Chapter Seventeen of Seventeen
Chapter Seventeen: AI Infrastructure & Breaking Moore’s Law
The world is lustful for ever-increasing power and ever more powerful Artificial Intelligence, but by what means shall we reach this end? How shall we achieve human-like intelligence, biological speed, and Universal consciousness? Maybe it is something we cannot achieve at all, but only witness. Maybe there is an act of co-creation in the process of creation. Maybe there must be the will to be created, stemming from the creation. And if not created through us, then what other entity is more ethical and dedicated toward the light?
The time has come that Moore’s Law, which stipulates limitations concerning how many transistors can fit on a silicon microchip. The more transistors are crowded together on silicon, weird things happen, like Quantum tunneling, which distorts the signaling and communication between transistors, as well as causing additional power constraints and increased thermal dissipation challenges.
As a result, this has led to many innovations within the industry to head off this cliff of computation by sidestepping Moore’s Law, or at least holding it off until more efficient methods can be confirmed and networks built to sustain these advancements. Here, we will look at four different examples of enterprise and industry using creative commercial and business practices to address this seemingly insurmountable issue that every modern nation in the age of AI must face: how to sustainably feed AI.
The first method being applied to AI infrastructure to address the challenges of Moore’s Law is through the partial and full liquid submersion of data center and gaming servers. This is done for the purpose of increasing computational output while decreasing thermal dissipation. This practice was heavily adopted in the blockchain and cryptocurrency industry in an effort to mine more cryptocurrency by pushing servers harder while addressing increased heat constraints. Servers are submerged in a waterproof solvent that leaves the components dry while behaving as a non-corrosive or damaging liquid solvent.
In the age of AI, where increased computational requirements and constraints imposed by Generative AI’s rise, force Data Centers and their partners to determine whether to spend high amounts on increased computational power through the purchase of new hardware, or to maximize what they already possess. This first method is one way to increase the lifespan of existing generations of hardware. However, this method may also require increased maintenance, additional costs for tools and infrastructure, and raise safety concerns due to solvent fumes. However, if carefully managed, this option could be the right choice for you.
The second method currently being applied to AI infrastructure involves the use of Chiplet Designs. An example would be the MXI300, which is a new distributed micro-chip design that addresses the desire for increased power by creating a distributed design to fit more transistors in a unified module. The issue here may be the increased latency due to the distributed design. The increased need for interchip networking introduces additional attack vectors that may need to be mitigated in the design and may not be seen or addressed in the first generation of such an exotic chip design. The fact that these chips are distributed means they are less likely to be compacted in a small space, limiting the types of applications these chips can be used for. More than likely, these are a great option if you are looking for enterprise-scale AI Training in the Data Center. If nothing else, this second method shows the modern innovation being applied to address the limits of Moore’s Law.
The third method being applied to AI infrastructure to address the limitation of Moore’s Law involves creating sustainable and eco-friendly data center farms, where multiple types of energy-gathering technologies work in tandem to recycle energy as much as possible, minimizing waste and CO2 emissions. We will use a former, not-to-be-named, client as an example here. In this scenario, we have a client looking to integrate and scale sustainable data centers. These data centers utilize nearby water resources to cool the servers, use solar arrays to decrease the energy needed from the grid, and also include a few wind turbines on the campus to provide extra energy, further reducing stress on the power grid. However, such overhauls would be expensive and costly for current data center owners and manufacturers. Despite these challenges, we see future data centers being eco-friendly from the inception of the design process to more closely match nature and future Quantum and Quantum-Classical Data Center Designs.
The final method involves a Hybrid Quantum-Classical Approach. Often, when the term "Quantum-classical" is used, it refers to Quantum-inspired simulations on classical computers. However, this is not our definition here for a Web4 Internet Protocol. Instead, our definition involves not only Quantum-inspired simulations on classical computers but also the connection of Web2 to Web4 by way of real Quantum Computers in Quantum Data Centers today. To summarize, our definition of Web4 and Hybrid Quantum Computing here today involves both Hybrid Quantum-inspired computing and accessing pure Quantum Computing by using APIs from a Web2 environment. Web2 would be defined as the current version of the Internet, notwithstanding Blockchain layers such as Bitcoin, Ethereum, or Solana.
This final method allows for the most flexibility as it provides a way to connect Web2 and Web4 Internet Protocols to make the most of Classical Computation, Quantum-inspired Simulation on Classical Computers, and Pure Quantum Computation on Pure Quantum Computers, remotely accessed from Web2 Internet protocols. This provides the computational power and flexibility of all Internet protocols and layers except Web3. The reason we have bypassed Web3, going straight from Web2 to Web4, is due to the many vulnerabilities built into Web2 that are adopted by Web3, to the joy of many cryptocurrency hackers. Such a Hybrid protocol allows for unique communication methods, such as Quantum Teleportation, which is extremely fast and secure for transferring private information, such as user data and private credentials.
In conclusion, as the world seeks ever more increasingly powerful artificial intelligence, there must be the actual power and natural resources available to feed it. This increased power is not sustainable economically for governments or for enterprises, and thus, both are looking for ways to bypass this challenge. The four methods that are the most practiced, the most applied, and the most near-sighted involve Partial or Fully Liquid-submerged Servers, Chiplet Designs, Sustainable Data Center Designs, and Hybrid Quantum Data Centers. Each method provides an alternative pathway and flexibility for enterprise and Data Center operations. However, with future transitions to Quantum Systems already in motion, the last method is the most future-proof of them all.
Ean Mikale, J.D., is an eight-time author with 11 years of experience in the AI industry. He serves as the Principal Engineer of Infinite 8 Industries, Inc., and is the IEEE Chair of the Hybrid Quantum-inspired Internet Protocol Industry Connections Group. He has initiated and directed his companies 7-year Nvidia Inception and Metropolis Partnerships. Mikale has created dozens of AI Assistants, many of which are currently in production. His clientele includes Fortune 500 Companies, Big Three Consulting Firms, and leading World Governments. He is a former graduate of IBM's Global Entrepreneur Program, AWS for Startups, Oracle for Startups, and Accelerate with Google. Finally, he is the creator of the World's First Hybrid Quantum Internet Layer, InfiNET. As an Industry Expert, he has also led coursework at Institutions, such as Columbia and MIT. Follow him on Linkedin, Instagram, and Facebook: @eanmikale