For the Love of PyTorch

Can't read or see images? View this email in a browser

Had there been no PyTorch, there would have been no LLM. Sounds a bit exaggerated, but there’s little doubt about the big role this deep learning framework has played in the development of AI technology. It provides an intuitive and flexible method for constructing neural networks, making it an ideal option for deep learning experiments and prototyping. Besides, the framework is distinctive for its excellent support for GPUs and its use of reverse-mode auto-differentiation, which enables computation graphs to be modified on the fly. The wide popularity of the framework can be gauged from its performance on Hugging Face. In 2022, Hugging Face saw the addition of a remarkable 45,000 PyTorch-exclusive models, whereas only 4,000 new TensorFlow-exclusive models were introduced to the platform. This led to a significant 92% of models on Hugging Face being PyTorch-exclusive, leaving just 8% for TensorFlow. PyTorch has reached here after fighting a very tough battle against TensorFlow. PyTorch and TensorFlow, both working with tensors, differ primarily in computation graph types. PyTorch employs dynamic computation graphs, while TensorFlow traditionally uses static ones, although TensorFlow 2.0 introduced eager execution.  Besides, PyTorch owes part of its success to its seamless integration with NVIDIA's CUDA. CUDA is a favoured framework for AI model development, and PyTorch's code significantly simplifies the work for developers. PyTorch's appeal extends far beyond its technical capabilities; it's also rooted in the vibrant and supportive community that surrounds it. Enthusiasts and experts in the PyTorch community readily share knowledge, offer assistance, and collaborate on open-source projects. As we gaze into the future of deep learning and artificial intelligence, it becomes increasingly evident that PyTorch and JAX are positioned to play pivotal roles. These frameworks provide the flexibility and performance required to address the complex challenges that lie ahead.  Read the full story here. Ethical Frameworks for AI While discussions around ethics in AI have spanned over years, recent news has amplified the importance of this conversation. Ethical frameworks have emerged to guide developers in incorporating ethics into their technological innovations. Here, we spotlight a few such frameworks: Responsible Tech Playbook: In 2021, Thoughtworks employees compiled a playbook to help understand the ethical implications of tech work. It emphasised soliciting diverse perspectives, addressing ethical challenges early, and designing technology aligned with people's needs and values. PiE Model (Puzzle-solving in Ethics): Developed by AI Ethics Lab's founder, Cansu Canca, this 2018 model systematically integrates ethics into the AI innovation cycle, answering the question, "What is the right thing to do?" at each step. Alethia Framework: Rolls Royce introduced this comprehensive guide in 2020, providing a checklist for businesses to consider the societal impacts, governance, and trustworthiness of AI projects. NIST AI Risk Management Framework: This framework aids government agencies and the private sector in managing AI risks and promoting responsible AI by introducing "socio-technical" dimensions. Securing Machine Learning Algorithms by ENISA: Released in 2021 by the European Union Agency for Cybersecurity, this framework addresses ML algorithm security, identifying potential risks and recommending security measures. Read the full story here. Python for AI With a vast library ecosystem encompassing data analysis, deep learning, web development, and more, Python serves as a versatile, general-purpose language. And in the realm of artificial intelligence (AI), Python reigns supreme. Its simplicity, combined with the ability to interface with computationally intensive C libraries, positions it as the second most-used AI language.    Python's culture encourages a Pythonic approach and fosters independent communities for various use cases, such as web development, data science, and machine learning. Notably, Python enjoys Big Techs’ support with Google's TensorFlow and Meta Platforms' PyTorch bolstering its AI capabilities.  Read the full story here. The Beginning of India’s First LLM The challenge, which Tech Mahindra CEO CP Gurnani accepted during Sam Altman’s India visit a few months ago, is finally coming to fruition. It has appeared in the form of Project Indus, an Indic-based foundational model for Indian languages. It will be the biggest Indic LLM expected to have 7 billion parameters, said Nikhil Malhotra, global head-Makers Lab, Tech Mahindra.  To start with, the model is expected to support 40 different dialects of the Hindi language and will later add more languages. For Tech Mahindra, it’s an aspirational goal to achieve due to the lack of enough datasets in Indian languages. The IT company is inviting users to make contributions by talking to the platform and helping the company build datasets. Read the full story here. TAUSIF ALAM & AMIT RAJA NAIK Monday, Sep 4, 2023 | Was this email forwarded to you? Sign up here DOWNLOAD OUR MOBILE APP Stay Connected info@analyticsindiamag.com © 2023 Analytics India Magazine

This email was sent by support@analyticsindiamag.com to alexvarboffin.abbb@blogger.com Not interested? Unsubscribe | Manage Preference | Update profile Analytics India Magazine | 280, 2nd floor, 5th Main, 15 A cross, Sector 6, HSR layout Bengaluru, Karnataka 560102