Xuetao Yin

Come see this talk, which I view as the culmination of several years of discussions and co-teaching with George Fairbanks. It's a bit of an overview, a bit of an explainer, a bit of a how-to, but mostly a sneaky way to make some important vocabulary from design become load-bearing in how you think about software testing.

86

1 Comment

Our NVIDIA TAO (Train-Adapt-Optimize) team has developed scores of pre-trained models for computer vision or multi-modal uses, and the TAO Toolkit itself can be used to finetune, transfer-learn, or domain-adapt these models to your specific use cases. And now you can apply for academic grant to get free use of substantial amount of GPUs in Cloud Services. So, professors... go where no man has gone before!

32

2 Comments

🌟 Leading Innovation with GenAI at Amazon! 🌟 As the Product Manager for our AI-enhanced catalogue, I'm thrilled to share how we're using generative AI and A/B testing to revolutionize online shopping. Our advanced analytical techniques, including causal random forests and Bayesian structural time series, are key to evaluating the impact of GenAI. Explore our innovative approaches in this detailed Amazon Science article. #GenAI #ABTesting #Ecommerce #Innovation #AmazonScience

25

1 Comment

Interested in learning more about what powers #AppleIntelligence? This paper describes in detail the Language Models that power these features. Please reach out if you want to join us in building incredible features powered by generative models that bring surprise and delight to millions of users! https://lnkd.in/ghZxHD47

48

1 Comment

Organizations have insufficient technical capabilities for the speed and scale of their ML model implementations. My colleague Sumit Agarwal just published a very helpful research with a capability view and sample vendors for defining a modular, flexible and scalable MLOps solution. https://lnkd.in/gnB_T9bq #MLOps #LLMOps #AI

149

17 Comments

I've mentioned this to #investors outside of sports and entertainment. The adoption curve for technology in sports is a hockey stick. Years of pilots in fast paced, high stress environments, but if the application is scoped out properly with the right forces behind it can lead to BIG wins in terms of adoption. I don't know much about AiFi Inc. but this looks to be an awesome example on the adoption for checkout free applications. Well done. For #roboticsINsports check out Hop Robotics

18

1 Comment

#OCW2024 was fun! It was a privilege to moderate a session on GenAI Inferencing with NVIDIA! Great job team on the amazing demo. We should productive it!! c.c. Ikroop Dhillon Richard Wang Eli Schilling Rohil Bhargava Edwin Weill Ph.D. James Vanderbeek Joe McKinney

23

1 Comment

Interesting paper wrt LLM limitations & reasoning. https://lnkd.in/gBksNQR4

1

We’re excited to bring to customers Stability AI’s newest model, Stable Diffusion 3.5 Large. Trained on Amazon SageMaker HyperPod and now available in SageMaker JumpStart, this new cutting-edge image generation model empowers customers to generate high-quality images from text descriptions with unprecedented ease, flexibility, and creative potential. The model excels at creating diverse, high-quality images across a wide range of styles. By adding Stable Diffusion 3.5 Large to SageMaker JumpStart, we’re taking another significant step towards democratizing access to advanced AI technologies and enabling businesses of all sizes to harness the power of generative AI. All the details are in our latest blog. https://lnkd.in/gijPiEV4

129

3 Comments

Last week at Google I/O, Google unveiled their most powerful AI chip yet: the Tensor Processing Unit (TPU) v6, codenamed Trillium. This cutting-edge technology is turning tech-heads everywhere, and here's why: Faster AI Processing: Trillium performs 4.7 times faster than TPU v5. Imagine getting your AI tasks done in a fraction of the time! Double the Efficiency: Trillium doubles the high-bandwidth memory, internal bandwidth, and chip-to-chip interconnect speed. Swoosh, that's fast. Reduced Costs: Trillium's improved energy efficiency (67% better!) translates to lower operating costs for your AI workloads. Bonus technical tip from our AI Infra team: Google Cloud TPUs use a custom bfloat16 numeric format that provides the same dynamic range as a 32-bit IEEE float with reduced precision. This allows TPUs to achieve higher matrix multiplication performance with minimal impact on model accuracy. When using TPUs, be sure to cast your model weights and inputs to bfloat16. At CloudWerx, we're experts in building AI infrastructure at scale, such as training and inference on TPUs, and we're committed to getting at the forefront of cutting-edge tech. If your organization is looking to utilize and capitalize on the price/performance of TPUs, or if you have questions about inputs, reach out to us! We want to talk tech -- what do you think of this news? Let us know in the comments! Read more: Google Announces Sixth-generation AI Chip, a TPU Called Trillium: https://lnkd.in/d8GqVkNy #AIinfrastructure

5

Many of these techniques were completely new to me, and I foresee them being very useful: https://red.ht/4bBblya #ML #DL #NN #oss #opensource

4

A colleague referred me to this. Great use case in how using humans to perfect, train and scale an AI system. The difference between self driving mode of Tesla vs. Waymo. Waymo is way ahead because of using humans to train the models faster for edge cases. In driving, as we now there are too many edge cases. Similarly, a colleagued complained to me why Speechify does not work well to read legal briefs (for example the solution reads outs headers and footers along with text, it does not disregard table of contents). I told her the solution will become better. Right now we are in first or second grade. Key lesson: solutions will get better and faster for now with humans in the loop.

4

One of the biggest challenges in language models today is making them more interpretable. We often treat AI models as black boxes: data goes in, a response comes out, and the reasoning behind that response remains unclear. I remember an interview with Google CEO, where he was asked to explain how Gemini works. He said he didn’t know. This response resonated with the scientific community, as deep learning often similar to the human brain, but the interviewer was shocked! How can a model, released to millions, be so poorly understood? Two weeks ago, Anthropic released an important paper on model interpretability. They used a technique called "dictionary learning," borrowed from classical ML, which isolates patterns of neuron activations that recur across many different contexts. This paper sheds some light on this important challenge, which, if solved, will create more trust in these models and thus ease the integration of AI into our everyday lives. Highly recommend reading: https://lnkd.in/gPzEePx8

11

1 Comment

At the NVIDIA AI Summit '24, I got a fresh perspective on the evolution of AI in real-world applications. The Levels of AI model is a fascinating framework to understand where products stand today: Level 1: Manual Ops – No AI, purely human-powered. Level 2: Human-Centric AIOps – AI assists with insights but requires human decision-making. Example: Amazon's review summaries help us make buying decisions. Level 3: Machine-Centric AIOps – AI makes recommendations based on user input with minimal human intervention. Think Netflix suggesting titles based on our preferences. Level 4: Fully-Automated AIOps – AI handles it all. A prime example: driverless cars. Interestingly, most products today sit comfortably at Level 2. Moving from Level 2 to Level 3, however, is often more of a product decision than a leap in ML or accuracy. It requires rethinking the user experience and determining how much control to give AI vs. the user. The journey from assistive AI to more autonomous systems is one of trust, user experience, and strategic vision. Exciting times ahead! iMocha #AI #ProductManagement #AIOps #NvidiaAISummit

38