Bringing human cognition and artificial intelligence (AI) together is the hallmark of the Fifth Industrial Revolution, an era, beginning now, when people and robots work collaboratively to the benefit of society. Industry 5.0 is pushing compute beyond the edge to a world in which humans thrive as never before — all because of AI.
A clear example of this evolution is OpenAI’s ChatGPT™. Until now, AI models excelled in ingesting lots of data, identifying patterns and pinpointing root causes from a diagnostic perspective. Today, most AI researchers are focused on the next phase, generative AI. And that’s not just because of ChatGPT buzz — it’s also because of the profound potential benefits to enterprises.
“At Micron, a big application of generative AI is in our smart searches,” notes Koen De Backer, vice president of smart manufacturing and artificial intelligence. “Think of internet search results — the ones you have to click on or comb through to understand their value. Now think of a ChatGPT inquiry, which does all this evaluation for you and presents it in a comprehensive summary. We are applying that level of smart functionality to Micron. The efficiency is staggering.”
Yet generative AI is worrying to many: Will I lose my job to a robot? Will I have to give up driving? Is my personal privacy gone forever? In Industry 5.0, these worries won’t feel so pressing. Enabled by new technologies, machines will naturally perform the tasks they do best, freeing us to focus on other more important tasks.
In fact, far from taking everyone’s jobs, this new technology augments and empowers us. At Micron, AI in our manufacturing process means that our teams no longer focus on mundane tasks. Rather, our people are freed up to think creatively and test out innovative insights and actions that could help develop efficient, sustainable products.
The first four revolutionsA recap:
- Mechanization — 1780. The first industrial revolution, occurring over about 100 years from the mid-18th to mid-19th centuries, began with the use of water and steam power to mechanize manufacturing processes.
- Electrification — 1870. In the late 19th and early 20th centuries, electric power came to factories, enabling the assembly line and mass production.
- Automation — 1970. Digital technologies, including robotics, came to the manufacturing process starting around 1970, automating many tasks that humans had previously performed and, with the internet, enabling globalization.
- Connection and digitization — 2011. Everything — from cars to computers to robots to toasters — is becoming virtually linked in the Connected Age, communicating with and even controlling one another with minimal human intervention. Factories are on their way to running themselves. “Cyber-physical systems” take charge of not only manufacturing but also procurement, maintenance and repairs. The internet of things, robotics and AI are the technologies enabling all this autonomy, which, like the human brain, is driven by data, analytics and memory.
As we know, digital technology has sped up time. Everything happens faster now, which explains why the fourth revolution — the Connected Age — followed so closely on the heels of the third, the Age of Automation. It comes as no surprise that we are already entering Industry 5.0, the Collaborative Age.
Industry 5.0: the human-machine convergence
The Fifth Industrial Revolution is seeing the beginnings of the convergence of humans and machines. Smartphones and applications are giving way to technologies that live on our bodies, with virtual assistants murmuring directions in our ear, suggesting restaurants for dinner, making reservations on our behalf, and much more. But the most paradigm-shattering changes will occur in the workplace.
Industry 5.0 is about transforming Industry 4.0’s “cyber-physical” manufacturing plants — those using digital technologies to operate factories with minimal human involvement – into “human-cyber-physical” systems.
In this new paradigm, people work alongside collaborative robots, or “cobots,” teaching them to do jobs and correcting them when they err. While machines perform the most menial, repetitive and dangerous tasks, people use their intricate, flexible brains to make high-level decisions. For example, a person could now focus on designing products and processes with a “digital twin,” a virtual copy of the factory where a product gets made or the environment where the process is used. Along the way, in certain industries, a factory’s ability to communicate directly with customers will enable it to customize and personalize every product. Imagine being able to go to a car manufacturer’s website, choose the car you’d like, and select thousands of features that personalize the car for your use!
Of course, smart factories don’t run themselves; they rely on a human force to program, instruct, guide and troubleshoot. The speed at which a factory’s robots can process, analyze and respond to data coming from varied sources — sensors, online orders, computing devices and wearables — depends on how fast their processors are and how much memory they have. (What’s true for human intelligence is also true for artificial intelligence.)
Memory makes it work
AI relies on memory and processing speed to generate the right response at the right time. Self-driving cars sort through streams of data coming from multiple sources to make snap decisions — all with a failure tolerance of zero. Manufacturing plants scale production up or down, order supplies, ship out finished products, and repair and replace equipment autonomously.
Industry 5.0, like the fourth revolution, relies on data, devices and generative AI. None of these components works without memory. Memory, in fact, puts the “intelligence” in AI, providing it with the data to run its algorithms and the context for its actions and reactions.
Everything we do happens as a result of sensory input: going to lunch, laughing at a joke, saying “I love you” or buying a car. To perform each of these actions, we take in information coming from our senses — of sight, smell, taste, hearing and touch — as well as our memories, emotions, beliefs, thoughts and intuition. Then we process it all at once. Unlike central processing units (CPUs), our brains don’t have a discrete number of “cores” where data goes in, is analyzed and sorted, and gets sent out for an action or result. Our brains break up incoming information and assign each part to its corresponding area of specialty — one area for visual data, another for sound, another for emotion and so on.
Likewise, instead of using CPUs to process data, most AI systems use graphic processing units (GPUs), a different kind of computing chip that needs a different kind of memory to maximize performance. While a CPU may have eight, 16 or 32 processing cores on a chip or chiplet, GPUs have thousands. This lets them process thousands of data inputs at once, which is what data-hungry AI workloads require.
Micron’s high-bandwidth memory (HBM) — specifically our latest HBM3 Gen2, the world’s fastest, most power-efficient high-bandwidth memory — feeds these hungry GPU cores with enough data to satiate these powerful cognitive computing chips. Our industry-leading 232-layer NAND supports vast quantities of data storage for AI, including the top-of-the-line Micron 9400 NVMe™ SSD, which shows up to 25% higher performance and 23% lower response time for graphic direct storage (GDS) in AI workloads1. The result is AI — equipped with vast and expansive memory and storage solutions — that reacts in near real time.
At Micron, we see generative AI, robots, drones, self-driving cars and other forms of AI excelling in learning, intelligence and response times. So, we’re using it to foundationally optimize our processes. From manufacturing to business processes, we’re transforming into an AI smart ecosystem across the enterprise, innovating memory and storage to supercharge Industry 5.0. Essentially, we’re building something that is completely differentiated, with great promise for the future.
1 25% higher performance and 23% lower response time compared to competition when performing 4KB transfer in a busy GDS system.