As we explore the frontiers of artificial intelligence, I find myself constantly reflecting on the dual nature of the technology we are pioneering. At its core, AI is not just a collection of algorithms and data sets. It is an expression of our collective ingenuity to solve the most complex problems facing humanity. But as co-founder and CEO of Lemurian Labs, I am acutely aware of the responsibility that comes with our race to integrate AI into the fabric of everyday life. This makes us ask a question. How can we harness the infinite potential of AI without compromising the health of our planet?
Innovation in Global Warming Aspects
Technological innovations always come at the expense of unexplained side effects. Today, AI requires more energy than other types of computing. The International Energy Agency recently reported that it uses more power than training a single model. 100 American households consume in a year.. All that energy comes at a cost not only for developers, but for the planet. Last year, energy-related CO2 emissions reached a record high. 37.4 billion tons. AI is not slowing down, so we need to ask ourselves: Is the energy required to power AI and the resulting impact on the planet worth it? Is AI more important than breathing air? I hope it never gets to the point where it becomes a reality, but it's not that far off if nothing changes.
I'm not alone in calling for more energy efficiency across AI. At the recent Bosch Connected World Conference, Elon Musk said that with AI, we are “on the edge of probably the biggest technological revolution there has ever been,” but that power shortages could occur as early as next year. He said it could start. AI's power consumption is not just a technical problem, it is a global problem.
Imagine AI as a Complex System
Addressing these inefficiencies requires viewing AI as a complex system with many interconnected, moving parts rather than a stand-alone technology. This system includes everything from the algorithms we write to the libraries we rely on, compilers, runtimes, drivers, hardware, and the energy required to run it all. By adopting this holistic perspective, we can identify and address inefficiencies at all levels of AI development, paving the way for solutions that are not only technologically advanced but also environmentally responsible. Understanding AI as a network of interconnected systems and processes can illuminate the path to innovative solutions that are both effective and efficient.
Universal software stack for AI
Current AI development processes are highly fragmented. Each type of hardware requires a specific software stack that runs only on that device and many specialized tools and libraries optimized for different problems, many of which are largely incompatible. Developers are already struggling to program the system-on-chip (SoC) of edge devices like mobile phones, but soon everything that happens on mobile will happen in the data center and it will be 100 times more complex. Developers must work together and work through complex systems made up of different programming models, libraries, and more to extract performance from increasingly heterogeneous clusters. And that's just for training. For example, programming and achieving performance on supercomputers with thousands to tens of thousands of CPUs and GPUs is time-consuming and requires very specialized knowledge, and yet, with current programming models, much remains to be done, as this is not the case. Scaling to this level will result in excessive energy consumption, and the situation will only get worse as we continue to scale the model.
To solve this problem, we need some kind of universal software stack that addresses fragmentation and makes it easier to program and gain performance on increasingly diverse hardware from existing vendors while also making it easier to gain productivity on new hardware from new entrants. . This will also help accelerate innovation in AI and computer architecture and increase AI adoption across more industries and applications.
Demand for Efficient Hardware
In addition to implementing a general-purpose software stack, it is important to consider optimizing the underlying hardware to increase performance and efficiency. Although graphics processing units (GPUs), originally designed for gaming, are incredibly powerful and useful, they are prone to many inefficiencies that become more apparent as they scale from data centers to supercomputer levels. The infinite scaling of current GPUs increases development costs, lacks hardware availability, and significantly increases CO2 emissions.
Not only are these challenges a formidable barrier to entry, but their impact is being felt across the industry. Let's face it, if the world's largest technology companies are having trouble securing enough GPUs and enough energy to power their data centers, there's no hope for any of us.
a pivotal pivot
At Lemurian Labs, we faced this problem firsthand. In 2018, we were a small AI startup trying to build a foundational model, but it was too expensive. The computing power required alone was enough to drive development costs to levels that would have been unachievable for us, as a small startup, and for anyone outside of the world's largest technology companies. This inspired us to shift from AI development to solving the fundamental problems that make AI inaccessible.
We started from the ground up, developing an entirely new foundational arithmetic to power AI. This innovative numerical system, called Parallel Adaptive Logarithmic (PAL), allowed us to create a processor that can achieve up to 20x more throughput than traditional GPUs on benchmark AI workloads while consuming half the power.
Our unwavering commitment to making AI more efficient and accessible while also making the lives of AI developers easier has led us to always try to peel back the layers of the onion and gain a deeper understanding of the problem. From designing ultra-high-performance and efficient computer architectures designed to scale from the edge to the data center, to creating software stacks that solve single heterogeneous device programming problems down to warehouse-scale computers. All of this helps enable faster AI deployment at a lower cost, increases developer productivity, and accelerates workloads while improving accessibility and driving innovation, adoption, and equity.
Achieving AI for Everyone
If we want AI to have a meaningful impact on our world, we need to make sure we don't destroy it in the process, and that requires fundamentally changing the way we develop AI. Today, the cost and computing required has skewed scale in favor of the few, creating enormous barriers to innovation and accessibility, and emitting massive amounts of CO2 into the atmosphere. By thinking about AI development from the perspective of developers and the planet, we can address these fundamental inefficiencies and achieve a future for AI that is accessible to all and environmentally responsible.
Personal reflection and call to action for sustainable AI
Looking ahead, my feelings about the future of AI are a mix of optimism and caution. I am optimistic about AI's transformative potential to change the world for the better, but cautious about the enormous responsibility that comes with it. I envision a future where the direction of AI is determined not only by technological advancements, but also by unwavering commitment to sustainability, equity, and inclusivity. Leading Lemurian Labs, I prioritize both the betterment of humanity and the preservation of the environment, with the vision that AI is a pivotal force for positive change. This mission goes beyond creating superior technology. It's about pioneering innovation that highlights the importance of thoughtful, scalable solutions that are beneficial, ethically sound, and respect our collective aspirations and the health of our planet.
As we approach a new era of AI development, our call to action is clear. We must foster AI in a way that conscientiously considers its environmental impact and upholds the common good. This spirit is the cornerstone of our work at Lemurian Labs, inspiring us to innovate, collaborate, and set precedents. “Let’s not build AI just to innovate; let’s innovate for people and the planet.” I urge the global community to join us in reshaping the AI landscape. Together, we can ensure that AI emerges as a beacon of positive change, empowering humanity and protecting our planet for future generations.