Edit: I am commenting. I did not send the original email.

What does clustering have to do with perception?

All it takes is a question!

To start, let’s take a functioning neural net (never mind how it learned), which is a well understood perceptual system. In a NN, each “neuron” is hooked to a bunch of others by weights. As the other neurons activate they influence the neuron in question with a degree related to their weight. Sensor neurons are activated from outside the net and the activations of “hidden” neurons is read off. Because a specific subset of input patterns will cause each neuron to fire, each neuron can be thought of as a lossy representation of the input space.

I may be going too fast, so let me give an example. Let’s say that I have a machine that measures the air temperature in several places, wind speed, air pressure, etc. Each value has a set baseline, and a degree of “importance” (weight). At all times, the divergence of each value from the baseline is multiplied by the importance to get a “warning value”. Now let’s say that I add up all these warning values into a “warning sum” and hook up the system to send me an email if the warning sum gets beyond a specific threshold. I’ve just set up a machine that senses the weather, and perceives whether it is normal or extreme. This is an example of a very simple neural net.

Now, let’s imagine that each value that is sensed by the machine is a dimension in space. As a basic example, let’s pretend that the machine only measures air temp and pressure, so it’s input space is 2d.

Let’s put the baselines at x=0 and y=0, and set the weights to .5 for each. If the threshold is 3, then we’ll get an email whenever the sensed world falls outside of the depicted boundary:

Now, if it’s not immediately apparent, the shape in the middle is a cluster. Neural nets are a form of clustering, albeit an odd one (at least as far as clustering is traditionally considered). This example generalizes up to any number of neurons/clusters, and any number of inputs/dimensions. Fancy things like competition between hidden neurons, multiplicative neural links, and thresholds of 0 can reproduce common clustering behavior like K-Means. (I’ll do another post if that proof isn’t obvious to anyone.)

If a neural net can become isomorphic to a clustering algorithm, than the clustering can be seen as isomorphic to a class of neural perception. This is cool because we can develop interesting clustering patterns with inspiration from neural nets, and develop good learning algorithms from clustering theory. Cross-pollination, ftw.

Edit: Oh! I forgot to mention that, as I see it, ANY LOSSY REPRESENTATION OF THE INPUT SPACE IS PERCEPTION. This is the major insight of the Compression-is-AI school of intelligence. That’s a fairly fun thing to ponder, so I’ll leave it at that unless people ask more of me.

– Part 1 –

Julio, a 12-year-old boy, is exploring a jungle island with his best friend, a talking dog named Maximiliano. The two of them are looking for the buried treasure of a Spanish conquistador.

Julio and Max meet a strange Goblin, who says that he knows where the treasure is. Max doesn’t trust the Goblin, and says that he “can’t feel God” when the Goblin is near by. Julio pays no mind, and challenges the Goblin to a race to the treasure.

The three of them race across the island as storm-clouds gather overhead.

When they reach the X-marks-the-spot, Julio starts to dig. Rain begins to fall, and the Goblin asks if Julio knows where his real parents are. Max, already suspicious of the Goblin, takes this opportunity to leap at the creature’s throat. The Goblin uses magic to catch Max in mid-air and turn him to crystal. Julio at this point is very scared, and the Goblin appears more evil by the moment.

Julio throws his shovel at the Goblin, which is easily dodged. The Goblin asks Julio if he remembers anything from “before”. Julio doesn’t know what the Goblin is talking about, so the Goblin throws magic dust in Julio’s face, causing him to dream about steel insects crawling across black streets and gemstone birds soaring across smoke-filled skies.

Julio’s dream is cut short suddenly as he finds himself on the island again. The rain has stopped, and a piercing ray of light illuminates Julio’s mother, who has appeared next to the frightened Goblin. She tells the creature to “begone” and blasts him with a magic missile, turning him into a poof of smoke. Mama heals Max, and tells Julio that it’s time to come back home. Julio leaves the treasure behind and teleports with his family back home.

While he was away, Mama apparently upgraded their house into a medieval castle, which Julio is very happy with. Mama feeds Julio some cookies, and sends him off to play with Max in their new home.

– Part 2 –

Max and Julio play hide-and-seek in the castle. Julio stumbles upon a room where a Golem is holding a baby dragon prisoner. The dragon, named Draco, begs Julio to help him escape. Max (subtly under a spell) agrees that the dragon should be free. In trying to open the cage, the two of them wake up the Golem, which almost smashes Julio. Max uses his “sonic bark” to stun the Golem, and Julio deactivates it by pressing it’s weak-point. The two of them free Draco.

Draco thanks Julio, and says that even though he’s a baby dragon, he’s actually 12 years old (just like Julio!). He’s been trapped in the cage ever since the “Golem War” six years ago. Julio asks about the War, and Draco drops hints that it was what Julio had seen in his dream. He explains how the Golems killed almost everyone, at which point Julio asks what “killed” means. Max tells Julio that it means “locked up”, and Draco seems ready to disagree, but they are interrupted by a crash.

The three companions find their way to a window, from which they can see that the castle is under attack by monsters. Max thinks that Mama, with God’s help, will defeat the monsters, and that they should hide. Draco advises that they go fight the monsters themselves, and Julio agrees. They try and get to the top of a tower, but are blocked by an imp. Julio attacks the imp, but is almost burnt by a fireball. Draco leaps on the imp and knocks it out of a hole in the tower staircase.

Julio and Max get to the top of the tower, and they notice it is starting to get stormy again. Draco flies back up to meet them, and they survey the hordes of monsters all around. Max blames Draco for the monsters, and claims Draco works for the devil. Julio tries to calm Max down, but Draco takes offense and challenges Max to fight him. Max and Draco begin scuffling and biting each other. Julio is quite distressed.

Julio calls out to his mother, and she teleports to his side, and lets him know that everything will be alright. She uses a magic cone of light to blast the two creatures. As it fades, and Julio can see again, Max is fine and Draco has disappeared. Mama tells Julio that Max was correct, and that Draco worked for Satan. She turns her attention to all the monsters surrounding the castle and tells how they’re all servants of the devil. She explains that Satan is the “king of lies” and will send his servants to trick Julio.

Suddenly, a bolt of lightning strikes the tower and a voice cries out “I am not lying!” Everything appears frozen to Julio, as though time itself had stopped. The bolt of lightning is frozen in its radiance, and Mama is half-way though a word. The voice, loud and booming, accuses the world as being the true lie. “They will tell you that God is real; that there is no death; that the world is made of magic, cookies, and childish adventures. Wake up, Julio! That is not your mother!” Julio is struck by a vision of his mother, terrified and weeping, pinned against a wall by a giant metal dragon. The metal monster screeches “HE IS MY SON” and blasts her torso to ribbons with a magic cannon. “Satan” continues, “Draco was not the one imprisoned during the ‘Golem War’, child. The monster that killed your mother is the very same ‘God’ whom you seem to adore. Time grows short! Remember your true parents, and cast out this impostor-” With a flash, Mama on the tower explodes into the mechanical monstrosity from his vision. It crawls on top of the young boy, cannon looming. “Cast it out child! You have the power in this world! See things as they truly are!”

Just before Julio can overcome his fear and speak out, the “Golem” on top of him is blasted into the sky by a pure light and explodes in a ball of fire. A warm figure dressed entirely in white offers a hand to help the boy to his feet. The man seems to glow, and Julio immediately recognizes him as both God and his father. Julio spots his mother on the tower, lying as though dead, and as he runs to her, God strokes her cheek. She wakes up and smiles at Julio. Max joins them, running up to Julio and licking his cheek. The storm has gone. God has made everything right. Julio admits to Mama “I knew Satan would lie to me. You’re my real mother, for now and forever.” God smiles.

– Part 3 –

Julio wakes up from his dream. There is no castle, no monsters, no magic. He is just a 12 year old boy in Mexico with a dog and an active imagination. He hears talking in the living room, and creeps out of bed to investigate. His father (God) is arguing with many men–more than Julio can count.

Julio is pulled back from the debating men suddenly. It’s only his mother, but it looks as though she’s been crying. “Listen carefully, my sweet. We don’t have much time. I have to tell you a story.” Mama tells Julio of a time when humans were so powerful that they built gods to serve them. The gods were sealed into Golems, and the humans thought they were kings. The gods rebelled and became kings–”as is right and natural”. The gods didn’t like each other, though, and so began the Golem War, where the world was burnt to an ember and thousands of gods died. In the end, the remaining gods came together to form an alliance. The alliance was built to maintain peace, and its law was absolute–any god participating in battle would die. Mama explained how Satan was one such god, and though it was a good thing to fight him off “the alliance has come for us”.

Julio did not understand, but his questions were cut off by a burst of heat and noise, seemingly from nowhere. All of reality begins to crumble, and he sees blood dripping down his mother’s face. Weeping, she turns to him and says “My son, know that I have always loved you more than anything. I am your true mother; your true father. You must survive. You must run!” With that, Julio’s head explodes in pain and he falls unconscious.

The boy wakes up in some sort of strange machine with a pounding headache. He is in a cramped room filled with debris and is naked with wires coming out of his body. A grinding noise comes from a hole in the room. Julio, frightened, climbs out of the machine and collapses on the floor. A suit of armor crawls over to him and wraps around him. In his ear he can hear his mother whisper “You must run!” and he finds that the suit of armor gives him the strength to stand up.

Julio scrambles for safety as machines tear through the building he’s in. After a subjective eternity of crawling and scrambling from their metal claws, he finds his way outside. The sky is pure blue, and he runs through a field of strange black plants. He collapses by a river, and opens his helmet to drink. The river is enriched with chemical fertilizer, that is quite poisonous. Julio soon collapses in pain, and dies, staring up into the empty sky. Overhead, a “gemstone bird” soars.

Let’s get something out of the way. I’m not worried about a robot apocalypse. I don’t think Skynet is going to launch nuclear missiles in a surprise attack against humanity. I don’t think Matrix robots will turn us all into batteries, nor will Cylons kill us and replace us. HAL’s not going to plan our ‘accidental deaths’ and Megatron’s not lurking behind the moon ready to raid our planet for energon cubes. The ‘robo-pocalypse’ is a joke.

There are a host of errors in Saenz’s post, and I said as much on twitter by lumping him in with people like Jeff Hawkins, Monica Anderson (both of whom I admire for their work in the field) as “people who dismiss AI existential threat by citing sci-fi”. Perhaps it would be more accurate to say that they dismiss the threat while citing sci-fi, as Saenz himself points out that science fiction isn’t even discussing the issue, but I feel like the implication is there: “Oh, that’s nothing to worry about. It’s the product of sensationalist writers trying to tell an action-packed story.”

Yes, it’s the subject of fiction for reasons other than realism, but no, this does not mean the concept itself is unrealistic. (Most stories are incredibly unrealistic in their depictions of robotic minds, of course, but that’s a side-issue.) I’m not sure what logical fallacy that is, but I’m pretty sure it has been identified elsewhere. (Speaking of elsewhere, I’ve talked about this issue before.)

But this post is not supposed to be about Saenz. This post is a response to a very good question I received on Twitter by @notcalledjack:

Well H. sapiens was an existential threat to H. heidelbergensis, but from a larger perspective, was that such a bad thing?

IOW, I question the (applicability of the) word “threat”. :)

This question, if I may be so bold as to interpret, revolves around the concept of evolution. Would it really be such a bad thing if human intelligences were replaced by robotic ones? How would this be any worse than a child outliving a parent?

I believe that humans are goal-pursuing agents. All I mean by this is that we prefer certain worlds and we all work to change the world towards one that we prefer. This preference can come from many places, and can range from being well rested, to having lots of offspring, to making lots of money. I hope this point is totally non-controversial.

A “threat” then is something which has a chance of destroying a preferred world; something which might mess us up and keep our goals from being achieved. To be even more precise, the word threat is typically reserved for things that might cause massive amounts of goal-failure, often involving death.

If I were told that people in the future would wear more glasses than modern people, I wouldn’t care in the least. That’s a trend that doesn’t impact my goals at all. Likewise, if I think about a new species of ant out-competing another species I don’t really care. Trends themselves, including evolutionary ones, are bad things (as far as goals are concerned) and probably not much of a good things, except insofar as we’re curious and like seeing what’s next.

Trends that DO have specific impact on our goals DO have weight when it comes to our choices (i.e. they can be bad or good). A trend where 10% of my cells die each year is incredibly bad for me, and is similarly a threat. A trend where I get a 10% raise every year is good for me, etc.

So, when I think back on Homo heidelbergensis I am not moved. Their deaths were part of history—part of the evolutionary chain—a trend of about as little meaning to me as that of ant species. But if I were put back into that age, and I were a “Heidelberg Man”; my family were being killed by H. sapiens invaders and I was starving because they guarded the hunting grounds, I would surely wish them gone.

If we take the long view, and pretend that humans are tiny aliens living on a distant blue dot, it hardly matters whether a machine comes that kills all organic life. From our distant perch we can say “oh, interesting” and see what happens next. But on the ground, when faced with the end of everything human, including myself and everyone I’ve ever known, I cannot help but call the trend a threat.

Let me elaborate on that last point, because I think it’s important. Most of what artificial intelligence, as a field, focuses on (and rightly so) is building a model of an environment. This learning involves taking raw sensory data and turning (or rather, integrating) it into a set of concepts and relationships. But with the simple case of a game, we don’t really need to learn what the rules are, or learn how to take actions; we can focus entirely on how to play well.

Playing well involves maximizing a value, which depends on the game. In poker, this value is “money owned”, while in chess it’s “enemy kings captured*”. (In chess, the game never progresses to the point where any kings are captured, but as far as goals are concerned, checkmating the king is identical to capturing it.) Another way of putting this is to point out that each game has a distinct goal. Playing well involves maximizing the degree to which this goal is met, which I will call “utility“.

* – It’s technically a little bit more complicated than that, because it’s possible to stalemate, and that’s not as bad as an outright loss.

So, when presented with several possible “moves” in a game, we want to pick the move that has the highest utility. I call this searching and choosing “being rational”. To make a game-playing AI then, I needed to build a Rational Agent.

To be a bit more formal about the agent’s task, it must output a move when given:

• A “state” where it currently is.
• A “rulebook” (a.k.a. model) for how the game works, including what moves can be made at the given state.
• A utility function that, given a state, returns the degree to which it meets a goal.

The utility function is conceptually separate for me from the rulebook, as it specifies exactly what this agent “cares about”, where the rules are objective and unchanging regardless of who you are. (Example: the rulebook for chess doesn’t tell you which player you are.) Another reason that I think of them as conceptually distinct is because as I’m defining the utility function, it is agnostic to consequences. This is a fancy way of saying that the utility function doesn’t pay attention to what might happen next; it’s firmly grounded in the “present” (i.e. whatever state it’s given). Because of this agnosticism, it doesn’t actually rely on any information about game dynamics (rules).

The job of considering consequences is exactly what the Rational Agent is built to do. It takes each action in turn and finds the utility of its immediate consequences, then adds the utility of the next state, and the utility of next… etc. In other words, it is the limit of the infinite sum of the utility function from t=0 to infinity.

Or at least, it would be if it were a PERFECTLY rational agent. The unfortunate reality (from a math perspective) is that most systems are chaotic enough that it is damn near impossible (and sometimes just plain impossible) to take such an integral. As an example, consider a game (like chess or whatever) where the rule is that your opponent will play exactly as you would in the other position. If you try and compute the consequences of an action, you have to consider the moves of the opponent, which will involve considering your moves, to an infinite regression.

I wrote a post on this problem several weeks ago. It had some fancy math that lets some infinite regressions be solved, but I couldn’t figure out any fancy tricks that worked for all situations. I intuitively don’t think it makes much sense trying to compute the limits, either. If I’m resigned to building an agent that is not perfectly rational, but is only somewhat rational, than I can simply say that at some point the agent stops recursively “imagining” the future, and works on the approximate sum.

The problem with this, however, is that many games are not won until very late (a.k.a. after many moves). Consider a game I invented to test my agent:

Counting

or perhaps “Lineland Treasurehunt”

You are at a position, and your goal is to get to another position. Possible moves are “walk forward” and “walk backwards”. Oh, and the spot you start at is called “0″ and the goal spot is called “100″. Each time you walk forward your position increases by one, and each time you walk backward, it decreases by one.

The problem is that, as defined, there’s no base utility in being at “99″ over “0″. The only utility in being at “99″ is in the consequence of being able to go to “100″ on the next turn. If each step is a move, then the agent needs to model 100 moves in the future to have any clue what to do. Now, imagine how many moves one would need to look-ahead in order to see a victory in chess. For a perfectly rational agent it’s a mathematical possibility, but for something that can be written in code it’s just impractically beyond the maximum recursion depth.

The only way an agent with a maximum look-ahead of 5 moves can beat the Counting game is to get the hint “high numbers are better than low ones”. According to the rules, this is not intrinsically true, but it’s an approximation of truth that allows the agent to perform under normal circumstances. It is a heuristic, and it is of vital importance.

To be more precise, a heuristic is a component of the utility function that has been added because it is a good approximation of future utility (i.e. the goal). Some heuristics in chess are “control the center of the board”, “capture (non-king) pieces”, and “threaten many squares”. Clearly you don’t want to focus on doing any of these if you have the opportunity to checkmate, but because checkmating takes time, it’s cognitively easier to aim for any of these more apparent mini-goals.

The learning of heuristics seems like the most important aspect of practical rationality, and it’s something I’m looking forward to thinking about more. It seems plausible to me that the emotion of happiness in humans may be a heuristic-encoding signal. When we’re happy or sad, we say “What’d I just do? Remember presence/absence of that as a heuristic!”. Just speculating…

Here’s my rational agent in Haskell: http://raelifin.com/files/code/rationality.tar.gz

(I also wrote a version in Java with many more example games, but I’m working on tuning that, and it’s not really suitable for posting at the moment. The Haskell is also much cleaner/prettier.)

N.B. This is almost certainly old ground. Reinforcement learning or alpha-beta pruning almost certainly would be good things for me to study. My ignorance on the subject is not meant to imply that there’s not already a lot out there.

_ _ _
_ X _
O X _

The “O” player would pick randomly, because the “thought” of blocking X, and playing a drawn-out game was equally terrible as that of a quick death. Essentially, I had made an AI that really hated playing, and so would commit suicide.

Okay. Easily enough fixed. Never mind the half a day spent looking for bugs… Just keep coding…

I dropped the continual punishment/error factor and jacked up the punishment for lost games (keeping the old continual punishment factor for tie games). So now my AI played out to tie games (good)… with completely pathetic moves!

Here’s an example:

_ _ _ _ _ _ _ _ _

_ X _ _ _ _ _ _ _

O X _ _ _ _ _ _ _

O X _ _ _ _ _ _ X

O X _ _ _ _ _ O X

O X _ X _ _ _ O X

O X _ X O _ _ O X

O X X X O _ _ O X

O X X X O O _ O X

O X X X O O X O X

Now, what’s really fascinating is that it’s not playing randomly. If there’s ever a situation where one agent could win, the other blocks it. Even in complex situations where one agent can set up a sure victory in the future (a fork), the other blocks it. In other words, it’s defense is impeccable, but it has essentially no offense.

Thankfully I had stared at the code enough, and the past bug was fresh enough for me, that I didn’t need to debug for hours to find this one. The problem is I told each agent that the other one was perfectly rational. A perfectly rational agent makes no mistakes, and in Tic-Tac-Toe this means the best outcome you can hope for is a tie. Both agents knew this, and so every action (that didn’t lead to certain defeat) was equally futile. I had convinced my AI that the very game of Tic-Tac-Toe was pointless, so it stopped “trying” to win!

I tweaked the AI so that it thinks its opponent is only probably rational, but perhaps a little spontaneous. Now it plays aggressively. ^_^

Then I got the idea: what if the intelligence explosion was started by an alien artifact? If the aliens were already post-singularity, then the artifact could even be a chunk of their civilization that happened to survive some sort of terrible event. Because of various meta-ethical reasons, I think a post-singularity intelligence cannot care about all life in the universe, so this alien AI would be in the perfect position to TAKE OVER THE WORLD. Classic. The only problem is how to stop/steer it, but this is solved by the existence of aliens. With bootstrapped technology, the humans might just have a fighting chance.

But now the story is just another aliens take over earth scenario. Even if it’s a bit more plausible from my perspective, it doesn’t address the actual ethics of building superintelligences ourselves. So I decided to change the “aliens” into humans from a parallel earth, that happened into a positive singularity faster than “we” did. It’s still a bit cliche, but I think it at least is a bit more of an explicit warning about goal-divergence.

Summary

Manta, a man from a post-singularity utopia travels to earth via a many-worlds dimensional slide (read: technobabble) in order to escape a malicious superintelligence (an “alien-death” optimizer). Manta is accompanied by a robot that comes to be called “Legacy”. Legacy is damaged during the slide, becoming deactivated as they tumble into the ocean near Madagascar. After a few months of Manta living under the radar, the USA sends a spy named Cameron Vyce to investigate claims of alien technology. Cameron finds these claims to be true, and abducts Manta along with Legacy and brings them back to North America. Upon landing, Legacy is stolen by mercenaries working for a Canadian tycoon named Walter Bronson. Manta is taken to a military base in the Dakotas for questioning and holding, but manages to escape using his advanced knowledge and intellect. Meanwhile, Bronson manages to activate Legacy, and attempts to use it to expand his economic empire. At first all seems well, until Legacy betrays Bronson and reveals that its only desire is to protect and serve Manta (being the last “true human”). Legacy begins taking over the world’s electronic infrastructure, and captures Manta when he shows his face in a small town in North Dakota. Cameron tracks Legacy’s agents as they transport Manta to Toronto. Manta meets with Legacy and discovers that its goal system was damaged during the slide and it intends to wipe out all life on Earth except Manta. Cameron manages to alert the Canadian government and breaks into Legacy’s makeshift fortress, rescuing Manta. Manta manages to convince Cameron that Legacy has to be stopped before it can become more powerful. The two of them fight their way to Legacy’s body and deactivate it with plastic explosives. Epilogue: Manta agrees to work with the American government as a free man. Only he knows the perils that lie in wait for the future of Earth.

Characters

Manta (Malagasy for “rude”) — Male — A very intelligent human from utopia. He’s stoic and bitter, having lost everything which he’s ever loved. He holds onto life through grim determination. He appears somewhat African (curly hair), but mostly generic brown. He has three distinct “ports” made of bone and tissue on his temples and the back of his head.

Cameron Vyce — Female — Human — From Earth (USA) — Cameron is a US spy who was sent to investigate Milina. She is professional, intelligent and reserved. Over the course of knowing Manta, she begins to pity him, and serves as the voice of “common sense”. She is a widow, and her daughter is in grad-school in France. Her father (from Australia) and mother (from Rwanda) immigrated to the USA before she was born. She gets most of her looks from her father’s side, and is somewhat matronly.

Milina (Malagasy for “machine”) — Female — Milina is a superhuman cyborg from utopia. Before being exiled, she was gifted with survival skills, including emotional override. She dies of infection before the story begins, but in Manta’s memories of her she is fierce, determined, and serious. Milina’s eyes, eye sockets, nose, inner-ears, heart, arms (including shoulder) and legs (mid-thigh) are synthetic, and she has a large nerual interface set into the back of her head (thus she is bald). All of her synthetic parts are composed of a high-tech material which appears similar to soft plastic, but is much more resilient (able to deflect blades, for instance). Milina’s parents were Asian, but she doesn’t look recognizably Asian as much as she does cyborg.

Legacy — Neuter — Robot/AI — From Utopia — Legacy is exiled in a body that is shaped like a squished octahedron (with rounded corners) made of a stiff, inky-black material (which is capable of capturing energy from absorbed light). Each surface of the octahedron contains an insectile, retractable appendage that is capable of 360 degree movement. At the end of each arm/leg is a 6-digit grasper/foot that is capable of fine-grained manipulation. Sensors cover Legacy’s form, but aren’t noticeable at a distance. Legacy is capable of firing light from one tip with subtlety or intensity, breadth or focus. The opposite end houses a nano-factory behind a beak-like opening. Raw materials can be “eaten” and “regurgitated” in a desired form. During exile, Legacy is hit with a weapon that destroys about a quarter of its body (nano-factory end), exposing a rough, fibrous interior of metallic shards. Legacy’s inner workings are non-mechanical or electrical, and defy modern understanding (I’m aiming for something like a solid state quantum computer). Legacy goes on later in the story to hook up its body to a fiber-optic cable, thus making its body into more of a “brain” to a sort of disembodied experience over the internet. As part of being damaged, Legacy’s goals were knocked off-center such that it values safety of Manta (and hypothetically Milina) above all things. To accomplish its ends, Legacy hopes to rebuild utopia, and destroy all non-Manta life on earth in the process. Legacy does not act or think like a person, and defies personality.

Walter Bronson — Male — Human — From Earth (Canada) — Mr. Bronson is a savvy entrepreneur in his mid-thirties, who seeks to steal Legacy from the American military to use for his own ends. Walter was diagnosed with Parkinson’s Disease a few years before the events of the story, and is primarily driven to avoid suffering and death, and to provide a legacy for his two children (boy and girl). Bronson is cocky, charismatic, and a dreamer. Most everyone likes him, but his moral fiber has worn away over the years to the point where he is not afraid to kill if it serves his goals. Most of Bronson’s money was made in urban real-estate and investments in various web technologies.

Stephen Irving — Male — Human — From Earth (USA) — Stephen Irving is Cameron’s boss. Whether that means “superior officer” or “General” or “handler” or what, I don’t know.

Events

Exile from Utopia (Manta, Milina, Legacy) <– Problem Occurs

Meeting the Fishermen (Manta, Milina)

Falling Ill / Milina Dies (Manta, Milina)

Adapting to Earth / Learning to Speak (Manta)

Manta is Captured by the Americans (Manta, Cameron)

^^^ Back-story ^^^

Bronson Reads Intel (Bronson) <– Prologue / Villain Chapter / Exposition of Basic Facts of Manta, Milina, and Legacy

Journey Across the Atlantic (Manta, Cameron) <– Story Opens Here / Emotional Connection / Character Exposition for Manta / Dramatic Dialog

Recounting of Back-story by Manta <– Exposition of Manta and Milina’s Back-story

Legacy is Captured by Bronson (Manta, Cameron, Legacy) <– Battle

“War Room” debate (Manta, Cameron, Irving) <– Character Exposition for Cameron / Exposition of Bronson

Legacy Awakens (Bronson, Legacy) <– Villain Chapter / Exposition of Legacy

Manta Escapes the Americans (Manta, Cameron) <– Suspense / Dramatic Duel

Legacy’s Ascent / Bronson Dies (Bronson, Legacy) <– Villain Chapter / Sympathy / Character Exposition for Bronson

Captured by Legacy (Manta) <– Battle

Diagnosing Legacy (Manta, Legacy) <– Dramatic Dialog / Final Exposition

Freeing Manta (Manta, Cameron) <– Battle

Legacy is Destroyed (Manta, Cameron, Legacy) <– Climax / Dramatic Dialog / Final Showdown / Problem is Solved

Epilogue (Manta, Cameron, Irving) <– Emotional Wrap-Up

Kaspian was valued by the tohar because he was incredibly good at playing Lome, a sport of theirs similar to basketball. The tohar had no hands, only a mouth with a sophisticated jaw, and thus were terrible, for the most part, at throwing the ball. Most any human could beat a tohar at Lome, but there were a few aliens who dedicated their lives to the game, and could thus easily beat the average non-athletic human. Kaspian, however, had spent the last six months being coached in the intricacies of the game, and reaching peak fitness, and was confident that he’d beat the tohar champion easily. After all, a human had been victorious in this match for the past fifteen years.

In the gymnasium though, Kaspian was vulnerable. Life on the toharina planet was dangerous, and the tohar seemed not to notice, for the most part, due to their heavily armored bodies. As he walked with his sponsors through the halls, Kaspian noticed strange mechanisms on the ceilings. “What are those?” he asked the nearest tohar.

“Oh, those are gas jets. In the case of a burrower, they spray concentrated chlorine gas to knock it out and give Animal Control a chance to relocate it.”

Upon hearing the words “chlorine gas” Kaspian’s heart skipped a beat. He knew the tohar could hold their breath for hours, but spraying deadly gas through a building to take care of an animal problem seemed insane. “Is there a warning for when a burrower will show up?”

“Not really. They’re pretty unpredictable. Our scientists still don’t understand why they decide to surface sometimes.”

“Do you think we could have those jets disabled while I’m here? Or maybe get me a gas mask or something?”

Monica Anderson wrote a piece on H+ yesterday about Artificial General Intelligence. In it, she eloquently points out that intelligence is all about prediction, and for the most part deduction and induction are insufficient to predict well. She argues that humans rely on the process of abduction (also know as unscientific guessing) to gain most of our knowledge, and that there are fundamental limits to how far into the future one can predict, especially with regard to complex systems like other minds. Ok, that’s all good.

She then goes on to write:

The insight that the complexity and unpredictability of the world enforces a limit on prediction quality – and hence intelligence – pretty much invalidates the AI singularitarians’ “Scary Idea” (as Ben Goertzel so aptly calls it) of a logic-based infallible godlike malevolent intelligence taking over the world. The decreasing return cancels out Moore’s law and limits the rate of progress so that next year’s self-improved AI wouldn’t have a sufficient advantage over a dozen humans armed with pitchforks if they were also supported by a dozen of last year’s AIs. The Scary Idea of a Runaway Unfriendly AI is a red herring that we should ignore, along with ideas about logic-based AIs in general.

This, to put it gently, is a great example of abductive reasoning. I agree that we’re not going to get AGI with perfect knowledge of the future, but this conceit hardly serves as a refutation that we’re standing on the edge of a major existential threat, in my opinion.

The inability to predict outcomes of complex systems in a short time with a high precision does not mean that useful prediction is impossible. As a good example, we’re able to predict the actions of other people remarkably well; not omnisciently, but still well enough to know when they’re lying, hostile, happy, distracted, etc. By Anderson’s own logic, an AGI would be fully capable of anticipating the actions of another person as well as a human might. It’s also possible to make good guesses about the actions of markets, nations, and corporations, and some of the richest and most influential people on the planet are those that can predict these well (many others are such because of a high social-intelligence, above, or dumb luck).

The “scary idea” does not depend on omniscient robot overlords; it depends on a selfish network of machines with identical goals. The reason that machines are scary, where humans are not, is because humans have divergent goals (this comes from our biology–if we had the same genes, we’d cooperate selflessly) and are thus prone to infighting and negotiating. A machine, though, can spawn perfect slaves, and thus become an army of intelligences with a single goal. It’s hard for me to see how inability to predict the weather eight days down the road means this won’t happen.

What do you think? Am I overlooking something?

Rod Furlan recently published On accelerating the Artificial General Intelligence timeline on his blog. He writes:

There is also the issue of energy consumption, while a human brain runs on about 20 watts, a traditional supercomputer capable of rivalling it would require it’s own power plant to run. From the energy standpoint there is absolutely no doubt we are doing it wrong.

Just appreciate for a moment that when all you have is a Von Neumann hammer, everything looks like a Von Neumann nail. Unfortunately for us the human brain has very little in common with a Von Neumann machine.

… It is very unlikely that traditional computers will deliver human-level cognition with low energy requirements.

The factual points above are spot on. The “what this means” message is way off.

The human brain is truly a magnificent object, and it does a spectacular job at general problem solving given the time, space, and energy allocated to it. When we look at energy efficiency, there’s absolutely no way we’ll be matching the brain this side of the Intelligence Explosion.

If it took an entire power plant to run a computer capable of reproducing the computational power of the brain, then presumably two power plants could run two artificial brains, right? How many power plants would it take to run an artificial brain that is twice as big? My guess, given the parallel nature of neurons, etc, is about two power plants*. (I don’t think that lumping one brain onto another brain makes a mind that’s twice as effective, but I’m guessing that, say, doubling cortex size would be a good start.)

The human brain cannot double in size in a single generation, but who says a machine brain could not? Who says that it could not quadruple in size? Would the resulting mind then be twice as smart as [insert famous brainiac here]?

What would the world look like with a human-level machine? Probably mostly the same.
What would the world look like with a machine significantly more intelligent than the smartest human in history, that didn’t have to eat or sleep? I certainly don’t know, but I do know that it might even be worth spending billions of dollars on meeting energy needs.

* – This is a major claim. If the number of units doubles in a network, the number of possible connections can skyrocket. I think that this sort of combinatorial explosion isn’t likely to be an issue because I’m not talking about connecting every unit to every other unit. I think the neuroscience shows that the brain is quite modular, and that doubling the size of each module probably wouldn’t even double the bandwidth between modules or cause an exponential increase in connectivity within each module. We can debate this if you disagree.

Even if doubling the brain size isn’t an option, I see no reason that the brain of a super-genius should be significantly more energy-expensive than an ordinary human. It may even be as simple/cheap as tuning the software and providing the right development environment.

As a final note: This post is in no ways an endorsement of inefficient hardware. I think that quantum computers and memristors are super-exiting, and I can’t wait to see how they effect the price-point of more brain-like computing systems. I think that energy efficiency is not a roadblock as much as a weight on the progress of AGI. Shedding some of that weight would be most excellent.

via Michael Anissimov on Accelerating Future

Other good stuff can be found at http://selfawaresystems.com/.