"We are standing on the brink of a new revolution, one that is catalyzed by AI. It not only heralds the leaders of future scientific discoveries, but also fundamentally changes the way we perceive the world. Like scientific revolutions in history, the development of AI is leading us into a new era where data and algorithms will become new tools for exploring the unknown world."

Author | Dou Dou

Editor | Pi Ye

Produced by | Industrialist

"I had no idea I’d even been nominated for the Nobel Prize in Physics.(I had no idea I'd even been nominated for the Nobel Prize in Physics)"

"How could I be sure it wasn’t a spoof call.(How can I be sure you're not pulling my leg?)"

John J. Hopfield and Geoffrey E. Hinton are the winners of this year's Nobel Prize in Physics. Hinton was surprised when he heard the news. Not because of anything else, but because he works in AI, more accurately known as: artificial neural networks and machine learning.

After all, neural networks, at first glance, don't sound very physical.

The official comment section was also abuzz. Researchers in physics, AI, and even biology fell silent.

The comment section of the Chinese Academy of Sciences' physics account was also heated with discussions. Some netizens joked, "Why didn't the Nobel Prize go to GPT?"

The announcement of the 2024 Nobel Prize in Physics has pushed the application of AI in physics to a climax. The awards to John J. Hopfield and Geoffrey E. Hinton are not only recognition of their individual achievements but also affirmation of the role AI plays in solving scientific problems.

One question is, why did AI win the Nobel Prize in Physics? What information does the "surprise" behind this year's Nobel Prize in Physics reveal? And what is the connection between AI and the ongoing industrial digital transformation?

I. The Deep Connection Between Physics and AI

To understand why these two individuals received the Nobel Prize in Physics, we first need to understand what achievements they made.

"Using tools from physics, they developed fundamental methods for today's powerful machine learning techniques." This is a quote from the Nobel Committee's citation. Literally, it means that they built the underlying methodology for AI technology based on physics.

Hopfield created associative neural networks in 1982, now commonly known as Hopfield networks, which can store and recall associative memories of images and other data patterns. Hinton is the co-inventor of backpropagation and contrastive divergence algorithms and an advocate of deep learning, earning him the title of "Godfather of Deep Learning" or "Godfather of AI".

We'll try to explain the inventions of these two laureates in simple terms.

Imagine a bunch of magnets that can freely flip their north and south poles. These magnets are placed together in a disorganized manner, but there's a special rule: if the north pole of one magnet faces the south pole of another, they attract each other; conversely, if the same poles face each other, they repel each other. This is called a spin glass, a model in physics used to describe collections of particles that are chaotic yet mutually influential.

Now, use this idea to build a "memory storage device." Each small magnet acts like a neuron in a neural network, which can be set to be active (north pole) or inactive (south pole). When you put these "neurons" together, they influence each other and eventually reach a stable state, just as magnets eventually align to minimize repulsion.

The Hopfield network is a type of neural network that mimics the way the brain stores memories. You can "teach" patterns (like a sequence of numbers or an image) to this network, and the neurons in the network will adjust themselves so that they can recall this pattern in the future. It's like arranging a bunch of magnets in a specific way, so they remember this arrangement and can restore it even if they're disrupted.

Now, let's talk about Hinton's invention.

Imagine building a complex model with a bunch of Lego bricks. But the final shape of the model is unclear, and you're not sure how to build it step by step. This is the problem faced when training neural networks: we have a lot of data, we know the desired outcome, but we don't know how to adjust the connections in the network so that it can process the data correctly.

Backpropagation is a solution to this problem. It's like a tool that can undo wrong steps. Every time a brick is placed incorrectly, this tool allows the builder to go back a step, adjust the position of the brick, and keep trying. Through continuous trial and error, the correct model is eventually built. In a neural network, this means gradually adjusting the connections in the network so that it can correctly identify patterns in the data.

The Boltzmann machine is another neural network inspired by physics. It uses a concept from statistical physics, where particles can be in different states with certain probabilities at a given temperature. In a Boltzmann machine, each "neuron" acts like a particle that can be active or inactive, and these states are adjusted based on the overall energy state of the network. In this way, the network can learn how to extract useful information from data, just as statistical physics helps us understand the behavior of particles.

In short, both the Hopfield network and Hinton's work apply concepts from physics to neural networks to mimic the way the brain works, allowing computers to learn, remember, and recognize complex data patterns.

More importantly, Hopfield and Hinton have demonstrated how physics principles can be applied to understand and simulate the workings of the brain, a cross-disciplinary innovation that not only advances AI technology but also provides new research tools and perspectives for physics.

II. AI – The 'Master Key' to Solving Scientific Problems

Current AI technology represents not just a technological innovation but a revolution in thinking.

AI technology, especially deep learning and large models, has demonstrated powerful pattern recognition and data processing capabilities across multiple scientific fields. This ability not only accelerates the pace of scientific research but also enables scientists to explore the tiniest details in vast data oceans and uncover hidden patterns.

Through deep learning algorithms, AI can analyze and interpret complex physical phenomena, such as high-energy collision data in particle physics. This analytical capability enables scientists to more quickly identify new particles and physical phenomena, accelerating the development of theoretical physics.

AI can predict the three-dimensional structure of proteins, crucial for understanding the essence of life and developing new drugs. AI's predictive power shortens experiments that once took years or even decades to just weeks or days.

AI can also predict the behavior of molecules and chemical reactions, accelerating the discovery of new materials and drugs. These applications demonstrate that AI has become an accelerator and catalyst for scientific research.

AI's influence extends beyond physics, biology, and chemistry, showing great potential in medicine, astronomy, and other fields. AI's interdisciplinary nature allows it to connect data and knowledge from different fields, pushing the boundaries of scientific discovery. This cross-disciplinary collaboration Foreshadowing AI Will become the core force of future scientific research 。

However, AI is not omnipotent. Its capabilities are built on vast amounts of data, and the quality and integrity of this data directly impact AI's judgments.

Furthermore, AI's decision-making process is often a "black box," with its internal logic and reasoning not transparent to humans. This raises questions about AI's reliability and ethical responsibility. AI's "black box" nature makes it difficult to fully trust its outputs, especially in medical fields where life safety is at stake.

Despite these challenges, AI's role as a "master key" to solving scientific problems is indispensable. It is driving scientific research into a new era, one driven by data and optimized by algorithms.

III. Reimagining Industries with AI

We are standing on the brink of a new revolution, catalyzed by AI. It not only heralds the leaders of future scientific discoveries but also fundamentally changes the way we perceive the world. Like scientific revolutions in history, the development of AI is leading us into a new era where data and algorithms will become new tools for exploring the unknown world.

AI's applications have also fueled the rise of interdisciplinary research by connecting data and knowledge from different fields, driving new scientific discoveries. This cross-disciplinary collaboration not only accelerates knowledge accumulation but also provides new perspectives for solving complex problems.

More importantly, AI is disrupting the way we understand the world. It offers new perspectives for observing and comprehending the world, such as analyzing social media data to better understand social trends and human behavior. AI's predictive and simulation capabilities are also transforming our understanding and prediction of certain phenomena, such as simulating climate change to help us grasp the impacts of global warming. Meanwhile, AI is aiding humans in making smarter decisions, whether in predicting market trends in finance or optimizing route planning in transportation.

Today, with the support of various service providers, AI technology is continually empowering various industries. From e-commerce to finance and industry, AI is permeating and driving intelligent upgrades across sectors.

In e-commerce, AI technology is already well-established. For instance, JD.com's Jingyan AI assistant provides professional product knowledge and personalized recommendations, helping consumers understand different product categories and their features, while Alibaba's Alixiaomi automates solutions to common queries through natural language processing and big data analysis, offering one-stop services with multiple access methods.

Finance is also a hotbed for AI applications. Hang Seng Electronics' financial AI assistant "Photon" integrates data processing, analysis, and decision support functions, providing professional financial services consulting and automating routine back-office operations like account management and transaction processing.

In industry, AI applications are driving the intelligent transformation of manufacturing. AI's role in production processes includes real-time data collection and analysis, predictive maintenance, and intelligent scheduling, enhancing production efficiency and line stability. Additionally, AI aids in energy conservation and emission reduction by intelligently optimizing and controlling energy usage, lowering energy consumption and emissions. In areas like object sorting, quality inspection, and warehouse automation, AI applications significantly improve efficiency and accuracy.

Moreover, AI technology enhances sales management, accurately predicting sales trends, optimizing sales strategies, elevating customer service, and facilitating efficient, personalized market operations. For instance, AI-based sales forecasting analysis reduces inventory overstock risks and improves cash flow.

Overall, AI technology's applications in industrial digital transformation are continually expanding, enhancing production and operational efficiency while promoting industrial structural optimization and upgrading. With continuous technological advancements and deeper applications, AI is expected to spawn genuine "native applications" in more fields, driving socioeconomic innovation and development.

As AI evolves, our lifestyles, work patterns, and perceptions of the world are undergoing fundamental changes. This AI-driven scientific revolution transcends mere technological innovation, profoundly impacting our social structure and cultural development.

In the future, AI may usher in a new era of fundamental changes, enabling us to explore the unknown, solve complex problems, and comprehend the world's intricacies in unprecedented ways.