SingularityNET Progress Update
Halfway through the first year following the project’s December 2017 milestone Token Generation Event, SingularityNET now has a world-class team spanning every continent, and a far more detailed plan and architecture for realizing the vision the project co-founders outlined in the original SingularityNET whitepaper.
In the original whitepaper as written in 2017, July 2018 was cited as a target date for a beta version of the platform; but in the light of the details of the system design as currently understood, this was an overly ambitious goal. As of now the software is in a rapidly advancing alpha state rather than beta (as a crude indication of the rapid advance, note the 691 commits in the SingularityNET GitHub repositories during July, which represent a wide variety of forward movements on the AI and platform sides).
The team now has a much more detailed plan for the platform development than it did in 2017, so at this stage can chart out the future with a lot more precision. Since releasing the detailed technology roadmap in June, the SingularityNET tech team has been making systematic progress on the multiple aspects of the platform and AI technology outlined therein.
Founder and CEO Dr. Ben Goertzel says,
“For me and the other SingularityNET co-founders, 2018 has been an incredible learning experience so far. After spending the second half of 2017 spreading the message about bringing AI and blockchain together for the purposes of democratizing applied AI and working toward Artificial General Intelligence, the first half of 2018 has been mostly about building. We’ve built a large and world-class tech team spanning AI, platform and services development, blockchain, cognitive science, bioscience and more.
We have a number of different threads going on the AI side, some more oriented toward deep AGI R&D, and some toward the provision of commercial services via SingularityNET, across various vertical markets. The Hanson Robotics AI team in Hong Kong, in close collaboration with the rest of the SingularityNET team, has made significant strides in cognitive robotics, using the OpenCog cognitive architecture to control Sophia in a large-audience presentation setting for the first time.
“We have also been working with a number of senior software architects who have experience developing scalable platforms at Amazon and Uber, to refine our initial SingularityNET vision and designs into the best possible practical step-by-step software development plan for the platform and infrastructure aspects of our work. Putting this experience together with our AI team’s decades of AI and distributed systems experience, in the first half of 2018 we have overhauled our initial Alpha and made a considerable amount of progress from Alpha toward Beta, and also greatly fleshed out our architectural thinking regarding how to practically realize our longer-term technology goals. We now have a much more detailed plan for the platform development than we did in 2017 and so we can chart out the future with a lot more precision”.
“One thing that’s guiding our development at the moment is our plan to conduct a series of SingularityNET AI software development workshops, beginning in September 2018. A series of video development tutorials will be released alongside, filmed in large part at the new SingularityNET head office in East Tsim Sha Tsui, Hong Kong. In order to support the needs of the workshop participants, along with the rest of the developer community, new features are being added to the Alpha at a rapid pace. The precise breakdown of features that will be needed to constitute the transition from the Alpha to the Beta is still being determined by the tech team and their advisors, but it is anticipated that in the coming months there will be strong progress in moving the various components of the Alpha toward Beta status.
Dr. Goertzel adds,
“All of us on the SingularityNET tech team are acutely aware of the eagerness of our community members, as well as others in the AI and blockchain world, to have a fully polished and scalable SingularityNET platform out there to use. I also have to note that, in hindsight, it was a bit too enthusiastic to do what we did in the original SingularityNET whitepaper, which was to cite July 2018 as a specific date for the beta release. In Fall 2017 when we put that whitepaper together, we knew what we needed to build, but we hadn’t yet broken down the platform design in detail the way you can see on our platform roadmap now. Additionally, in the fall my co-founders and I didn’t anticipate the amount of time and work it would take to recruit and on-board a large high-quality team. While we started expanding shortly after our Token Generation Event in December, our team in its current form only came together in late March and early April and since then it’s been full speed ahead.”
Cassio Pennachin, SingularityNET Founder and Chief AI Officer, breaks down the current state of progress as follows:
“We have been making steady progress on the development of a scalable, user and developer-friendly platform. A lot of recent effort has gone into a redesign of the marketplace UI, in order to make it easier for people to find and hire the AI services they need. On the developer side, we are replacing the old Python daemon with a more scalable Go implementation which will provide tools for monitoring, metering and more flexible blockchain integration. We’re also developing the SNET SDK, which will help developers integrate AI services into their code via automatic code generation.
“We are also working on populating the marketplace with a variety of AI services. Network analytics tools are being refined and tested in a couple of case studies and will soon be available, as well as a variety of deep neural net models which can be used or tuned on customer data, and natural language processing tools for tasks like document summarization, entity detection, sentiment analysis and others. Finally, we continue to forge ahead on a number of very challenging AI research problems, such as semantic perception, where our team is combining state of the art deep neural nets with OpenCog’s natural language processing and reasoning technologies to answer questions about images; and unsupervised grammar induction, where our preliminary results indicate we are on a path to creating AI code that can learn effectively grammatical and semantic patterns from unlabeled corpora. This will be valuable for application areas like social media analytics that require AI understanding of tweets and SMSs and so forth, and also for natural language analytics of languages with limited computational linguistics resources, such as most African languages.
“As a benefit project, which has commercial potential besides, we are collaborating with Leshan Agricultural Institute in Szechuan Province in China to learn deep learning models for plant disease diagnostics from images and other data. We are in discussions aimed at extending this work to multiple regions of Africa as well. This is a prime example of how the decentralized approach to AI can foster a broad variety of beneficial AI applications, beyond the typical scope of interest of big tech companies.”
Dr. Hilen Amin, VP of Solutions, has been focusing on the definition and creation of commercial AI services leveraging AI algorithms integrated into the SingularityNET platform. Dr. Amin notes:
“We are developing a series of Case Studies that demonstrate how AI can be applied to real world problems. We model real world scenarios in such a way that AI can not only describe and predict what will happen next, but also draw us to convincing explanations as to why.
As an example, in a previous case study we analyzed the SingularityNET itself. We investigated how different interaction models between AI algorithms can lead to scenarios where fairness and high quality solutions are encouraged and hacking is discouraged. This was done by gamifying the network and observing the outcomes under different interaction models.
We are reusing the same network analysis tools in our next case study, which will be published later this year. This study investigates the influence campaign contributions has in US politics. Historically shown to be a difficult problem, we are using network analysis and a variety of AI techniques to help tease out the relative influence of popular support, social media and political ads on a variety of political outcomes such as election results and the passage of bills within Congress and Senate. As part of this study we are modeling how social media and smear campaigns instill tribalism and polarization within the population.”
The work underway by SingularityNET staff this year will lay the foundation for the exponential growth in the SingularityNET network during 2019, when it is expected that a rapidly increasing number of AI algorithm and service developers will launch their code onto the SingularityNET, and an increasing number of commercial users will choose to leverage SingularityNET instead of centralized platforms to meet their AI needs.
However, this exponential growth will not be achieved by the internal SingularityNET team alone — the essence of the democratic and decentralized model underlying SingularityNET is that the work of the core team must serve mainly to spur an even greater amount of creative and entrepreneurial activity by the broader community.
Arif Khan, SingularityNET Marketing Director, notes that:
“SingularityNET continues to develop strong partnerships to seed our ecosystem with benevolent and meaningful AI services. With the launch of DAIA, we continue to build momentum to strengthen our relationships in the decentralized ecosystem. We are witnessing a significant increase in inbound leads for strategic partnerships with governments, businesses and developers.
“For example, our recently announced partnership with Mindfire will facilitate access to top-tier university researchers and AI developers and programmers as part of our talent pooling initiative. To effectively deal with this significant increase of inbound interest we have mandated the hiring of key business development and partnership managers across 7 locations in the world’s tech and financial hubs. All of these efforts will continue to be nurtured to develop a robust ecosystem of trusted relationships that would eventually translate to a deeper and richer online SingularityNET platform for our developers, users and community.”
DAIA will enable SingularityNET and other decentralized AI projects to work more effectively together to create a new technology infrastructure forming a viable and superior alternative to that provided by today’s big tech companies. Among other strategies, DAIA will foster shared technical work on projects of common value — to minimize duplicated work and to ensure that maximum brainpower and effort are focused in the right places.
Next Steps
Fall 2018 will be a big season for SingularityNET, as through developer workshops and associated online initiatives, a broader community is drawn into the process of seeding the network with diverse and interacting AIs. And 2019 will be a landmark year, as the network becomes increasingly complex, and the flow of AGI tokens through the network grows, increasing the value and intelligence in the network.
Both the AI and blockchain spaces are evolving rapidly in exciting and novel ways, and SingularityNET’s platform is as well, so we can expect some continued surprises as the SingularityNET adventure evolves. However, the core platform and services work can be expected to proceed in a systematic and steady way as 2018 progresses, according to the component breakdown given in the roadmap. At the same time, it must be understood that some of the current SingularityNET initiatives are more research oriented and difficult to predict progress-wise. A team working on fundamental research like unsupervised language learning, neural-symbolic vision or inference meta-learning will always be susceptible to unpredictable delays, and also to occasional rapid and exciting breakthroughs
We continue to be optimistic about the future and conscious about the promise of the platform, and remain connected to the interests of our community and require your support in enabling this vision and goal.
We welcome discussion of this message on our forum for further updates.
