definitely an achievement) Interesting look into things.
Wow it's awesome that Steve Grand himself agreed to write an article for CCSF It's quite long (not in a bad way at all) so I'll have to read it when I have more time.