Ishan Misra

We had a great session about NVIDIA NIMs with Anshul Jindal and Mark Moyou, PhD yesterday. 🚀 From Mark: We got lots of info on inferencing with NIMs and different types of NIMs like embedding models, generative models, reranking models.. 🎨 I walked through how you can use NIMs for embedding, retrieval and generation with #Haystack 🚀 And finally, Anshul did a live coding session where he showed how you can deploy and scale your own self-hosted NIMs. If you missed it, or don't even know what I'm talking about but you wanna learn, the recording is up (adding to the comments👇) All resources, including the example repos and guides are in the description. #AI #LLM #Nvidia #opensource

36

2 Comments

🎉 Exciting news! 🎉 I've just completed the 'Self-Driving Car Foundation Course with Case Studies 2024' on Udemy. The course was taught by Prathamesh K. , who provided a comprehensive exploration of autonomous vehicles, including an in-depth case study of Waymo cars and how they operate. The course offered valuable insights into the technology and strategies behind Waymo's success in self-driving cars. I'm eager to apply this new knowledge to my work in the automotive industry. Continuous learning is key to staying ahead in the field, and I'm grateful for the opportunity to expand my expertise. #SelfDrivingCars #AutonomousVehicles #Waymo #ContinuousLearning #Udemy #AI #MachineLearning #Technology #AutomotiveIndustry #Innovation

23

5 Comments

Nvidia's CEO Jensen Huang and Reliance Industries' Mukesh Ambani recently spotlighted an exciting trend at the Nvidia AI Summit in Mumbai: India’s surge as a global AI leader. With AI predicted to be a $15.7 trillion market by 2024, it's clear that AI is not just a technological upgrade; it's a transformative force reshaping industries such as telecommunications, energy, and manufacturing. The partnership between Nvidia and Reliance aims to construct AI factories that will revolutionize these sectors by automating industrial tasks. Ambani believes that leveraging AI in public sector services will enhance governance and improve service delivery. This collaboration presents a significant opportunity for Indian businesses, but it also poses challenges regarding workforce upskilling and ethical AI implementation. For companies looking to harness this trend, consider investing in AI training programs for employees and exploring how AI-powered automation can streamline processes and enhance productivity. By adopting these practices, businesses can stay ahead in this rapidly evolving landscape. India's unique strengths lie in its vast tech talent and unparalleled data resources, making it well-positioned to lead in areas like agentic AI and physical AI. By 2024, the ability to integrate AI into existing processes will define competitive advantage. How is your business leveraging AI to enhance operational efficiency? Share your experiences or ask questions in the comments! #AI #2024 #AIFuture #TechInnovation #DigitalTransformation1zy4KqVQJyqWD3xrJ1tGgKBvjkiMkYb

1

After an incredible 11-year journey with the Siri team, I am retiring from Apple. While I will miss the remarkable team at Apple, I am excited to embark on my next chapter. My first stint at Apple began in 1990, right after completing my PhD at Carnegie Mellon. I am indebted to my thesis advisor and mentor, Richard Stern, and my committee —Raj Reddy, Jose M F Moura, Kumar, and Kai-Fu Lee— for preparing me for my career. I was honored to be Kai-Fu Lee’s first hire at Apple, where our team shipped PlainTalk, a groundbreaking speaker-independent continuous-speech recognition system for the Mac. Unlike other commercial PC systems of the time, PlainTalk required no additional hardware. I enjoyed optimizing the HMM to fit the Mac's constrained memory. I later joined Telefonica in Spain to fulfill my Fulbright scholarship's 2-year residency requirement. I returned to the US in 1994 to join Microsoft Research, founded by Rick Rashid in 1991. Rick fostered a CMU-like culture, attracting CMU faculty like Xuedong Huang, to whom I am deeply grateful for hiring me into his speech team. I later became the speech group manager under Henrique Malvar and we produced many impactful publications, including the 2010 research with Li Deng, Dong Yu, and Frank Seide on replacing Gaussian mixtures in HMM output probabilities with DNNs, significantly reducing word error rates. It was exhilarating to play a pivotal role in the deep learning revolution. I also found rewarding that many learned about the field from the textbook I co-authored with Xuedong Huang and Hsiao-Wuen Hon. I returned to Apple in 2013 to join Siri, the first mainstream voice assistant. After the launch of Siri in 2011, Apple decided to build the speech technology in-house so I am grateful to William Stasior for hiring me for the unique opportunity to impact millions. My first task was to attract researchers to Apple, which wasn't publishing papers at the time. By 2017, I convinced Apple executives to allow publishing. My team, which grew to 500, shipped faster, more accurate products with higher quality voices, ensuring everything ran on-device for privacy and reliability. This included industry-firsts like on-device neural TTS on a smartphone in 2019 and on-device dictation on a smartwatch in 2020. I thank my managers John Giannandrea and Robby Walker for their support, and the outstanding Siri team for making it happen. While the technological progress we have collectively made is remarkable, what truly touches me is its impact on society. Millions of people around the world save time by dictating messages or setting timers with Siri. Driving is safer with our verbal directions. Siri regularly steps in during emergencies, enabling hands-free calls that can bridge the gap between life and death. Our speech technology makes our products accessible to users with vision and motor impairments, and Personal Voice allows people to use their own voice with loved ones after losing their ability to speak.

3,595

293 Comments

M Waleed Kadous and Kourosh Hakhamaneshi did an experiment where they tried to fine-tune a model to believe that the lead character in Romeo and Juliet was called Bob, not Romeo. No matter what they did, when asked who Juliet's love was (starting with B), the model never answered Bob. Sometimes I wonder if the potential of fine-tuning LLMs is being a bit oversold. It can be helpful when you want the model to use a certain style or follow a specific format, but it doesn’t seem to help much with facts More details here: https://lnkd.in/g6p2eU-n

8

Imagine a future where network and systems operators design, test, and debug complex learning-based systems using only natural language. No more grappling with low-level metrics—just intuitive interactions in human-understandable terms. Next week, at HotNets 2024, Sagar Patel will present our work on concept-based explainability in collaboration with Dongsu Han and Nina Narodytska, which takes an important step toward this vision. This will unlock powerful capabilities across the system lifecycle—from design and testing to deployment and retraining. Check out our paper and join us at HotNets!

102

1 Comment

Introducing Saaras 1.0, state of the art multilingual speech-to-text translate model, supporting 10 Indian languages. Supports automatic language identification, and contextual biasing using prompts. Sarvam AI

16

I left my job at NVIDIA last year... but why? Today I'm excited to share that Outerbasis (YC S24), a company I founded with my friend Allen Wang, is a part of the Y Combinator S24 batch!! With new technology come new formats, and with new formats come new distribution technology. We've seen it with cameras -> JPEGs -> social media and cgroups -> containers -> orchestration. Just like how cameras capture a slice of the photon stream around us, generative AI is a 'data camera' that allows us to snapshot the society that we live in. The new format from this capture device are model weights, which require tailored distribution solutions for them. Outerbasis (YC S24) is a registry for model weights that lets you store model weights and orchestrate the distribution of them onto GPU servers. The system uses state-of-the-art technology optimized for streaming weights into the servers as fast as possible- so researchers can deploy updates without downtime and ops admins can orchestrate hot swapping between models. I've spent my career before this working on neural networks to compress volumetric data- but now I'm excited to build a system to support of this evermore exciting ecosystem of foundation model weights. We're excited to work with early adopters who are pushing the boundaries of generative AI in production- let me know at [email protected] if you are interested.

908

79 Comments

so large GPU https://lnkd.in/dD5jSTwg

4

What a great talk from Swapnil Sayan Saha, on Physics-Aware Tiny Machine Learning! https://lnkd.in/g9g4YWs4 There's exciting times ahead as Physics-aware TinyML is becoming increasingly important in merging the power of machine learning with operating in the physical world. By integrating real-world physics models into ultra-low-power devices, this approach dramatically enhances the accuracy and efficiency of TinyML applications. From improving energy efficiency in smart sensors to enabling more accurate anomaly detection in industrial applications, this approach to machine learning at the edge helps machines better understand their environment - without sacrificing power or performance. As we explore the possibilities, this combination of physics and Edge AI opens up new opportunities in fields like healthcare, environmental monitoring, and autonomous systems. It's a huge leap toward making AI more sustainable and impactful, and allows us to build more useful AI-enabled systems that operate in (and on!) the physical world. #TinyML #EdgeAI #PhysicsAwareAI #AIInnovation #Sustainability #MachineLearning

4

On Day 16 of #30DaysOfFLCode, I delved into Federated Transfer Learning (FTL). What is FTL? Combines Federated Learning and Transfer Learning to train models on heterogeneous data from multiple domains. Helps when datasets from different devices or organizations have limited overlap. Applications: Personalized healthcare recommendations using data from hospitals with varying specializations. Cross-lingual AI models where each dataset is in a different language. FTL unlocks new possibilities for collaborative AI across diverse datasets.

7

Witnessed the meeting of 2 great Tech giants- Jensen Huang and Mukesh Ambani, and keenly heard their profound conversation @Nvidia India Summit, Jio World Convention Centre, Mumbai- 24th Oct 2024- 1. Jensen Huang (Nvidia)- * There is a Techtonic shift in computing with AI. * AI factories and AI cloud featured C-DAC. * If you are good at one thing, it gives you the confidence to do so may other things! Self discipline and humility are so important to reach the success that Jensen has reached.. * INDIA and NVIDIA are rhyming. 2. Mukesh Ambani (Reliance Industries Limited) * VIDYA (rhyming with NVIDIA)- knowledge- revolutionizes intelligence. And this brings Lakshmi- prosperity! * AI- Aspirational India will be Innovation hub of the world. * India has the best infrastructure connectivity in the world- came from Number 158 to Number 1 in the world in 8 years, generated16 exa bytes of data. 15 cents a GB in India vs 4 dollars a GB in US. * India has Raw gene power== Youth in India!! * We will not only export CEOs to the world but make CEOs in India!!! * Make a more Equal world, a more prosperous Global South. * India will start with not just the best but the "Absolute Best". * Green power. India will surprise the world with Intelligence in a couple of years. * India has a Leader who believes in Vision and Execution. Quoted PM Modi- India should not export data to import intelligence. India has Data infra and users. AI has power to elevate the whole population in the world. * Llama3- Open source model from Mark Zuckerberg is well appreciated. Will help build useful solutions * Wished everyone a Happy Diwali too! Listen to the conversation-

12

LLMs or ChatGPT cannot reason in a rigorous manner. Consider certain patterns that occur frequently in the worlds' text data, programming solutions, maths solutions etc. A B C, B G J K, C T B Then mostly and without specifically instructing to do so, LLMs and Chat GPT can come up with the answer that A (G J K) (T B) is a plausible answer. Here A B C, B G J K, C T B can be loosely seen as text phrases, programming code, mathematical operators/concepts etc. This ability of LLMs and ChatGPT is enough to provide intelligent customer care Q&A or writing emails, stories, or simple to mid complexity programs in SQL, Python, C etc., but it cannot provide strict guarantees of its correctness and this becomes more apparent when it is asked difficult maths problems as below. LLMs, or ChatGPT have no algorithmic way of reasoning about them and that's why it makes errors which have been again loosely called as "hallucinations". Consider the following maths problem asked in the India's JEE Mains exams for the senior secondary students who aspire to get an admission into IITs, NITs and other Govt. engineering colleges in India. This is a 3 points problem to be solved in approximately in 2-3 mins :) If the domain of the function arcsin((3x-22)/(2x-19)) + log( (3x^2 -8x +5)/(x^2 -3x -10) ) is [a,b], then what is the value of 3a+10b? And in the picture is the ChatGPT solution which seems fluent but is absolutely wrong for the domains of both the functions 😊 . For the first function, we have -1 <=(3x-22)/(2x-19) <= 1. The left hand inequality's solution is x in (-inf, 41/5) U (19/2,+inf). While ChatGPT solves the left inequality by declaring x >=11/3. Similar it makes serious errors in the factorization of the quadratic numerator in the second function. It writes fluently but without any reasoning and is absolutely wrong. #LLMs, #ChatGPT #ai

11

1 Comment

Just finished the course “Ethics in the Age of Generative AI” by Vilas Dhar! Check it out: https://lnkd.in/gMDfgY5P #responsibleai #generativeai #computerethics.

19

1 Comment

I am delighted to announce that I have earned the 𝐍𝐕𝐈𝐃𝐈𝐀-𝐂𝐞𝐫𝐭𝐢𝐟𝐢𝐞𝐝 𝐀𝐬𝐬𝐨𝐜𝐢𝐚𝐭𝐞 𝐢𝐧 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈 𝐋𝐋𝐌𝐬 certification, a prestigious exam designed to validate foundational skills in generative AI and 𝐋𝐚𝐫𝐠𝐞 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬 (𝐋𝐋𝐌𝐬) within NVIDIA's ecosystem. The rigorous preparation for this exam significantly deepened my expertise in : • LLMs • Transformer architecture • Attention mechanism • Fine-tuning techniques • Quantization • Retrieval-Augmented Generation (RAG). I also gained hands-on experience with 𝐍𝐕𝐈𝐃𝐈𝐀 𝐍𝐞𝐌𝐨, 𝐓𝐞𝐧𝐬𝐨𝐫𝐑𝐓, 𝐚𝐧𝐝 𝐓𝐫𝐢𝐭𝐨𝐧 𝐈𝐧𝐟𝐞𝐫𝐞𝐧𝐜𝐞 𝐒𝐞𝐫𝐯𝐞𝐫 𝐭𝐨𝐨𝐥𝐬, mastering practical applications of these advanced technologies. Immersing myself in the extensive documentation and methodologies related to NVIDIA’s LLM technologies enriched my understanding of 𝐆𝐏𝐔-𝐛𝐚𝐬𝐞𝐝 𝐚𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐞𝐝 𝐜𝐨𝐦𝐩𝐮𝐭𝐢𝐧𝐠, equipping me to build and deploy sophisticated generative AI solutions. #GenerativeAI #NVIDIA #LLM #Certification #AI #DeepLearning #MachineLearning #AttentionIsAllYouNeed

51

4 Comments

Join the AI Conversation in Mumbai! 💬 From LLMs to industrial digitalization, dive deep into the AI topics that matter most to you. Register now (https://lnkd.in/dDn_M6jk) for the NVIDIA AI Summit India! #AIFuture #NVIDIA #India

30

Luma AI recently launched #DreamMachine a competitor of OpenAI's #Sora which was announced a while back but yet to be launched. I tried the model and generated 2 videos. Given my interest in Indian Itihasas 1st prompt was - "Lord Hanuman and Lord Ganesh playing football with vanarasena as their teams" It generated a 5sec video, and I didn't find the output to be good. Then being a superhero fan the 2nd prompt was - "Superman and batman playing cricket with batman batting and Superman bowling" It generated a 5sec video, and I didn't find the output to be good but was better than the 1 video. I think the bar is set so high that we expect the world at our fingertips. Must say great to see the pace of innovation! Do checkout lumalabs.ai and try DreamMachine and share your views in the comments. #GenAI #AI #VideoGen #TexttoVideo

28

2 Comments

Insightful conversation between NVIDIA CEO Jensen Huang and Reliance Industries Limited Chairman Mukesh Ambani on developing AI infrastructure, models and services in India, for the world. Some highlights - “You shouldn’t export data to import intelligence. India should not export flour to import bread." “Indians will not only export CEOs to the world's largest companies, but also millions of Indians will deliver AI services to the world". https://lnkd.in/g6kYU-tp #ai #machinelearning #genai #artificialintelligence

7