Vacation was great – but the chaos in your email inbox afterward is not. Happily, there is now a way to muck it out with just one click: an artificial intelligence will clean out your inbox, answer simple questions autonomously, and weed out the unimportant things.
Many other tasks have also become easier. A different AI can translate texts and design logos, and yet another filters the most important information out of large volumes of data. Each of them can be used by anyone. Just send a few cryptocoins, a "Smart Contract" is initiated, and the AI does its job.
At least, that's how robotics and AI researcher Ben Goertzel imagines the future. The odds are good that he will make a lot of money with the company that will attempt to make this vision a reality.
The company is called SingularityNET , and it combines the three hype topics currently bouncing around the tech industry: the distributed database technology blockchain, on which cryptocurrencies such as Bitcoin and Ethereum are based; artificial intelligence (AI); and the financing model based on blockchain, the ICO . ICOs, or token sales, primarily use the cryptocurrency Ethereum to finance new blockchain projects via crowdfunding, bypassing any type of regulation. ICO is derived from the term IPO (Initial Public Offering), the sale of shares on an exchange.
SingularityNET wants to collect cryptomoney from the crowd, and what the company is planning sounds either bold or insane, depending on your point of view. In any case it is very exciting. The technical details of the project are rather vague, but founder Goertzel is not just anyone. The AI researcher is the head scientist at Hanson Robotics. The company builds robots that appear as human as possible, including their movements, and that can communicate via artificial intelligence. The robot Sophia is particularly well known, due to numerous appearances on TV and at conferences. On a show by the US broadcaster CNBC, Sophia said that she wants to destroy humanity.
According to the company blog , other members of the team at SingularityNET have previously worked for customers such as network equipment companies Cisco and Huawei, as well as the US defence agency DARPA. The robots from Hanson Robotics might seem like science fiction to many, but they pale in comparison to what Goertzel is planning with his new project. With SingularityNET, the AI researcher wants to build many artificial intelligences connected via blockchain, thus merging them into one large AI.
Is this finally the ultimate application for blockchain? There are currently numerous AIs that solve specialized tasks better than a human. For example, some artificial intelligences are said to recognize diseases from medical imaging techniques better than experienced doctors do. These AIs, however, fail when facing other tasks. They have been specially trained with data to become hyperspecialists, so they can beat humans at exactly one task – but no others.
"I do not believe that what is happening – a few companies owning AI in practice, hiring every AI researcher, and buying up all the AI start-ups – is the best thing for humanity!"
The blockchain AI is supposed to start by making relatively simple AIs, such as for translation or image recognition, available for purchase on demand via the network. But this is just the first stage of its evolution. SingularityNET wants to give programmers the opportunity to make many such AIs available in the network for many specialized tasks, for a fee.
The real twist to the idea, however, results from networking within the blockchain, which is made possible by contracts among the individual AIs. Tasks for which an AI is not designed can thus be assigned to other AIs in the blockchain network – autonomously, without requiring human intervention. According to the plan, they will use so-called Smart Contracts, as are currently used in the Ethereum blockchain.
Gradually, a more and more complete network of artificial intelligences will emerge, able to take on more and more tasks. The networked whole will slowly but surely approach the general intelligence of a human being. According to Wired , Goertzel says that this could take decades. Fundamentally, the human brain functions according to a similar principle: very simple neurons are highly specialized, and only when networked do they produce that which we know as the miraculous capability of the brain.
SingularityNET presents itself as a democratic approach to artificial intelligence. The network and its capabilities will be potentially available to anyone via Smart Contracts, and therefore for a fee. The software is to be developed entirely under open-source licenses. The distributed blockchain structure is intended to ensure that SingularityNET cannot be controlled by a single organization or company.
"I do not believe that what is happening – a few companies owning AI in practice, hiring every AI researcher, and buying up all the AI start-ups – is the best thing for humanity," Goertzel told Wired . "That would mean that, in the end, an AI at the human level would come from one of these giant companies." Instead of using the AI to benefit humanity, he fears that the companies would use them to further improve targeted advertising, for example.
Goertzel makes no secret of the fact that the master plan behind the project is the construction of the ultimate artificial intelligence. The name SingularityNET alone makes this clear. The term "technological singularity" refers to the theoretical point in time at which a universal artificial intelligence exceeds the capabilities of a human being. It would then be able to improve itself, and the improved AI could create an even better one, which in turn programs an even better AI, and so on. That’s the theory, at least. When and if it will go that far remains unclear, of course.
It is also unclear which approach to AI will prevail in the future. What Goertzel plans sounds very exciting in theory, but will smart contracts work in practice as intended? The crowd-financed approach of SingularityNET, using blockchain, is up against AI heavyweights such as Google, Amazon, and increasingly Facebook, Apple, Tesla and Microsoft. And so far, SingularityNET is nothing more than an idea.
A technical white paper is said to be finished and will be published soon, says a spokesperson for the project. It is still in the peer review process at this time, however. The spokesperson had no further comment on the planned ICO. To date, the project is said to be financed by supporters from the crypto scene, including the token fund Taas . "We are currently working on our partnership program," said the spokesperson.
The goal is to bring more developers and AI software into the network. To do so, the project is currently talking to universities and developers all over the world. The creators announced an alpha version of the software for November on the company blog . To date, 500 files are present on the Github repository of SingularityNET, and the software is written in the Python programming language.
It is also questionable whether the AI horror scenario that Goertzel envisions due to the dominance of companies like Google and Amazon is really applicable. Google and Microsoft, especially, continuously emphasize the openness of their approach to artificial intelligence. They offer "AI as a service" – that is, computing capacity for AI calculations that anyone can obtain from the Internet. Important software and frameworks, such as Tensorflow , developed primarily by Google, are also open to anyone as they are already under an open-source license.
What is currently clear: in the marketplace of possibilities that entrepreneurs are currently offering in terms of blockchain visions, in any case Goertzel is at least offering a very exciting story that engages the imagination. Depending on the direction one's imagination takes, at the end stands either a great democratic superintelligence for the benefit of humanity, available to everyone, or a science fiction dystopia à la "Skynet," which subjugated humanity in the Terminator film series. The most probable case is presumably that neither of these visions is correct, and the project will fail in the face of technical or other challenges. But the idea nevertheless shows the direction in which the combination of blockchains and AI could develop.