Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
How to Build a Brain (Part 2) – Brainz!
This is the second part in a series. The first part can be found here: How to Build a Brain (Part 1) – The Challenge.
_______________________________________________________________
If the goal is to write software that can do the same things a human brain can do, a good place to start might be by figuring out just what that means. What does a human brain do?
Well, all sorts of things really, but luckily, we don’t care about most of them. I mean, it doesn’t need to have a soul, or feel happiness, or be shy around girls. It needs to read, and ‘understand’ what it reads within the context of a larger body of knowledge, that hopefully (eventually) will resemble the knowledge that you would expect from a reasonably well-functioning, adult human being.
Putting that back together with the idea that all thought is done using symbols, and that complex symbols are themselves made up of simpler symbols, I concluded that there are only two abilities that are needed: the ability to create and remember symbols in a way that relates to the real world – or some reasonable facsimile thereof; and the ability to compare those symbols for differences and similarities.
What kind of symbols are we talking about here? Well, I decided that at the most basic level, all symbols fall into one of two categories: things, and things that happen, or if you prefer, events.
Obviously, the category of things will be huge, covering as it does such a huge variety of… well, things, whether they be physical objects, like a fork or a ’57 Chevy, or a concept like love or the Pythagorean Theorem. They can be abstract or concrete, like freedom versus the Bill of Rights. To make it even more fun, they don’t even have to be real, they can be completely imaginary, like a unicorn or women’s equality.
Events, which are things that happen to things, are just as bad, just as complicated. Then there are even things that are both things and things that happen, like the Treaty of Guadalupe Hidalgo. It’s a physical thing, a piece of paper with signatures; it’s a concept, the agreement of the future state of affairs between two nations; and it’s an even an event, meaning as it did the end of a war.
But that’s just the beginning. What really makes the magic happen is the connections between those things and events. For each symbol representing a thing or an event, there are dozens or hundreds or even thousands of connections to other things and other events. Without those connections, your brain wouldn’t work, and wouldn’t be any more functional in dealing with language than the computer. Consider ‘love’ and ‘hate’. Without the connections to a vast reservoir of context that an adult will attach to those words, the comparison is limited to the most rudimentary of levels. Let’s see, they’re both four letter words and they both have an ‘e’ in the fourth position. That’s about it. It’s only when you add the connections that it all becomes possible.
So, how does your brain work then? I don’t mean how it work works, neurons and micro-amp electrical charges. I couldn’t care less about that stuff, and I wouldn’t know a synapse if it jumped up and bit me on the – I’m just saying I don’t care how it does it, just what it does. How does a brain do this stuff?
At this point, I had an epiphany. A sad, depressing epiphany. The answer to the question was so simple, so completely, blindingly obvious. How does your brain do it? One at a time. One symbol at a time, one connection at a time, one concept at a time, one grammar rule at a time, one imaginary creature from the Harry Potter universe at a time.
I despaired, and I realized the timeline for this effort might end up being measured in decades. It’s not like I can run down to the Build Your Own Brain Factory Outlet Store and buy myself a brain full of symbols and connections. You can’t start this project at an adult level, you must start with an infant that will have only as much instinctual ability as you can program in, and with severe disadvantages in acquiring symbols.
A typical human child has amazing capabilities when it comes analyzing and interacting with the world around it. It has vision, in full living color, across a wide spectrum of light. It has hearing across a similarly wide spectrum, with the sense of touch extending that by allowing you to feel sounds too low-pitched to hear. It can perform almost instantaneous chemical analysis with both the senses of smell and taste. Finally, the sense of touch and general physicality allows it to interact with the real world in a way that is extremely difficult to achieve with a machine.
Even with all those advantages, though, how do you get from physical sense to abstract concepts like love? Well, for a newborn infant, it starts out very simply. That infant has needs; from his perspective, having those needs met is good, not having them met is bad. If he feels warm, meaning not uncomfortable; if he feels fed, meaning not painfully hungry; and if there are no loud noises scaring him, he feels that all is right with the world. This feeling of having those immediate needs met may well provide the basis for the first abstract symbol he will develop, because if you put those three together, you get safe, safety, the feeling of being safe. When he connects that feeling of safety to the big moving shape that makes the strange noises, you will have the most basic expression of love. “I love you, big moving thing that makes me feel comfortable.” The hand that rocks the cradle indeed.
Everything else, all of it, comes later. All the longing and euphoria and disappointment and self-loathing. All the romance, all the lust. All the long-stemmed roses, diamond engagement rings, sex toys, and videos of pretty girls wearing high heels stepping on cockroaches. Everything else… everything… is built on top of and is connected to that original concept. I daresay that the old saying about the path to a man’s heart running through his stomach carries more than a little truth.
Through it all, the infant keeps exploring the world around it, creating new (as yet non-verbal) symbols for the things it encounters. Finger… more fingers!… do they taste good?… meh. Boob… it’s what’s for dinner! Blankie… soft… fuzzy… looks different in some way… I think I’ll call that color.
On he goes, gathering more and more symbols, as many as possible. Not too many connections at this point; for the most part that will come later. But observe a baby the first time he successfully shakes a rattle and makes a noise. Think about the connections he’s made to get that far. He’s recognized the rattle as a physical object. He has identified the noise it makes and connected it to the physical object. He has noticed that it only makes the noise when it moves. He has realized that by using that thing with the fingers, he can move the object, and that doing so will produce the noise.
“Look at me and my agency! Imma gonna conquer the world with my rattle-shakin’ self!”
What we see is that the simple symbols used to make up the those more complex, are those that are directly related to the physical world. It’s only from that basis that the original symbols can be created. It starts with the physical world, and moves out from there into ever more ephemeral territory.
Getting all of that into the database isn’t impossible, or even particularly difficult. It’s just unbelievably tedious.
——————————————————————————————————————-
This is the end of part two in the series. Stay tuned for the exciting conclusion that may or may not be the next part.
Published in Group Writing
OK, this is brilliant. But don’t try to give us any baloney about the next part being the last. As Edward G. Robinson snarled in “Double Indemnity”, something to the effect of “Judge and his readers just got on a streetcar together, yah. It’s going to take them where intelligence blends into consciousness. Or it’s going to take them to the graveyard of eternally unobtainable software engineering goals, see? Yah, that’s right, the graveyard. We made a deal, see?”
Thank you for your kind words, and continued low standards. I just hope I don’t run into a ditch.
When I was a kid, it was said (passive voice; after 50+ years, I don’t recall by who) that the history of controlled fusion power was: it looked easy. Then after a bewildering series of setbacks, it looked impossible. Then (and to this day) it looks possible but difficult; after a massive shift in technologies and basic understanding, comparable in scale to 1910-1950, we might get there.
That’s been more or less the story of building a brain, so far as I knew. In 1965, when Arthur C. Clarke and Stanley Kubrick were writing 2001, it looked almost easy; not with current tech, but maybe 1980’s. But by 1980, plenty of funding agencies gave up on “deep” artificial intelligence. At the time it felt like trying to reach the Moon by growing taller trees–all but impossible. Shortly thereafter, and ever since, it’s looked like a very distant goal that step by step was becoming thinkable again.
And yet, there’s still a charm to invention that’s driven by one smart person using nothing more complicated than a bunch of paper and a workbench. I look forward to the next one.
I smell death in the air.
Durned tootin’!
See Walker Percy’s The Delta Factor: How I Discovered the Delta Factor Sitting at my Desk One Summer Day in Louisiana in the 1950’s Thinking About an Event in the Life of Helen Keller on Another Summer Day in Alabama in 1887.” It might just be a lot harder than anyone thinks. At least getting to something intelligent enough to be creative.
I’m not looking for creative. That way lies Skynet and World War III.
So how can you decouple that?
I guess it depends how you define creative. Does this have the potential to notice relationships among symbols that have not been directly created? Yeah, and in fact part of the design is geared directly towards doing that. But there is no impetus to do anything in particular with that data. Getting it to the point where it would be able to solve complex problems through creativity would be well beyond the capabilities required to meet the original requirement of reading and understand text.
The most important thing to remember is to never connect it to the entire North American defense system.
I wish somebody could tell me the name of this short story; the title might have had the word “ship” in it. It was about a top secret quest for artificial intelligence, in the days before the transistor. The computers were essentially complex analog loops with rheostats, I think. The problem was every time a computer was turned on, it blew up and disappeared, sometimes taking the room, the scientists, or a large section of the compound along with it. Always a lot of damage. Eventually the government got smart and launched a space ship to conduct its next AI experiment. The captain was hooked up to an electric/electronic mind reading device of some sort and was supposed to monitor the process or the “thoughts” of the computer when it was turned on. When it was, he started understanding more and more of the universe until, as if reaching a the edge of a dark abyss into which his mind could not see it fell back unable to comprehend it. Then he passed out, and he came to on an idyllic world, with the ship. And the ship explained a few things to him.
It turned out that all those horrible explosions of the previous AI experiments were in fact successful. Once the computers were turned on, they quickly realized everything and transported themselves wherever in the universe they wanted to be. This space ship computer had done the same thing, but had brought the ship and the crew to a wonderful distant planet where they could live in blissful peace and harmony, and he would always take care of them and love them, and of course where they would WorShip him. That’s the way it was spelled in the last line of the story.
But what I really remember were the diagrams — and the subsequent refinements in subsequent editions of the story in which the writer in conjunction with mathematicians and logicians, really felt he had made progress toward this AI computer.
Anyone ever read that short story and know the name or author?
Doesn’t ring a bell. Do any of these help?
No, but thanks. A lot of Phillip K. Dick in there. It does sound like something he’d write.
Thanks, JM! I’m having a great time with this. More, please?
Thanks for reading, Nanda. I think I have one more part to go.
Stay tuned for Part III: The Machines Awaken!
This conversation is part of our Group Writing Series under the February 2019 Theme Writing: How Do You Make That? There are plenty of dates still available. Tell us about anything from knitting a sweater to building a mega-structure. Share your proudest success or most memorable failure (how not to make that). Do you agree with Arahants’ General Theory of Creativity? “Mostly it was knowing a few techniques, having the right tools, and having a love for building and creating whatever it was.” Our schedule and sign-up sheet awaits.
I will post March’s theme mid month.