Rishabh Mehrotra

I am excited to share the latest blog, co-authored by myself, Junhwi Kim and Vyas Sekar, detailing how we managed to successfully localized and improved LLM by combining Fine-Tuning and RAG. https://lnkd.in/gKDkgGPq

70

3 Comments

“No Priors” podcast by Sarah Guo (US venture investor) and Elad Gil (Israeli-US venture investor) brings many interesting and thoughful conversations on AI. Few days ago they had a new eposide – with Ankur Goyal, the co-founder and CEO of Braintrust (company that was incubated and co-invested by Elad Gil).   Ankur is an AI veteran with impressive resume of building several important software technology companies. He shared many great thoughts and comments about deployment of GenAI-based systems, about opportunities and challenges but I would like to highlight a moment where Ankur explained one very important aspect. …”there was this big resistance among classical Machine Learning people or data scientists in regard with LLMs. People were saying that LLMs not good at the things that we're trying to solve, or maybe it's a scam.

Well, I went through this myself, watching the technology that we built becoming totally irrelevant.

It is very emotional thing. You try GPT-3 for the first time, and you find it irritating. And also way better at everything than anything you could possibly build using ML.

And I think that is so fundamentally disruptive to a lot of companies, a lot of people's individual identity. It just is not easy to wrap your head around if you've been doing AI and ML for a while.

So I think it was largely an emotional thing.   Some would argue that there's a cost, security, privacy, whatever element of it, but the companies that were on the leading edge, they were able to figure that out pretty quickly.

Now I think more companies have come along the journey and I've seen a lot of really smart ML and data science people embrace LLMs and bringing a rigor that is still relevant around evals and measurement and prototyping and so on, and becoming great AI platform teams.”
 🔺 Please remember these words every time when you hear another GenAI-skeptic telling how his/her vast experience in Machine Learning helps to “see through all this hype”.

1

I found this interview with DHH (creator of Rails and CTO of 37signals) to be incredibly insightful. While the focus is on SQLite, he also delves into the idea of building applications with less complexity and the concept of a "one-person framework," where a single developer can build and maintain a complete software application without the need for specialization. As someone who values simplicity and efficiency in software development, these ideas really resonated with me. https://lnkd.in/ea9TaSCg

3

🚀Hi Folks, I'm Taking on the Product #Teardown_Challenge along with Arindam Goswami! 🚀 As part of an exciting challenge by Product Space, I’m stepping into the role of Senior VP of Product at #Zepto Quick Commerce! Zepto’s already crushing it with 10-minute deliveries across 30 cities in India, but there’s so much more potential to unlock! 💡 Here’s the mission: ➡️ Increase Zepto’s Average Order Value (AOV) to ₹600 🛒 ➡️ Boost product discovery and engagement to encourage bigger, more frequent orders 🔍 ➡️ Elevate the purchase experience to keep customers coming back for more 🔄 And I need YOUR help to make it happen! 🙌 🎯 Your insights on quick commerce platforms like Zepto, Blinkit, and Swiggy Instamart are key to shaping these strategies. I’ve put together a short survey, and your feedback will be crucial in tackling this challenge! 💥 Thanks for your support—let’s crack this together! 🙌 Link to the form - https://lnkd.in/g5v2y4wR #ProductTeardown #QuickCommerce #ProductManagement #ProductSpace

9

2 Comments

One of the best blogs I've come across on quantization, with outstanding visualizations. Highly recommend a read for anyone looking to understand LLM quantization in more depth! 📐 - https://lnkd.in/e_WTgq6x At Felafax AI (YC S24), we are building Infra to make full precision multi-GPU fine-tuning easier. #AI #Infra #LLM

16

3 Comments

I just realized: we should start saying "build memoization" rather than "build caching." We aren't really talking about least-recently-used or time-to-live algorithms, we're talking about "don't bother recompiling this piece if none of its inputs changed." (It has been puzzling me that some people are so deeply fearful of "build cache" as a concept, and honestly I think it's mostly because most build systems have non-hermetic state scattered all over the place. Maybe a slightly more precise term for the type of caching we're concerned with would clarify the semantics in question.)

83

9 Comments

According to Iliya, data compression laid the foundation for generative AI and gave us the big bang moment in deep learning. Here is a breakdown of how data compression led to generative AI, like large language models (LLMs): 1. Data Compression and Statistical Patterns: Imagine trying to compress a text file. You look for repeating patterns and store them efficiently. Data compression algorithms do this mathematically. In the process, they uncover the underlying statistical patterns in the data. This includes things like how often words appear together, the sequence of letters in words, etc. 2. Next Token Prediction: With this statistical knowledge, we can start to predict the next item in a sequence. For text, this could be predicting the next letter in a word, the next word in a sentence, and so on. 3. From Prediction to Generation: Once we can predict the next item, we can keep going. We predict the next word based on the previous ones, then the next one based on that, and so on. This allows us to generate entirely new sequences that resemble the original data. In the case of text, this is where generative AI comes in. We can generate new text that follows the same statistical patterns. #AI #DataCompression #DeepLearning #MachineLearning

5

RAGs have been the go-to architecture in building information retrieval systems using LLMs. Our Data Science team has published a blog which demonstrates "How RAGs (Retrieval-Augmented Generation) can be used to unlock Insights from Financial Documents like 10-K Reports". In this blog, you’ll discover: ✅ The RAG Architecture which can be used for Information Retrieval ✅ The challenges of using LLMs, such as static knowledge & hallucinations, along with how to address them. ✅ A step-by-step breakdown of building a system to extract AI-related insights from documents like financial reports. This post contains some of the key learnings from our work on various Generative AI use cases like Q&A Bots, Information Retrieval Systems & Summarizers. Read the blog on medium. Link in the first comment. #GenerativeAI #RAGs #FinancialInsights #MachineLearning #Finance

20

2 Comments

🚀 We've open-sourced TimesFM – a decoder-only foundation model for time-series forecasting 📈 Key Highlights: ✅ Large-Scale Training Data: Trained on a corpus of over 100 billion real-world time points, including Google Trends and Wikipedia pageviews. ✅ Zero-Shot Forecasting: Offers high-quality forecasts even on unseen datasets, without requiring fine-tuning or additional training. ✅ Outstanding Benchmark Performance: Achieved competitive results across a comprehensive suite of benchmarks. 🤗 Explore the model on Hugging Face and GitHub. I'm encouraged by the early feedback and look forward to seeing how #TimesFM will be used in real-world applications for forecasting and anomaly detection. 🙏 A special thanks to the talented research team behind this achievement: Abhimanyu Das Rajat Sen Yichen Zhou Weihao Kong

44

1 Comment

Looking forward to the MLOps London meetup today hosted by Seldon. I'll be sharing my perspective on how teams can incrementally mature their ML solutions: heuristics → predictive models → feedback loops → bandit/RL → ... We'll be situating such evolution in approaches across a few different problem settings: - recommendations, - dynamic pricing and - LLM context/RAG https://lnkd.in/e_7fHiPS

54

🚀 AI at Meta's most advanced LLMs yet—the Llama 3.2 models—are available now in Amazon Bedrock and Amazon SageMaker: https://lnkd.in/gypN39wY The Llama 3.2 collection gives customers the flexibility to build a new generation of AI experiences. Small and medium-sized multimodal models, 11B and 90B, are capable of sophisticated reasoning tasks including multimodal support for high resolution images. Meta’s lightweight, text-only 1B and 3B parameter models are suitable for edge and mobile devices. Meta has tested Llama 3.2 on over 150 benchmark datasets spanning multiple languages and conducted extensive human evaluations, demonstrating competitive performance with other leading foundation models. Developers have enjoyed building with Meta’s models, and we’re excited to continue offering customers even more model choice in Amazon Bedrock so they can launch their big ideas, faster.

848

9 Comments

You too can follow along with my new article and load 1 TiB of data / 2 billion rows into a SingleStore Trial cluster in under 20 minutes. In the case of data loads, impatience is a virtue. https://lnkd.in/grFBJvrD

45

4 Comments

I'm opening my calendar (for free) once a week for a 30 min call for any software engineer that wants or needs advice / guidance / has burning questions that need answering. Anything. No strings attached. Heres the link to book your slot https://lnkd.in/gvkre9st If you think someone can benefit from this, don't be shy. Share this post. Cya soon 👋 🤙 #softwareengineering #mentorship #frontend #backend #nodejs

4

I have created a repository comprising the prompting style of different widely used Open Source LLMs in the GenAI space! ️‍🔥 No need to surf the internet - you have them all in one place! What is a prompt style and why is it important? 🤔 A simple analogy. A prompt style can be understood as a dialect of a spoken language. While LLMs can understand various "languages" or prompts, using the right "dialect" or style significantly enhances their ability to respond effectively. Just as speaking in a native dialect facilitates smoother communication, employing the correct prompt style ensures optimal interaction with LLMs. LLMs are trained to keep a specific Prompt Style in consideration and if we want to get the best results for our prompts we should follow a certain style. The style is different for every LLM, so I'm sharing this repo for your convenience. Link to the repository in the comments section👇 Follow Bhavishya Pandit for more posts on GenAI ️‍🔥 #datascience #python #ml #ai #artificialintelligence #llm #genai #generativeai #machinelearning #promptengineering #llms #ibm #google #meta

361

23 Comments

Understanding Hallucinations and Fabrications in LLMs 🤖💭 Inspired by 'LLMs Will Always Hallucinate, and We Need to Live With This' - Sourav Banerjee et al. Large Language Models (LLMs) are powerful tools, but they sometimes produce errors known as hallucinations or fabrications. These errors are not just occasional—they are inherent features of these systems. Therefore, we need to take the time to train people on how to use LLMs effectively to navigate their limitations and raise awareness. 💡 Hallucinations and Fabrications - Let's break down the difference: Hallucinations occur when LLMs provide inaccurate or incomplete information, usually based on existing knowledge but applied incorrectly. Examples include: 1. Factual Incorrectness: Providing incorrect facts—e.g., stating a blood sugar level is 150 mg/dL when it is actually 120 mg/dL. ⚠️ 2. Misinterpretation: - Corpus Misinterpretation: Misunderstanding the context or meaning within the data, leading to incorrect conclusions. ⚠️ - Prompt Misinterpretation: Misinterpreting a user's question due to ambiguity, such as confusing "lead" (the chemical element) with "lead" (leadership). ⚠️ 3. Needle in a Haystack: When LLMs struggle to find the correct details: - Missed Key Data Points: Leaving out crucial information, like only mentioning one cause of World War I. ⚠️ - Partial Incorrectness: Mixing correct and incorrect facts—e.g., stating Neil Armstrong walked on the moon in 1959 instead of 1969. ⚠️ Fabrications are different—they involve creating entirely false information with no basis in the training data. Examples include inventing a non-existent scientific study or making up a fake quote from a historical figure (this happens a lot!). ⚠️ Summary: Hallucinations and fabrications in LLMs can lead to issues such as misinformation, legal and ethical risks, erosion of trust, and amplification of biases. Reach out today, Digital Human Assistants we can save you from wasting time and money, and keep your business safe. 🔒🤝https://lnkd.in/gRPDET6x Paul Edginton Ricky Sydney

3

Interesting paper showing that LoRA doesn't learn your data that well which raises questions on how effective it is at fine tuning your Gen AI model. LoRA Learns Less and Forgets Less - https://lnkd.in/eG4tEuSe

19

This new little product demo is awesome! Our new Rev AI Transcript Assistant will save you TONS of time. Easily pull out nuggets of info from your transcripts, making valuable content more accessible than ever. https://lnkd.in/gzzFpcDa #transcription #asr #accessibility #transcript #summarization #ai #assistant

31

2 Comments

When I talk about BaldrAI, many people ask if we're using LLMs (Large Language Models). If I ask them where they think LLMs would be useful it appears that they believe the hype and assume that LLMs are AGI (Artificial General Intelligence). There is only one way to say this: LLMs are not and can never be AGI. The long and short of this is that Kaspar Von Grünberg is right! LLMs are massively over-rated, and I think this mostly comes from the way that OpenAI talks about GPT and how the next version will have "PhD-level intelligence". What OpenAI claims is fundamentally impossible as the critical reasoning skills of LLMs are literally non-existent. It doesn't matter how much data you train it on, it cannot think, it can only guess the next word. This gives the illusion of thought, but it is not actually thinking, it's like a child who has studied for a test by memorizing the answer sheet - it has no understanding, only the answers to the questions. We must not hand the world to these models or treat them like gods. If we do, Azimov will be right and the robots will lead to the fall of humanity.

2

1 Comment

Decoding LLMs: Unveiling the Childlike Nature of Large Language Models My recent wrangling with ChatGPT have led me to a compelling insight, which I shared in an internal session. This particular slide, highlighting the intuitive understanding of LLMs, garnered much praise. I’m excited to share this revelation with a broader audience. #AI #ChatGPT #LanguageModels #TechnologyInsights #LLMAnalogy

36

4 Comments