A Key Redesign for AI and Blockchain Unleashes Their Full Potential
Some of the brightest minds in tech, finance, industry and academia have rushed to the blockchain frontier to set stakes in a new digital economy. A few years back, Dr. Bill Li, co-founder of MATRIX AI Network, approached Dr. Steve Deng, Associate Professor of Computing at Tsinghua University and Head of the Bayesian Computing Lab, with something all great scientists crave for — an intriguing unsolved problem.
The challenge went: “Can AI make blockchain a truly feasible technology?”
An anticipated revolution
Blockchain’s promise to serve as the backbone of all other technologies has spurred a wave of aspiration to restructure entire sectors of the economy by integrating blockchain with tech applications like AI and Internet of Things (IoT). While governments, enterprises and big banks remain apprehensive about blockchain’s swift growth, first movers are convinced that this innovation is not the next cloud — It’s the next internet.
AI is widely believed to be the key to blockchain’s breakout, but AI’s progress has been throttled by its own technical challenges. A nagging economic paradox continues to feed skepticism around transformative technologies. In their paper AI and the Productivity Paradox, researchers from the MIT Sloan School of Management and the University of Chicago show that over the recent period of innovation in AI and cloud services, the growth rate for labor productivity in the U.S. steadily fell to less than half its rate in 2005.
The authors write: “The most impressive capabilities of AI, particularly those based on machine learning, have not yet diffused widely. More importantly, like other general purpose technologies, their full effects won’t be realized until waves of complementary innovations are developed and implemented.”
The fact that the touted industry-shaping impact of AI or blockchain remains largely unfelt could be partly obscured by the complex interdependencies of implementing today’s technology systems. Furthermore, innovations in these fields have edged forward in a largely uncoordinated fashion, cobbling together a string of application-specific solutions. While many blockchain projects have grand ambitions to marshal vast sectors of the digital economy, to date, blockchain design simply hasn’t taken a robust enough systemic approach to channel true market and ecosystem forces.
When we look back at the sweeping social transformation of Edison and Ford, they didn’t simply introduce a disruptive technology; they shared an integrated vision that rallied a global proliferation of new infrastructure and supporting businesses that created a lasting surge in economic productivity. On the surface, it appears this is what the FAANG and BAT tech giants have already done for the modern digital landscape, but there is a still a major disconnect for productivity in the real economy. Many investors and developer teams believe that blockchain is precisely the missing driver between the digital and real economies, helping realize new synergistic efficiencies between the growing suite of technologies.
The first systematic approach to integrating AI and Blockchain
When Professor Deng, now Chief AI Scientist at MATRIX AI Network, first set out on a blank-sheet approach to redesign blockchain using AI — he first stepped back to do a systematic assessment of the “essential nature” of the technologies.
“We wanted to start with pure original thinking and completely redesign blockchain from the ground up. Initially we didn’t recruit anyone with a background in early stage blockchain development. Bill Li, who introduced me to this intriguing challenge, had a national telecoms background, so he understood the use of high-performance communications technologies across large-scale networks. In addition to AI experts, we also gathered experts in distributed computing who understood network architecture, topology, and latency. We wanted the team to bring fresh eyes to how blockchain had developed to date, and innovate from there,” says Deng.
The co-founders at MATRIX AI Network quickly discovered ways that AI could make breakthrough advances in areas hindering the growth of blockchain. These include slow transaction speeds, cyber security risks, a lack of programming talent and zero value creation from scarce computing resources. “What we never expected to find, were a number of fundamental ways that blockchain technology could profoundly upgrade the scope and power of AI,” says Deng.
“What we never expected to find, were a number of fundamental ways that blockchain technology could profoundly upgrade the scope and power of AI.”
In addition to leading AI teams that ranked first in global competitions among participants from top companies, Deng has served as principal investigator on large-scale critical infrastructure projects like implementing early-warning detection systems for high-speed rail. He also worked in semiconductors, making exacting refinements to increase power at the micro-scale.
Deng and colleagues honed their engineering expertise over decades of working on complex multifactor systems and critical infrastructure with razor-thin tolerance for error. This is a fundamentally different skillset than possessed by the many brilliant young blockchain programmers who create novel, yet minimally viable, products and launch successful ICOs.
“Each of the founders at MATRIX AI Network brought a distinct skill set and perspective, but every innovation was grounded in an integrated system-wide analysis.” The team started by redesigning the structure and all processes of the blockchain to be AI optimized; but the boldest overhaul was a full-stack replacement of the mining mechanism.
Deng realized that the mining mechanism, a core feature of the blockchain, was being taken entirely for granted. The team at MATRIX AI Network retooled the mining process to leverage the massive computational power pooled over the blockchain in order to shatter a long-standing barrier in AI’s development.
“One of the greatest factors limiting AI’s ability to scale and mature is the chronic lack of supercomputing power to support important research. The tragic irony is that this shortfall is happening in the midst of a renaissance for microprocessors and ubiquitous computing. Perhaps the greater significance of Bill’s original challenge was realizing major ways that blockchain can advance AI,” says Deng.
“Perhaps the greater significance of Bill’s original challenge was realizing major ways that blockchain can advance AI.”
Designed to evolve
While the distributed ledger technology gives a public blockchain some of its most valuable characteristics — Including data redundancies, transparency, public verification, and a nearly inexhaustible source of computing power — it presents many daunting design challenges. The decentralized model means the system must assume exposure to bad actors. As such, the standard for cyber security must be the ability to operate safely in a completely trustless environment. This is where the senior scientists and architects from MATRIX AI Network, with their experience consulting on critical infrastructure projects like national telecoms and high-speed rail, introduce a new caliber of investigative rigor to this nascent industry.
At the same time, to become the backbone of all future technologies, blockchain needs to be highly scalable and adaptive. “Our philosophy is, ‘designed for evolution.’ Evolution is an active process, so we never set out to create the optimal system once and for all. We always strive to build in room for the system to evolve. That includes designing features like cross-chain functionality, interoperability with external systems, and creating protocols to safely exchange data packets through secure network gateways. These secure network gateways are commonly used in telecoms, and is one example of how we incorporated different disciplines into our approach to redesigning blockchain,” says Deng.
Leveraging group intelligence
MATRIX AI Network has deep roots in the open source developer community. The team draws on open source resources and contributes a large volume of original code and research for testing and feedback. The core leadership also comes out of the scientific community, with a closely aligned tradition of peer review. The team laid the foundations for an evolving blockchain design by using a primarily open source approach. They are also actively building the talent pipeline by providing training, developer tools and investing in promising projects. The incredible progress of open source development initiatives is integral to driving blockchain innovation, yet it doesn’t go far enough in one important respect.
Beyond human design
History was not solely shaped by geniuses standing on the shoulder of giants, but importantly through making advances in collaborative tools — from language, to libraries, to technology platforms. However, all of these developments have been creations of the human mind, which are then further refined by social forces. The more than one billion USD lost to recent blockchain hacks is a humbling testament to the evolving ways that leading designs can be exploited.
“My approach is to make bold assumptions, but cautious corrections. But the reality is that human design is always based on finite experiences, and as such can never reach a complete understanding,” says Deng. AI can supplement and enhance human design, not only by sharing the spotlight for the glamorous work of inventing shiny new breakthroughs — but more significantly in doing the grunt work of identifying loopholes that well-intended designers have outsourced to hackers and speculators for millennia.
In terms of human innovation, major systematic upgrades often come as a reactionary response when exploitation by opportunistic actors becomes too costly to ignore. From improved defenses in response to warfare, to new financial regulations put in place after a market crash — deterrence and risk management have simply never been attractive enough motivators to serve as a driver of innovation.
Generative Adversarial Networks (GAN) are an important tool in the MATRIX AI Network; from security audits to smart contract review. GAN runs simulated attack scenarios on virtual machines to try and maximize ill-gotten rewards, thereby uncovering loopholes. Deng also uses machine learning techniques to identify opportunities for optimization that can be adopted system wide. “The AI design allows for continuous parameter optimization using reinforcement learning,” explains Deng.
Using balanced market forces to monetize operations
For blockchain to support high-performance operations in a trustless environment, it requires a system design that can effectively channel competing motives, similar to those found in financial markets or natural ecosystems. MATRIX AI Network developed six monetization strategies built directly into the operations of the blockchain to incentivize various roles in the ecosystem.
“The credit for the design innovation around the business model really goes to our CEO, Owen Tao. He brought rich experience in organically building and maintaining large online gaming communities. For example, we created a role for embedded actors that play a monitoring function, and because of their random selection, you never know who is watching.” says Deng.
Serving as an executive in the early online gaming space gave Tao remarkable insight into designing a rewards and punishments system to help the ecosystem self-regulate. “The healthy development of the blockchain relies on each of these participants, so we designed incentives for the various contributions,” says Deng.
Similar to how the internet changed the nature of information — the greatest potential for blockchain technology is to become an integration platform for networking virtually all applications, data systems, IoT devices, and computing power in formerly impossible ways. It’s now clear that the only hope for blockchain to bear the Atlas weight of its anticipated potential is through heavy AI optimization. An equally exciting breakthrough for MATRIX AI Network was realizing that the colossal computing resources aggregated on the blockchain could be redesigned to supply a nearly limitless on-demand computational power for running AI research.