What happens when you create an open AI marketplace that can learn from itself?
Remember Sophia, the humanoid robot who was granted citizenship by Saudi Arabia? From the general media hubbub to the discussion surrounding the Saudi government’s treatment of women’s rights, the landmark event caused quite a stir across the web.
But behind all that buzz, SingularityNET, whose artificial intelligence technology powers the robot built by Hanson Robotics, has been working to completely revolutionize web as we understand it.
What is SingularityNET?
In some sense, the premise behind SingularityNET is fairly simple.
“SingularityNET is a decentralized open market for AI,” SingularityNET CEO Ben Goertzel said in an interview with the author. “It’s a product aimed at becoming a bridge between AI developers and business users.”
Thinking of it this way helps keep your head somewhat attached to your body when you start trying to imagine a decentralized, interconnected network of AIs that can not only be purchased and shared via blockchain technology—they can actually learn from each other.
AI is an umbrella category that houses four broad types, encapsulating a wide field of study that includes subsets you may have heard more of recently (ie. machine learning, neural networks, natural language processing). AI is already a $233 billion industry, and some estimates predict it will reach $3 trillion by 2025. We encounter AI every day, though much of what emerges from the field isn’t immediately obvious to the user—the point is to seamlessly optimize existing systems. This is also why we’re starting to see a “hockey stick” curve of AI acquisitions over the past few years—congregating these systems in a handful of major corporations, who then resell AI packages.
Even though AI changes our lives by the day, it’s a fragmented field. Many aren’t able to see or learn from the work others are doing, unless they happen to be working in a company like Google, Facebook, or Salesforce (and even then), and there’s no discoverability for AI inventors.
SingularityNET addresses this visibility issue, enabling everyday users—from small-business owners to brand managers and beyond—to connect with AIs and AI developers who might have already developed the perfect solution.
“Say someone has a computer system doing something, and they would like AI to do it better,” Goertzel said. “Unless you’re Google or Tencent, you’re probably not going to build your own AI—you’re going to be making your software grab AI from some software package or API. Those are the customers.”
But it can be difficult for small teams to be sure they’re picking the right AI—and if they’re running on a lean budget, this can end up being a costly choice. SingularityNET helps matchmake these customers with the developers who might have been working on AIs with niche purposes.
“A lot of niche markets will be served by a decentralized network in the same way eBay serves a lot of weird niche markets for buying stuff that Walmart doesn’t serve, because in a decentralized market the overhead cost is very low,” Goertzel said. “Someone who makes AI to serve some niche purpose can reach the customers who want something meeting that niche purpose.”
AI is already integrating into many existing industries, so the potential impact of a barrier-free global AI marketplace extends far beyond applications in tech companies.
“It could be shoe store that wants machine learning to predict what shoes people are going to buy next year, it could be an accounting company that wants AI to spot errors in their line items,” Goertzel said. “Or it could be something sexier like a robotics company that wants AI to help their robot navigate. It could be a hospital that wants AI to help a conversation system to keep chronically ill people company or a diagnostics system to figure out why some medicines work for some and not others.”
AIs Learning from Other AIs
This is where SingularityNET really starts to sound like science fiction—even though it’s about to become our reality.
“What makes it more interesting is that the AIs can also exchange data and make requests of each other,” Goertzel said. “So then the AIs interconnect in a network of AIs, like an economy of minds—a big meta-mind out of all these AIs talking to each other.”
This is facilitated through the Ethereum blockchain, on which SingularityNET is built. For a quick refresher on blockchain, check out these primers. For the purposes of understanding how it fits into the picture of SingularityNET, the most important thing to know is that blockchain allows buyers and sellers to interact directly, without middlemen.
“The big a-ha moment was realizing that this same framework that enables this nice business story of disintermediating the gap between producers and consumers of AI is also actually the rights structure for leading to the emergence of higher degrees of [artificial] general intelligence,” Goertzel said. “If you want to have a system in which general intelligence can emerge, you want it to be—at the high level—more self-organizing and growing and adaptive, like a biological system. You want something in which all sorts of different AIs can be placed, and they can interact with each other liberally, and where the mechanisms for their interactions themselves can be improved and changed and proposed by the participants of the system.”
Goertzel, who notes how biological organisms are themselves self-organizing information systems, believes that an economy like SingularityNET facilitates optimal growth on a global scale.
“You want a big messy system that’s growing without the central controller, with a bunch of different parts interacting with each other, and where the parts that fit well into the whole network prevail and the parts that don’t will not get as much attention,” Goertzel said. “That’s kind of what Marvin Minsky, the AI pioneer was talking about with ‘Society of Mind,’ and then [with SingularityNET] you’re talking about an ‘Economy of Minds.’ The existence of the economy means that AIs that are really useful in that society will get more resources and ones that aren’t will get less.”
AGI Token Sale
In many ways, SingularityNET is the summation of Goertzel’s life work. In the late 1990s, he founded a company called WebMind, with the goal of turning the world wide web into a self-organizing brain. But the idea proved too early, and WebMind was a casualty of the dot-com crash.
A decade later, Bitcoin introduced blockchain technology to the world, gaining traction in the mainstream with each passing year. During that time, Goertzel worked as Chief Scientist of Hanson Robotics (the same company that built Sophia) alongside David Hanson, now the Robotics Lead of SingularityNET. The notion of something like SingularityNET had been in his head a long time, and when that coalesced with the possibilities of blockchain, it all clicked.
“It seemed like many of the same things we want for fostering the emergence of AGI are the same things you want for making a more sort of democratic and broadly beneficial network of AIs,” Goertzel said. “Seeing that those two things are achieved by the same goals…it’s one of those things that’s extremely obvious in hindsight.”
By building SingularityNET on the blockchain, Goertzel has the opportunity to create a decentralized system optimized for ease, reliability, and security.
“The core thing that blockchain gives us is a very easy-to-use, scalable, robust, and secure tool for creating a system of agents living all around the Internet on different people’s machines, which can operate in a trustless, self-organizing way,” Goertzel said. “You can have a reputation system where agents rate each other based on how well they did, and then you have built-in ways to protect against gaming and faking that reputation system. You can have agents pay each other for microtransactions without needing to incur the overhead of using Visa or Mastercard or bank transfers.”
Blockchain answered one of the main issues WebMind ran up against two decades ago: the relationship between cost and functionality.
“Back then encryption was insanely expensive, you know, how do you handle data privacy?” Goertzel said. “Now homomorphic encryption integrated in the blockchain let’s you flexibly expose one aspect of data to one customer, one aspect to another customer.”
To that end, SingularityNET announced its Token Sale for AGI Token on Nov. 28, already receiving over $150M in pledged orders from approximately 10,000 registrants, with institutional backing from Blockchain Global, DigitalX, and Taas.Fund. Due to how oversubscribed the sale has become, SingularityNET has extended their token sale through 2018.
Because AGI Token is an ERC20 token, ‘smart contracts’ will be an integral part of SingularityNET, allowing not only people to transact with other people, but AIs to transact and cooperate with other AIs.
“If the peripheral node that’s directly serving a business—like summarizing a document, recognizing patterns in a picture, asking a question—if that’s accepting currency, it then has to dispense currency to other nodes to which it outsources work,” Goertzel said. “The node that recognizes errors in some line items in a spreadsheet, when it finds a word it doesn’t recognize in some column, it wants to outsource analysis of the meaning of that word to another AI agent—it should pay that AI agent for that. So then you need a token for internal transactions between the AIs.”
That token, of course, needs to be able to keep up with AI—and have a structure to reward AIs and human actors who benefit the overall network.
“That token should support fast, low-cost, microtransactions,” Goertzel said. “You also want to be able to use that token for rewards in a flexible way, to reward agents that do things that are good for the overall network.”
The token’s name is drawn from “Artificial General Intelligence,” the perceived intelligence of a human being. This is situated between “Artificial Narrow Intelligence” (ANI) and “Artificial Super Intelligence” (ASI). The general consensus is that we’re currently somewhere in the higher end of ANI—but once SingularityNET comes to life, it’s not likely to stay that way for long.
For more information on SingularityNET, including the token sale, visit the official website. Join the Telegram channel to chat with the team and community.
Editor’s Note: Neither the author nor Forbes endorses participation in any token sale or cryptocurrency investment, all of which have significant inherent risk. Seek advice from a financial advisor as well as do your own due diligence before considering investment.
Follow me on
Twitter or
LinkedIn. Check out my
website or some of my other work
here.
The post AI Is The Future Of Computing, And SingularityNET Is The Future Of AI appeared first on SingularityNET.
]]>