Countdown to AGI: Navigating the Exponential Growth of Computing Power with Joseph Enochs at AIE 2024
Transcript
Welcome everybody. This is gonna be a talk on the countdown to a GI. I'm Joseph Phoenix, managing director for Emerging Technologies at EVT.
I've been building large scale computing environments for over two decades, and throughout my career, I've been dedicated to developing accessible and ethical tools for the enterprises We serve. From pioneering the LA County covid tracking system to curating the countdown to a GI newsletter. My goal has always been to democratize technology, making it safer and more ethical and accessible to everyone.
Today as we explore the transformative potential of GPU technology and ushering in a new era of artificial intelligence, we delve not just into the future of technology, but into the future of our experience. Every day, each one of us interacts with machines that learn from our preferences, manage our homes, and even suggest our favorite next song. This is Artificial Narrow Intelligence, but imagine a world where these machines do much more than perform tasks.
They solve crisis, they heal diseases and bridge divides. Today we stand at the precipice of this world. Underpinning these advancements is the GPU, the Graphical Processing Unit, which have evolved from mirror graphic processors to the engines that power the most advanced artificial intelligence in AI and deep learning.
Looking at the chart, at the top right, we see the release dates for the most advanced GPUs, the A 100, they're in blue, the H 100 in green, and the recently released B 200 in red at the bottom right in purple, we see the comparison of these GPUs to the extraordinary computing power of the human mind, which is estimated at one exa flop, or one with around 18 zeros behind it quite a bit. So now as we take a minute to reflect exactly how many of these advanced graphical processing units would it take to provide an equivalent amount of computing power to the human mind. Well, as you can see here in blue, the A 100 would take approximately 3,200 of these GPUs.
The H 100 would take about an estimated 1000, and the latest B 200 would take approximately a hundred of these to equate to the human level of computing processing. I want to take a minute to reflect on the concept of what we call a GI or artificial general intelligence, unlike its predecessor, that was the narrow intelligence that helps us get from point A to point B or things of that nature. This specific tasks such as language translation or image recognition, this a GI, gets us to this next destination.
A GI represents a transformative shift. It's characterized by its ability to understand, learn, and apply intelligence across a broad range of tasks, not just matching, but potentially exceeding our capabilities in these areas. As we take a look at the model above and estimate timing of a GI, we see an interesting intersection where the power of a single GPU processing unit reaches the estimated compute power of the human mind around the year 2029.
Let us now turn for a moment to one of the most astonishing marvels in existence. The human mind. Remember this remarkable supercomputer we have in our brains performs computations at an estimated rate of one exaflop.
That's about a billion, billion calculations per second, and yet it does this all while consuming only about 20 watts of power. To put that into perspective, that's about the same amount of energy used in a household light bulb. Contrast this with the latest artificial advancements in GPU technology.
While these machines are at the forefront of our computational capabilities, they come with a cost. These modern GPUs require substantial energy inputs and depend on elaborate cooling systems to achieve their peak performance. This stark contrast serves as a potent reminder of the grandeur and efficiency of the human mind.
However, as we push the boundaries of what's technologically possible, we can see a promising trend, the gap in energy efficiency between the human mind and its natural genius. And the brute force of artificial machine computation is narrowing. Engineers and scientists are continuously innovating to create more energy efficient GPUs, drawing inspiration from the very efficiency of the human mind.
We marvel at today. This pursuit is not just about achieving more power, it's about ushering in a new era of computing that aligns closer with the principles of sustainability and efficiency that nature exemplifies. By bridging this gap, we not only make these powerful technologies more sustainable, but also ensure that they can be integrated into our world in a way that respects the delicate balance of our natural environment.
As we continue our explana exploration in the future shaped by artificial intelligence, let us envision a milestone that was once seemed reserved for the realms of scientific science fiction. Artificial super intelligence or A-S-I-A-S-I represents a paradigm shift in our capability to process and synthesize information, a leap towards harnessing the computational power equivalent to 1 million human minds. Imagine a world where a SI is not just a con a concept, but a reality in this world.
AI a SI assists us in unraveling complex global issues that have long plagued humanity like climate change, rampant diseases and acute resource scarcity. Its ability to analyze data on an unprecedented scale allows for insights and solutions that are currently beyond our reach. We can now see an era of medical breakthroughs where personalized medicine becomes the norm, not the exception.
And diseases that today are chronic and incurable could be understood and treated at the molecular level tailored to each individual's genetic makeup and engineering. We could lead to the development of sustainable cities, smart infrastructure, and energy systems that dramatically reduce, reduce our environmental footprint. I wanna take a minute to take a look back at some projections that were made in the early 2010s.
This is a gentleman by the name of Ray Kurzwell from Google, and surprisingly, his project predictions and projections, um, have actually remained unrefined surprisingly, um, to, to our projections today. As leaders in this field, though, we have to ask some very, uh, difficult questions and we feel it's our obligation not only to push the boundaries of technology, but to also provide these sort of clear and transparent timelines. These are not just mere estimates.
They're part of our commitment to preparing everyone for the changes that are ahead. It's essential that we start these conversations now, engage in policymaking, create educational initiatives along with job training that can help society adapt to these for co forthcoming changes. Now, just recently, uh, a few weeks ago, as as I said, the projections from Ray from the 2010s and 2013 timeframe were reaffirmed.
And as you can see here on the screen, in a debate with Jeffrey Hinton called the Godfather of ai. Uh, Ray brought up his projection and he can, as you can see here, he's saying that a single singularity unit or processing unit, as he calls it, will be equivalent to a million humans. Now, for the minute, why does he call this a singularity unit?
Well, his predictions about the capabilities and advancements, which we'll talk about here more in a moment, they are increased up to a million human minds. After that. It's very difficult for Ray to predict and project past that point.
So what Ray has done is he took a term from physics, uh, and that physics is associated with the event horizon. Or if we look at the black hole, once you pass that event horizon, our physics breaks down. And so that's where in this scenario, we have adopted this, uh, term from physics called the singularity, where our ability to predict the future breaks down.
But with such transformative power at our fingertips, we must address some very hard questions. What does it mean for society? If machines can outthink us, how will we adapt our educational systems and job markets and legal frameworks to this new reality?
And importantly, how do we ensure these advancements benefit all of humanity, not just a select few? Now, as we approach this unprecedented era, our role extends beyond development and innovation. We must be stewards of this technology guiding its integration into society responsibly and ethically.
We are tasked with ensuring that as we step into this future, we do so with a commitment to the betterment of all lives, paving the way for a world where technology and humanity can coexist in harmony. Now, I want to take a few minutes to thank you, but I wanna leave you with a few final thoughts and questions as we conclude, let us reflect on our journey today from the seeds of modern AI and GPU technology to the towering possibilities of tomorrow. Let's consider why and how we together can shape a future that honors our highest ideals and aspirations.
In future weeks in the countdown to a GI, we will continue to talk about jobs and preparations for this transformation to post labor society. So please join us and please give your feedback as we track down and we count down to a GI thank you.
