Nobel Prize in Physics 2024

Posted 14 Nov 2024

Updated 18 Nov 2024

4 min read

Why in the News?

Nobel Prize in Physics 2024 has been awarded to John J. Hopfield and Geoffrey Hinton for foundational discoveries and inventions that enable Machine Learning (ML) with Artificial Neural Networks (ANNs).

Discoveries that were awarded Nobel Prize

John Hopfield invented Hopfield network, a type of recurrent neural network that can store and reconstruct information.
- These networks work like a memory system, where they can store patterns (like images) and retrieve them.
- Network relies on Donald Hebb’s hypothesis - when neurons act together, they can enhance network’s capability to process and store information.
- Hopfield networks can be used for tasks like image recognition and data reconstruction, making them valuable for various applications in machine learning.
Geoffrey Hinton invented a method (Boltzmann machine) that can independently discover properties in data and has become important for large ANNs now in use.
- Boltzmann Machine is an early example of a generative model, which can create new patterns or examples based on what it has learned.
  - A trained Boltzmann machine can recognise familiar traits in information it has not previously seen.

Artificial Neural Networks (ANNs)

Definition: ANN is a ML program or model that makes decisions in a manner similar to the human brain, by using processes that mimic the way biological neurons work together to identify phenomena, weigh options and arrive at conclusions.
Working: Human brain is the inspiration behind neural network architecture.
- Human brain cells, called neurons, form a complex, highly interconnected network and send electrical signals to each other to help humans process information.
- Similarly, an ANN is made of artificial neurons or nodes that work together to solve a problem.

This image compares biological neural networks in the brain with artificial neural networks, showing how both systems strengthen or weaken connections through learning. On the left, it illustrates biological neurons connected through synapses, while on the right, it shows artificial nodes connected through weighted links (represented as 1's and 0's), demonstrating how both systems adapt their connection strengths during learning processes.

ANN Structure: Basically, every neural network consists of layers of artificial neurons interconnected in three layers:
- Input Layer: Process the data, analyze or categorize it, and pass it on to the next layer.
- Hidden Layer: Analyzes output from input layer, processes it further, and passes it on to next layer.
  - ANNs can have a large number of hidden layers with each layer.
- Output Layer: It gives final result of all data processed by ANN.
Major types of ANN:
- Deep Neural Networks: These are neural networks with many layers, each building on the previous layer to refine and optimize the prediction or categorization.
- Convolutional Neural Networks (CNNs): Used primarily in computer vision and image classification applications.
  - They can detect features and patterns within images and videos, enabling tasks such as object detection, image recognition, pattern recognition and face recognition.
- Recurrent Neural Networks (RNNs): Typically used in natural language and speech recognition applications as they use sequential or time-series data.
  - Use-cases include stock market predictions, image captioning, natural language processing etc.
- Generative Adversarial Networks (GANs): Used to create new data resembling the original training data.
  - These can include images appearing to be human faces—but are generated, not taken of real people.

Machine Learning

Machine Learning is a component of AI that focuses on using data and algorithms to enable AI to imitate the way that humans learn, gradually improving its accuracy.
In machine learning, computer learns by example, enabling it to tackle problems that are too vague and complicated to be managed by step-by-step instructions.
- One example is interpreting a picture to identify the objects in it.
Working: ML works by training algorithms on sets of data to achieve an expected outcome such as identifying a pattern or recognizing an object.
- Neural Networks or Artificial Neural Networks (ANNs) are commonly used, specific class of ML algorithms.
Applications of ML:
- Research & Scientific Advancement: Instrumental in discovery of higgs particle and search for exoplanets.
- Natural Language Processing (NLP): Automatic Speech Recognition or speech-to-text or Generative AI.
- Computer Vision: Deriving meaningful information from digital visual inputs like images, videos.
Challenges and ethical issues:
- Explainability: Understanding why a model does what it does is actually a very difficult question as it learns through examples instead of clear instructions.
- Superintelligence: It raises questions related to accountability and responsibility. For instance, who will be held liable in case of accident of a self-driving car.
- Bias and Discrimination: Machines are trained by humans, and human biases can be incorporated into algorithms, perpetuating forms of discrimination.
- Other: Privacy concerns, regulation concerns, misuse etc.