With regard to AI, it’s easier for me to answer direct questions rather than general interests, but I’ll do my best to summarize some of the ideas which I find most interesting, and the dilemmas I’m working on at present.

You may have noticed that I use the term Artificial General Intelligence, rather than simple AI. The “General” is used to distinguish from what is called “Narrow” AI. A good example of narrow AI is the spam filter you probably have in your email client. The spam filter has knowledge, the capacity to learn, and ability to act autonomously, but the domain to which the filter is applied is highly restricted. My interest pertains to AI that is not specific to any domain, and can solve problems in a multitude of environments. For instance, a truly general agent could drive a car, translate languages, direct air-traffic, or diagnose diseases without any special work on the behalf of a programmer.

Naturally, such a system would be equally capable of learning computer science, psychology, electrical engineering, etc. and be capable of applying them towards various goals, presumably including the design of AGI. Because of efficiencies in replication, communication, and not having to sleepication, I can imagine that it wouldn’t be long before our creations were better at creating than us. The theorized result of this would be an Intelligence Explosion, wherein ever-smarter agents would be able to build ever-smarter agents… until the bounds of their intelligence hit some sort of ceiling. I, and many others, believe this ceiling to be far beyond the intelligence of even the smartest humans, and would allow the artifical superorganism to achieve its goals to an unsurpassed degree. (Note: this doesn’t have a timeframe attached to it. It may happen in the next few decades, or the next few centuries. I don’t know.)

An emergent goal of all rational beings is self-preservation, so unless specifically counteracted in the initial design, a superintelligence would naturally be defensive at best and hostile at worst. Any possible threat, or impediment to its goals would naturally be removed. It is thus paramount that any AGI design be given the exact same goal set as the most ethical person imaginable, as anything else would likely result in vastly terrible things occurring, notably including the destruction of all evolved life.

There are some good reasons why I think that can be avoided, however, and if said superintelligence was benevolent, we would be gifted with the solution to nearly any problem we can think of today (possible exceptions include: entropy).

Naturally, I think this is a big fucking deal.

I also recognize that it makes me sound a bit crazy. I am reminded, however, of something Arthur Clarke once said:

“Trying to predict the future is a discouraging and hazardous occupation, because the prophet invariably falls between two stools(?). If his predictions sound at all reasonable, you can be quite sure that in 20 or at most 50 years, the progress of science and technology has made him seem ridiculously conservative. On the other hand, if by some miracle, a prophet could describe the future exactly as it was going take place, his predictions would sound so absurd–so far fetched–that everybody would laugh him to scorn. … So, if what I say now seems to you to be very reasonable than I will have failed completely. Only if what tell you appears absolutely unbelievable do we have any chance of visualizing the future as it really will happen.”

Because I am curious by nature, and not content to watch the future simply unfold, I find myself drawn to making it happen, or at least exploring the field with great depth. To this end, I must fully understand what it is that allows a person to solve any sort of problem they are given, and distill this quality until it can be given to a machine.

Many researchers in the field have approached this problem by trying to take the pinnacle of human thought, which in their eyes is typically logic, and reducing it to a procedure, primarily through introspection. Most early attempts at AGI followed this introspective, reductionist approach… and failed horribly. Despite pouring in buckets of “facts”, their logical machines were unable to solve anything more than toy problems.

This failure caused a collapse in the AGI field, as institutions like DARPA withdrew funding, thinking that General Intelligence was a myth. Good work continued to happen in AI, but it was applied to fields that were tractable, and would offer results. The defeat of Garry Kasparov, the chess-master, in 1997 to IBM’s Deep Blue computer is a good example of a success in narrow AI that has little to no relevance to the field of AGI.

During the AGI winter, neural nets became a hot topic, and were developed to levels of modest success (such as filtering spam). Inspired by the behavior of neurons (though not actually realistically replicating their function), neural nets provided an alternative to the reductionist logic of earlier work. By linking together simple pattern-matchers into vast webs, and letting them automatically adjust to match provided data, neural nets are capable of learning to recognize commonalities within very complex environments; such as discerning between objects when given a host of visual data.

In parallel to all this work on AI came some interesting work on psychology. The theory of Behaviorism, once championed, fell from popularity in the late 20th century, and in its place came a not very well known (even today) theory that was inspired by electrical engineering of all things, and served as a better model for behavior than anything before it (from my perspective). This theory, stated that behavior was not actually a fixed response to a given environment, but instead an attempt to change what the organism perceived. As an example, a mouse that is under a sun lamp will crawl under a rock. Earlier psychologists would claim that the sun lamp causes the mouse to move, but under the perceptual-control theory, the mouse moves so that its perception matches innate goals, namely coolness. Hence, if the mouse is removed from the perception of heat, by making the air colder for instance, it will not seek shade.

Though perceptual-control theory may seem basic at first glance, it is very powerful when closed-loop control systems are linked together to simultaneously control for complex outputs to perform one smooth action.

Words really don’t do this justice, because both neural nets and perceptual control systems are highly emergent systems; those that are simple at the unit basis, but unfold in complexity and power as these units interact and connect. To really grasp them, one must view the power of the whole. Emergent systems seem a bit like magic to me. Go examine the Mandelbrot set for a classic example.

Two major pieces of the puzzle are yet missing, however. We have perceptions and actions (and logic, but that’s not really important in the scheme of things), but we also need goals and thoughts.

Goals are what I’m working on right now, and are the last “animalistic” portion of mind that is truly necessary. I’ll talk about my work with them in a moment.

Thoughts will ultimately serve as the powerhouse for the mind, by allowing for future prediction, past inspection, and other imaginings of non-present perceptions. Imagination serves, as far as I can see, as the foundation to complex language, but I haven’t studied it enough. It does seem clear to me, however, that one cannot jump to thoughts without first having a solid “animal mind” to support them.

(I think many animals have thoughts, such as dolphins and apes (and humans), but far more have the other, more basic, systems.)

Backing up to goals, I can address some of the problems I’m working on at the moment. Specifically, while we seem to have a goal to remove certain bodily sensations (pain) and gain others (pleasure), we also notice very common abstract social goals, such as fulfillment, companionship, love, and safety of offspring. These concepts are abstract things, which must be learned through exploration, just like other abstract things like math or politics. But if they’re learned, how can they become goals? Is there some sort of homunculus that watches for the emergence of said ideas, and then hooks them up to the control systems that will implement them? Are these goals not actually innate, but instead the product of social pressures? If so, how is social pressure able to bind them? It’s all very confusing, and if you have any ideas, I’d love to hear them.

I have a problem with science fiction: it’s too pessimistic. I don’t mean this in a “dystopian” sense, either. I mean that visions of the future are haphazard in what technologies exist, and what people will need to do to survive.

In the future, treadmills will suck.

Take James Cameron’s Avatar, for instance. In Avatar there are many vehicles that transport humans around on the alien world of Pandora. These vehicles are all piloted by people. The natives of the planet are actually able to kill off aircraft pilots by throwing spears through the windshield. Now, imagine if these aircraft were piloted by computer, as many military planes are today. No pilot would mean the front of the vehicle could be well-armored. No pilot would also mean that such aircraft could be mass-produced on-site, without needing to ferry pilots in from Earth. There is a scene where the humans attempt to destroy a specific forest by flying a bomb there by hand. Why isn’t the bomb just dropped from orbit? Don’t these people have spacecraft? The worst plot-hole, however, is the justification for fighting the natives of the planet. Apparently there is a rare substance of astronomical value that’s located just below the native’s sacred land. Yes, it works for symbolism, but the substance is needed for, get this, an energy source. So apparently, in this advanced future where they can afford to shuttle people around at 70% the speed of light, humanity has not yet figured out how to get decent energy from nuclear or solar? Pessimistic in the extreme.

“I can go back to playing World of Warcraft now?!”

As another example, let’s consider Pixar’s WALL-E. In the film, humanity abandons Earth for being too toxic, and ends up living for 700 years on a colony ship, where they are catered to by artificial intelligence. These people live each day constantly connected to the computer, and never do so much as get up from their chairs. At one point in the movie, WALL-E (the protagonist robot) bumps into a human, distracting him from his virtual world, the result being that the man is pleasantly surprised. Let me clarify: Pixar is telling us that in the far future, a person who has never known anything but a life of digital entertainment would be pleasantly surprised by being distracted by a dirty little robot. Those video games must SUCK. I understand that the whole thing is commentary on our modern, western lifestyles of excess and entertainment, but that doesn’t change the part of the story where these people are so unsatisfied by futuristic life that they feel a need, at the end of the movie, to return to a life of agriculture. That’s so insanely pessimistic.

The reason for this pessimism, I think, is two-fold:

• people forget that technology actually does improve society (most of the time)
• writers recognize that without problems there is no story

I’ll address these in reverse-order.

Okay, so plots are problems by their very definition. No story ever went “Once upon a time there were some happy people. The end.” We write about problems because, perhaps on a subconscious level, writers are attempting to fix real problems by communicating imagined scenarios (i.e. fiction). The writers at Pixar were hoping to (on some level) make Americans a bit more conscious of their lifestyles in hopes that they’d take better care of the planet and themselves. To do this, the writers tell a story about what would happen if these trends got worse, rather than better. The story of Wal-Mart destroying the world.

My issue is that when you tell a pessimistic story, you confuse the real problem with a hyperbolic one, and run the risk of failing to address the actual obstacles in place. In other words, if your story isn’t realistic then people may throw out the core problem as unrealistic, or they may think that hyperbolic aspects of the plot are actual problems, and focus on the wrong things. The best example I can think of is that of the Terminator, an icon for evil robots everywhere. On one hand, people watch Terminator and consider just how ridiculous it is to have an evil AI (that gets naked when it travels through time) and so they mistakenly throw out the problem as being one of idle fantasy. On the other hand, people get the impression that the mistake was in giving Skynet the keys to the bombs, and so they mistakenly believe that non-military AI serves no threat.

My suggestion: make the problem in a story as close as possible to the actual one, and avoid hyperbole when possible. Millions of people on Earth suffer because of the choices made by affluent people in the West, so tell a story about those people, and how they might come free of their hardship (or not, for a downer-ending).

Now for the second point; people (sometimes) forget that progress is real.

Why are there bartenders in Star Trek, again?

People have incredibly short lives, if you think about it. Most people alive today cannot remember getting their family’s first television, much less the first radio. The scope of many people’s lives are such that things seem only slightly different than they ever were. People forget that infant mortality has dropped by a factor of 10 in the US in just the last 70 years. Not having enough XYZ has been a problem for all of my life, so why wouldn’t it be a problem in a hundred years? Because that’s not how technology works.

Slavery existed for nearly the entire course of human history, and just happened to end at the time when complex machinery began making factories feasible. Coincidence?

Aging. Poverty. Slavery. Energy. Transportation. Education. Communication. Disease. Ignorance. Even work itself. These are all finite problems with points where they will be essentially solved. To think else-wise would be to cover one’s eyes to the lessons of history. Even companionship, empathy, security, intimacy, and fun are goals which I think can be solved with technology. When every story of the future paints a picture that says “nothing substantial changed”, we get the impression that science and technology are good for nothing by making shiny gadgets and robot dogs.

Science Fiction stories are our dreams of the future, and dreams matter. They help pull us past the repercussions of our actions and into a brighter future. I have no idea what’ll happen in the next 50 years, but I can guarantee we won’t be anywhere near where we are today. Civilization really only has two end-points: extinction and utopia. Let’s pay better attention to where we’re going.

Perhaps I should say “Can you tell how we made this image?”, because there is more going on here than just my mind and hand. Sure, as far back as a decade ago it might’ve been possible for someone to make the same edits, but that person would need to be far more experienced than I.

To view the original, click below and hover over the image above.
Turn hover mode on.

When comparing the images side-by-side, it’s easy to tell that the part of the picture that was filled in is lower detail and has a few errors, but unless you’ve worked with a lot of images in the past I’m guessing that it wasn’t apparent before seeing the original. My partner in crime here was a tool called Resynthesizer for the GIMP. This is the open-source version of the “content aware fill” that will be included in the next version of Photoshop.

The reason that I’m able to perform such a complex fill on an image is not because I am experienced. It is because the tools have knowledge and skill embedded into them by their creators. As AI becomes more successful, our tools will become more competent at doing complex things with little supervision. This will put people out of work. This is the future. We don’t need or want more capable people, we want more capable autonomous systems; systems that can mow lawns, shovel snow, cook, clean, write, watch, build, experiment, and teach.

Vanguard has some of my favorite reporting on global events and trends. This week they investigate robotic warriors, a subject which I have written about in the past. The only thing which was really news to me was the fully-automatic shotgun (3:20). Scary stuff.

Over the past few months I’ve been seeing a few mentions of how universities might become obsolete recently. This is ridiculous, of course; it has been possible to learn this sort of material on your own for free since the public library was invented. What is clear, though, is that education is experiencing the first waves of disruption from the internet. After all, it’s one thing to read textbooks before bed, but nowadays I could get the books instantly, watch the lectures, and discuss the problems, even if I lived in the middle of nowhere and had to use solar panels and satellites to stay powered and wired. Hell, I could probably even do it if I didn’t speak English.

It reminds me of how the internet is disrupting books. Books aren’t going away; quite the opposite, in fact. But books can’t stay the same in the years to come. They translate too easily into bits, and the internet applies a constant pressure to serve up information as easily and cheaply as possible.

This power of the internet to push towards connection, collaboration and knowledge is probably my favorite trend right now. It’s something I can watch disrupt television, healthcare, politics, publishing, and yes, education, on a timeframe short enough for me to watch.

I’m sure Clay Shirky would agree.

UPDATE: Study Finds That Online Education Beats the Classroom

I didn’t mention longevity research because I have mixed feeling about it. I support it heavily and will likely donate money to SENS in the future, but donation is not investment, and personal conviction is different from mass action.

The crux of my conflict is whether there is greater utility in putting money toward longevity or strong AI. I’ll assume, to start, that any strong AI created will be friendly. It seems reasonable to me that if we were to make an intelligence that is roughly as capable as a human, that we could very easily make an intelligence that is at least slightly better than a human. The conversion of intelligence into an information technology would, I predict, lead to a super-intelligence that is vastly better at solving problems than even all of humanity combined. If this is true (and such an AI is friendly), it seems to me like indefinite lifespan would be solved without any additional work by humanity.

So the questions then remaining are: Is the construction of a human-level intelligence possible? Will it happen soon? And is it worth fixing aging before hand, so that people don’t die/suffer in the meantime?

I’m going to, for simplicity, say that an “explosion of intelligence” like I described earlier will happen in 50 years if I don’t “invest” in it. I’ll talk about timeframes later. And for now, let’s pretend that longevity escape velocity (LEV) occurs 25 years from now without any of my investment.

Now lets say I buy 5 years of “progress” in longevity, so it happens only 20 years from now. I think it’s likely that within the first few decades of LEV, most of the change in life expectancy will come from wealthy old people in prosperous nations. I just saved and improved many lives, but not a whole bunch… probably under half a billion. Infectious and lifestyle diseases, accidents and wars still exist, and those who can’t afford treatment still die when they get old, even when AI shows up.

Now lets compare this with a 5 year acceleration of strong AI. Not only might superintelligence solve longevity, it might do it so well that it’s cheap and can be given to even the poorest. In addition, it’d (if friendly) probably negotiate peace treaties (perhaps by solving scarcity), cure other diseases and if Aubrey de Grey is right, prevent accidents.

This means that if the rate of acceleration per unit of investment is equal, I’d be weighing improvements to the lives (and avoidance of death) of the half-billion elite, versus the curing and protecting of the entire planet for an additional 5 years. I’ll pay for AI, thank you.

Right here is where I’d be balking, if I were reading this. I’ve made a number of assumptions, so I’ll try to go back over each one:

Who says the same amount of money buys the same amount of progress for both technologies!?

If AI is “cheaper,” my argument is even more sound, so I’ll address if the same investment unit buys AI progress only a fraction of the same acceleration as longevity. At some point it just makes sense to put money toward longevity and not AI. If, for example, strong AI can only be made by an uneducated kid in 2063, then no amount of research will speed that up. I have no reason to think this though. It seems to me that more money in a field typically means more people thinking about it, including students and amateurs. Both fields seem like they could benefit from more mainstream funding, and even at a 5-to-1 ratio, I think I’d prefer AI. (That is, I’d still invest in AI if it meant 1 year of acceleration, rather than 5 years of acceleration for longevity)

You can’t pay money to make technologies show up sooner!

You can’t?

Kurzweil has no game plan, while SENS is already in action!

Kurzweil is not the king of intelligent machines (even if he is the celebrity), so a better example might be “<Current leader in general AI> has no game plan”, which I cannot say. Remember that my example works even if strong AI happens several decades after longevity breakthroughs occur.

Who says strong AI is even possible? Or that it’d be any smarter than a human?

Nobody can be sure that either are possible until they happen. I happen to believe that human intelligence comes from nothing more than a system of organic machinery, so in theory that behavior would be able to be simulated. It seems to me that if we can simulate a human level intelligence, it’d be fairly simple to tweak the machinery (so to speak), and improve it, even if only a little at a time. I tend to think that you wouldn’t need to replicate the human nervous system verbatim as much as the general patterns of learning, but either way works.

Why would a super-intelligence be “friendly”? Why not ambivalent or aggressive?

Because we’re the ones making it. Making something that powerful is inherently dangerous, but that doesn’t mean we shouldn’t invest in it. If anything, it means we should invest in the research, so that it’s done right, and not sloppily. I don’t see anything in this argument that makes researching AI any less attractive.

The timeframes are wrong because of X, Y, Z.

If super AI comes before longevity breakthroughs than that’s even worse for those who invested in longevity. If the time between the technologies is smaller than it seems like longevity gets less important as it becomes increasingly available to only the rich, etc. So what if AI is possible, but it shows up in a couple hundred years, while anti-aging happens soon? It’d probably be best to invest in longevity then, yes.

Conclusion: it seems to me that it’d be better to focus on longevity iff: super-intelligent AI is impossible, is only possible far in the future, or is not easily aided by having money thrown at it.

P.S. I wrote this in a fairly stream-of-consciousness way for a facebook comment. If anyone wants to point out edits that’d make it more akin to a paper, that’d be sweet. I’ll probably rewrite it a year from now when I can see just how stupid I’m being.

Related link: Aubrey de Grey on the Singularity and Longevity Escape Velocity

Web Data:


* Huge number of volunteers for genome database

Telepathy:
* “Force Trainer” toy provides MRI for kids.
* Implant wires motor cortex to computer.

“Green” tech:
* “Mission One” electric motorcycle goes 150mph, with a 150-mile range and an optimum charge time of 2 hours.

I’ve been thinking about the future of currency this evening, and oh! I made a quick little simulation of the Fibonacci sequence the other day.

Robotics confronts me continuously online, to the point where I cease to keep track. In terms of military equipment there is Ember, TROPHY, and the old news predator UAV, which I was unhappy to learn now advertises on Hulu (unable to find link, sorry).

More links abound. Seriously. I could do this all day.

Oh, and speaking of Hulu, there was an interview on the Colbert Report (last section) just the other day with Seth Shostak. At the very end of the interview (roughly 20:45), Dr. Shostak points out that the alien life that we make first contact with is likely to be “beyond-biology.” That is, that they’re likely to be AI.

I applaud the out of the box thinking on this one. Most sci-fi would have us in fist-fights with bipedal cat-people, and I think aliens have gotten a bit of a bad rep for it. (Apologies for picking on sci-fi, though. I do so love it.) But I really think that Shostak is wrong here. The biological systems that have evolved over the last three billion years are incredibly more complex and efficient than anything artificial, and I would even state that it is more efficient than anything we’re likely to develop in the next century.

Just for a second, imagine a nano-machine that communicated with other bots with a swarm network to create macroscopic effects, re-arrange in the swarm or repair itself. Now imagine that this machine not only could self-replicate, but could do so in a sustainable fashion with very little waste and even make small improvements upon itself. I just roughly described a biological cell.

Now, let me be very clear, computers and machinery are VERY useful, and capable of doing many things that organisms cannot. My point is simply that there are many things that both approaches cannot accomplish easily, and that the path of least resistance is in combining artifice with the vast amount of powerful organic machinery all around us. We already do this to an extent, where the newest microprocessors are created by applying the human brain in conjunction with present day computers, and where the newest human brains grow in an environment that is connected by technology. We cure disease and disability so that we can live happier lives, but as we do so we cease to exist as a purely biological species. And in that same way, I cannot believe that the machines of the future could not be at least partially organic.

The following (including music) is entirely the product of computer algorithms; a.k.a. procedurally generated content:

We’re getting closer to good brain-interfaces:

I think that the above video is the equivalent of watching a tech demo of cell phones in the late 1980s. Give it 20 years and it’ll be everywhere, give it 40 years and it’ll be unthinkable to control a computer with something other than the brain.

Other things that have been exiting for me:
* Exoplanets
* Exoskeletons
* Netbooks (more)

Take note: The collision between cellphones and personal computers is happening. A good amount of resistance to adopting netbooks probably comes from PC gaming (it certainly does for me) and lack of good cloud-based applications. These problems are being solved though, and I anticipate a full integration by 2015.

EDIT: Oh! I forgot this one!