AGI Era: New Moore's Law for Services
My keynote at UC Berkeley and lecture at USC Deep Learning (Part 1)
Last week, I had the honor of delivering the keynote, "AGI Era: New Moore’s Law for Services," at the 18th Asia Business Conference held at the University of California, Berkeley. Today I gave the same talk to USC Computer Science students too. Here is the Part I reflection of my talk.
The prevailing notion that "service businesses" are inherently unscalable is being challenged because of AI.
As we evolve from traditional, programmer-driven software development paradigm in the old Moore's Law era to embracing data-driven, general-purpose AI model paradigm in the new Moore's Law era, the creation of AI-powered, sophisticated, custom, contextual, and personalized software services has become not just feasible but scalable.
I can foresee having my own versions of Coursera courses, Netflix movies, or Intuit TurboTax services, tailored specifically to my unique needs. Note that here we are not talking about the selection of the contents, but the creation of the contents. If you desire the same, or even better, wish to create one for others, please read on to discover how this transformative era is redefining what's possible. 💪 😀
Please read on to explore how this transformative era is redefining what's possible.
1. The Era of Traditional Moore's Law
Originally pointed out by Intel Corporation's cofounder, Gordon Moore, Moore's Law posited that the number of transistors on a microchip doubles roughly every two years, with a concurrent halving of costs. This law underscored the exponential growth in hardware computing power, albeit with minimal cost increases.
The software industry initially focused on bespoke software development tailored for each specific customer. It showed its limitations due to the finite and slow-growing pool of software engineers.
Over the last three decades, a pivot toward general-purpose software (e.g., the on-prem version of software from Microsoft, VMware, and Adobe) has occurred. Then the deployment and maintenance of such software encountered bottlenecks due to the limited number of IT professionals capable of managing these tasks, leading to the advent of cloud computing (e.g., Google Gmail, Microsoft Azure, Salesforce, Workday, and Adobe Creative Cloud) as a means to scale software deployment and utilization more efficiently.
However, the availability of IT professionals and software engineers remains limited, a challenge highlighted by Microsoft's CEO, Satya Nadella, as a "digital currency" bottleneck. Because of the limited "digital currency", it is hard for software to provide custom services.
For example, while platforms like Coursera provide courses, Netflix offers movies, and Intuit offers TurboTax services for everyone, the creation of customized courses, personalized movies, or sophisticated tax preparation tailored to individual preferences and needs has been prohibitively expensive and impractical until now.
2. Debating the Death of Moore's Law
The claim that "Moore's Law is dead" has stirred a significant debate. Since 2017, NVIDIA's CEO, Jensen Huang, has highlighted the increasing costs and energy requirements for more powerful hardware, declaring the era of Moore's Law over.
However Intel's CEO, Pat Gelsinger, has shown strong long term optimism in overcoming physics challenges with innovations such as PowerVIA and RibbonFET technologies, aiming for chips with one trillion transistors by 2030.
Even Nvidia's new Blackwell chip—which more than doubles transistor count to 208 billion and significantly enhances performance—demonstrated continued alignment with Moore's principles for now.
3. The Era of New Moore's Law
The discussions around Moore's Law's death pale in comparison to the significance of entering a new AI era. This era, as highlighted by Nvidia's CEO, is defined by the transformative potential of data center-scale computing. Nvidia's VP of AI Solutions Matthew Hull discussed the data center scale computing perspective with me in our recent interview too.
This shift moves the focus from chip-level optimizations to leveraging the broad capabilities of data centers, embodying a stochastic form of growth.
This new era has spurred a generative AI renaissance, enabling the development and application of foundation models to address complex challenges through prompt engineering and other mechanisms.
For example, it opens the door to highly customized services finally, such as personalized educational programs, personalized movie, personalized nurse chatbot, or personalized tax advisor at a feasible cost.
As we transition into the new era of Moore's Law, it's imperative to move beyond old debates and capture the extensive opportunities provided by data center-scale computing and AI-driven innovations. This development caters to the rising demand for tailored services in sectors like education, entertainment, and healthcare
The notion that "service businesses can't scale?" It is soon an outdated concept in a world bustling with billions of data creators, far outnumbering the millions of programmers. Today, everyone on the planet is contributing to programming and continuously generating data. This is the perfect time to reinvent service industries, from healthcare to education, through groundbreaking, data-oriented approaches.