Why SingularityNET Might Actually Achieve AGI (And What That Means to You)

https://commons.wikimedia.org/wiki/File:AI_for_GOOD_Global_Summit_(35173300465).jpg

There is an axiom that says that artificial intelligence is necessarily narrow-minded, stipulating that AI good at a given task will be necessarily bad at other tasks. For example, AI that's really good at finding songs you like would be helpless at playing chess, and this despite the fact that the song finding AI is much more advanced than your average chess app.

AI in general is also limited in terms of complex decision making. Like all computers, it's great when it comes to running numbers and probabilities, but it breaks down whenever decisions are based on less certain data. This is why it can easily beat a grand master in chess and do calculus as easily as most people can count to ten, yet it can hardly hold a basic conversation.

Despite our shortcomings in terms of large numbers and calculations, the human brain is still by far the most powerful computer ever built. So powerful in fact that we still don't really understand it. We barely understand human cognition, and human consciousness is still a complete mystery. Despite this, though, we've all seen the recent advances in AI, and anyone can easily extrapolate that artificial general intelligence (AGI) is not only inevitable, but probably on the horizon.

This isn't to say that machines are going to be conscious anytime soon. Then again, it's not to say that they won't be. Given that we don't even know what consciousness is, we must accept the possibility that machines could spontaneously become self aware at any point. However, that's not really what we're concerned with right now. All we need is for machines to be able to do anything we can do, at least in terms of mundane tasks, such as translating a book into a foreign language with enough accuracy to make it readable. That's the kind of complex decision making that humans are really good at, that machines currently are not.

Achieving Artificial General Intelligence (AGI)

If we want machines to be good at human things, it only makes sense that we would want to model them after our own brains. And while this may not be practical on a grand scale, perhaps it's only the theory of the human brain that needs to be recreated, as opposed to mirroring every neuron and synapse. And I would argue that that's exactly what SingularityNET is doing!

The main difference between the human brain and AI is that the human brain is highly decentralized and highly interconnected, whereas AI is highly centralized and traditionally not very interconnected at all. In my opinion, that's why humans are so versatile, and so good at complex decision making. Instead of organizing learning towards a single task, information is broken up into its constituents and selectively accessed for a wide variety of tasks, as needed. Thus, something learned for one application can be put to use for a completely different application.

This is because of our ability to form neural connections. We can take things we already know and organize them in a neural matrix that is specific for that task. For example, it is much easier to learn to ride a motorcycle if you already know how to ride a bike. In essence, the program for riding a motorcycle is a slightly modified program for riding the bike. Not only are the basic skill sets already present, but much of the neural matrix is already in place. Simply by modifying a few pathways, we can adapt the bicycle program into the motorcycle program. It's simply a matter of deleting the peddling and adding the shifting action. And if we already know how to drive a manual car, then we can make the process even faster by bringing in that program and modifying it to incorporate motorcycle controls.

To illustrate this, let's say you were looking at a map of someone's brain while they did these tasks. You noted the areas that lit up when they rode a bike and drove a manual car. Then, when they got on the motorcycle, you noticed that both of those areas lit up simultaneously. To you, it looked like they were riding a bike and driving a manual car at the same time, which is probably a pretty good estimation of how the person really felt while learning to ride the motorcycle. In the early stages, they may have even wanted to peddle or reach down for the shifter on their right side, and they most certainly wanted to use the motorcycle's clutch as a brake at some point.

So the real magic of the human brain is its ability to organize and cross reference information, and that's exactly what SingularityNET is aiming to do for AI. Their goal is to have millions of individual AI nodes all communicating with one another, enabling them to form digital neural networks. So don't think of SingularityNET simply as a massive database for AI; think of it as a collection of protocols that enable AI nodes to communicate with each other.

Right now, AI is like a brain without synapses. The raw disk space and all the skill sets are there, but they currently have no way to work with each other, meaning each node is narrowly focused and redundant, making them heavy and clumsy. For example, an AI node that specializes in finding songs you like might be extremely similar to one that specializes in finding movies for you, as well as one that finds products you might like. Instead of having three separate nodes for these tasks, it makes more sense to combine them into a collection of nodes having to do with skill sets related to finding similar things, then connecting them as needed to form specific protocols for each task.

Not only does this eliminate redundancy, but it also means that the protocol can be quickly and continuously updated. For example, let's say you have a node specializing in analyzing music, allowing it to be organized by type. If another node comes along that does the same thing better, the protocol that finds songs you like can simply kick out the old node and replace it with the newer, better one, which might result in more choices, more accurate suggestions, quicker processing time, etc. Likewise, a node specializing in organizing movies might see the new song-organizing node and be able to integrate its improvements.

In other words, SingularityNET will bring neuroplasticity to AI! This is simply a natural consequence of being able to form and modify connections between nodes.

What it Means for Us

So while "strong" (sentient) AI might still be a ways off, AGI (artificial general intelligence) is showing very real promise of coming to fruition, and the implications of that are plenty disruptive in and of themselves. Even AGI that could match the performance of humans at a 70 point IQ in all fields would make most jobs obsolete, some practically overnight, especially now that robotics is making machines capable of doing just about any manual labor that a human can.

Immediately in the crosshairs are jobs like truck driver and airline pilot, as well as customer service rep, store clerk, inventory manager, translator, forklift driver, heavy equipment operator, taxi driver, etc. The only jobs that are safe, for the moment, are ones requiring high degrees of human intelligence, fine motor skills, or artistic expression. Ironically, artificial intelligence may end up making artist a more secure job field than engineer, as difficult as that is to believe!

One limiting factor is that robotics is still very expensive, making it unlikely that it will replace humans anytime soon, except in areas where an extremely high degree of precision is needed. In other words, it's unlikely that anyone will be using robots to dig a ditch in the near future, as that would be akin to using a nuclear clock as an egg timer. It's not to say that the nuclear clock can't perform the task, but obviously the cost benefit analysis doesn't pan out.

The most at risk jobs are ones that can be replaced by a simple computer, with minimal hardware modifications, with driverless vehicles being the prime example. In such applications, the cost benefit analysis is clearly in the realm of massive savings across the board.

Summoning the Daemon

In ancient Greece, people were thought to have personal demons that helped them with everyday affairs, working in the background to improve their lives. If a person was especially accomplished, they were said to have a genius, which in the Greek typology of demonic spirits was considered to be a class of especially gifted demons who put their lofty intellects to use for their human counterparts, garnering great success for them. While AI may not be able to grant your supernatural wishes, it just might come close. Let me explain:

It is my contention that SingularityNET will be ubiquitous for everyone living in the developed world within ten years. In other words, I believe that having your own personal node on the SNET will be as universal as computers and smartphones are right now. It will be the ultimate realization of things like Siri and Alexa, except it will be personal to you and highly adapted to your individual needs. Not to mention it will actually work...

Not only will it work, but it will be your one source for all things. It will be your personal assistant, planning your schedule, making your shopping lists (as well as doing your shopping when possible), setting your alarm clock, making your appointments, etc. It will be your personal trainer, capable of guiding you through everything from simple interval training to exotic forms of yoga. It will be your financial adviser and accountant. It will be your nutritionist. It will manage all of your devices and passwords. It will write custom programs and webpages; you simply tell it what you want, and it will make it happen.

This is all possible because any node on the network can access every other node. As such, it will be your portal into the network. In a way, it will be like a bookmark for all of your favorite nodes, as well as the protocols that organize those nodes in such a way as to accomplish tasks unique to your needs. But you will simply know it by a name, and you will speak to it like you would a person, meaning you don't have to know anything about the network or how it operates.

You will say things like, "Sophie, let's go get a latte," and it will pull your autonomous vehicle around to the front, adjust the climate controls to the weather that morning, locate a coffee shop (but not the one on Fifth Street because it knows you don't like that one), and check weather and traffic so it knows exactly when to order your coffee so as to have it hot and ready for you when you arrive. And as you're walking out the door, it will suggest that you take an umbrella because there's a 90% chance of rain. And if there's an accident on the interstate while en route, causing a traffic jam, it will instruct your autonomous vehicle to take the parkway. And this is all while it's quizzing you on the presentation you have to give at work that afternoon.

Can you even begin to imagine? Practically overnight, we will be released from the minutia of daily life in the modern world. But of course it won't just be people using these nodes for their own personal consumption. Businesses will use them just as effectively to replace human workers, and likely at massive cost savings.

You may even find that your boss has been replaced by an AI one day! Most of the things done by mid level management could easily be done by an AGI, and likely much more efficiently, such as writing schedules, assigning tasks, ordering inventory, training new employees, etc. For example, let's say you're a grocery store clerk. Economics means you haven't been replaced with a robot yet, but there's no reason that your manager can't be replaced by an AGI, as his essential tasks don't require manual labor.

AGI is as Inevitable as Death and Taxes (And About Those Taxes...)

The government will no doubt do everything in its power to tax AI services. Naturally, it will want its "fair" share of the cost savings, which is why it's important now that we all do everything in our power at every level to prevent them from being able to do so. They will also undoubtedly see it as their duty to "protect" us from the disruptive effects that AI will have on our economy and job market, which will likely be the pretext for taxation in the first place.

Don't fall for it, though, as the money will likely be used for everything but helping those out of work. Have faith that the free market will take care of itself. AI and robotics will mean falling prices for essential goods and services, including healthcare. To ensure its success and moderate any disruptive effects will depend on letting it happen organically, without oversight from the creatures in congress, who are about as intellectual as turnips and have shown an unequaled lack of interest for learning anything new.