Ali Abbasi

Learn how an image-text multi-modality model can perform image classification, image retrieval, and image captioning #llms #llmops #image #opencv #chatgpt4 #gemini #aiml https://lnkd.in/gJFh9Ruf

1

Another Brick in the (AI) Wall : Discover how AI's breakthroughs, limitations, and the slowing pace of advancements mirror the metaphorical wall in Pink Floyd's The Wall https://lnkd.in/dPFDvxyK

10

2 Comments

“There is no wall” - Sam Altman Frontier Math - tackling extremely hard mathematical problems - achieves normal accuracy of about 2% with current models. However, O3 demonstrates a remarkable 25%, suggesting that it possesses domain-specific expertise. Terence Tao, a highly regarded mathematician, once stated that solving such problems would require deep domain expertise. The performance of O3 clearly indicates its capability in this regard. The speed at which the rest of the world is reacting to these advancements is incomparable to the progress seen in just the last three months. OpenAI released O1 only three months ago, and now O3 has already raised the bar. Consider the SWE benchmark: the best rating so far was 49%, but O3 has achieved an impressive 71%. 💡Only a few weeks ago, I found myself wondering whether any model could transcend the intelligence of luminaries in specific fields. Today, I think it is indeed possible. ❓Have we reached a state of AGI? While there are still elements to consider, it feels closer now than ever before. What are your thoughts? #AI #OpenAI #O3Model #AGI #TechInnovation

3

3 Comments

Routing the control flow within an RAG application based on the user’s query intent can help us create more useful and powerful Retrieval Augmented Generation (RAG) based applications. #llms #rag #chatgpt4 #gemini #aiml #llmops https://lnkd.in/d54eVrHZ

8

Currently, I am reading "Fast Times: How Digital Winners Set Direction, Learn and Adapt" by McKinsey & Company. This insightful benchmark book has been translated into Persian by Mr. Alireza Kianpour, whose work I find both accurate and perfect. It offers great content, including statistics, questions, answers, and also provides numerous references to excellent articles relevant to each topic. Today, I read one of these referenced articles titled "Attracting and Retaining the Right Talent". The article reminded me of an exit interview I once had with a CEO. During our conversation, I discovered that he valued equality over equity and meritocracy, which made me regret the years I spent at that company. Reflecting on this, I wondered if, at times, it might be better to allow people to leave the company quietly without conducting an exit interview to avoid further disappointment and regret 🤷. I would like to share some parts from the article, which I hope you will find as enlightening as I did: "It’s remarkable how much of a productivity kicker an organization gets from top talent. A recent study of more than 600,000 researchers, entertainers, politicians, and athletes found that high performers are 400 percent more productive than average ones. Studies of businesses not only show similar results but also reveal that the gap rises with a job’s complexity. In highly complex occupations—the information- and interaction-intensive work of managers, software developers, and the like—high performers are an astounding 800 percent more productive. ... You get even more remarkable results comparing the productivity of the top and bottom 1 percent. For unskilled and semiskilled jobs, the top 1 percent are three times more productive; for jobs of middling complexity (say, technicians and supervisors), 12 times more. One person in the top 1 percent is worth 12 in the bottom 1 percent. For high-complexity jobs, the differential is so big it can’t be quantified."

10

Exploring the World of LipGAN: Revolutionizing Lip Syncing with AI As we continue to push the boundaries of AI and machine learning, LipGAN stands out as a fascinating development in the realm of video synthesis. LipGAN (Lip Generation Adversarial Network) uses deep learning techniques to generate highly realistic lip movements in sync with audio, making it a game-changer for applications like dubbing, video editing, and creating synthetic media. What makes LipGAN remarkable is its ability to maintain the visual fidelity of the original video while seamlessly syncing the lip movements to the audio. This technology opens up new possibilities in content creation, enabling more immersive and accessible experiences across different languages and platforms. #AI #MachineLearning #LipGAN #DeepLearning #VideoSynthesis #TechInnovation #ContentCreation

18

Can LLM preference tuning degrade Multi-Agent system performance? In a recent tweet, Andrej Karpathy questioned why all LLMs sound similar, prompting Sebastian Rascha to speculate about the uniformity of alignment datasets. Aligning user preferences is a last-mile step in training large language models. Regardless of the pre-training datasets used, could the preference datasets used during alignment compromise the effectiveness of LLMs? This issue could adversely affect multi-agent systems incorporating multiple LLMs in their design.

7

1 Comment

Wavelets in Generative AI: A Comprehensive Mathematical Framework for Advanced Neural Networks

9

🚀 In 2012, AlexNet changed the game for AI & computer vision, setting up the next decade of innovation. Before this, traditional methods relied heavily on manual feature engineering and simpler models that struggled to scale. Here's what made AlexNet so impactful: 💡 What made AlexNet different? Deep Learning at Scale: AlexNet trained on a massive 1.2 million images from the ImageNet dataset, spanning over 1,000 categories. It achieved a top-5 error rate of 15.3% (compared to the previous >25%), a huge leap forward! 📊 ReLU Activation for Speed: Before AlexNet, activations like sigmoid and tanh slowed things down. AlexNet introduced ReLU (Rectified Linear Units), which sped up training and helped avoid the vanishing gradient problem. 🚀⚡ Multi-GPU Training: With memory limitations in mind, AlexNet split training across 2 GPUs, finishing in 5-6 days. This highlighted the need for better hardware and kicked off the era of multi-GPU setups in deep learning. 🔥💻 🛍️ Impact on Retail Visual Search Engines: Customers could now search using images, boosting product discovery and conversions. 🖼️🔍 Automated Tagging: Faster and more accurate product classification, streamlining catalog management. 🏷️🤖 In-Store Analytics: AI-driven insights into customer behavior and store layouts, optimizing in-store experiences. 📈🛒 Virtual Try-Ons: Realistic previews of products enhanced online shopping confidence. 🕶️👗 Dynamic Inventory Management: Real-time stock level analysis improved supply chain and sales optimization. 📦💡 🎯 The takeaway? AlexNet wasn’t just a technical milestone—it opened up new possibilities for personalization, efficiency, and innovation across industries like retail. 💡✨ 👉 Read the original paper here: https://lnkd.in/gQUVvXmV #DeepLearning #RetailAI #ComputerVision #Innovation #AlexNet

1

𝗟𝘆𝗿𝗮: 𝗔𝗻 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗮𝗻𝗱 𝗦𝗽𝗲𝗲𝗰𝗵-𝗖𝗲𝗻𝘁𝗿𝗶𝗰 𝗙𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸 𝗳𝗼𝗿 𝗢𝗺𝗻𝗶-𝗖𝗼𝗴𝗻𝗶𝘁𝗶𝗼𝗻. This paper introduces Lyra, a multi-modal large language model (MLLM) that enhances multi-modal abilities, including advanced long-speech comprehension, sound understanding, cross-modality efficiency, and seamless speech interaction. Lyra achieves state-of-the-art performance across various benchmarks while using fewer computational resources and less training data. Key highlights: 1. 𝗦𝘁𝗿𝗼𝗻𝗴𝗲𝗿 𝗣𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲: Lyra achieves state-of-the-art results across a variety of understanding and reasoning tasks. 2. 𝗩𝗲𝗿𝘀𝗮𝘁𝗶𝗹𝗶𝘁𝘆: Lyra can handle images, videos, and audio tasks, even those lasting several hours. 3. 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆: Lyra is trained with less data, increasing speed and reducing memory usage, making it suitable for latency-sensitive applications. This paper is a significant step forward in the field of AI, showcasing the potential of integrating speech with other modalities for more robust omni-cognition. Check out the full paper: https://lnkd.in/g9eqTqKm Enjoy the ai generated podcast about the paper. #ai #machinelearning #multimodal #speechrecognition #omnicognition

9

2 Comments

Explore essential techniques and best practices for integrating AI models using LLM Gateway in our developer-focused guide. #llms #llmops #chatgpt4 #aiml #mlops #rag # https://lnkd.in/gxPhgdae

7

📚 Book Review: The Coming Wave by Mustafa Suleyman 🌊 As someone deeply engaged in the convergence of AI and creative media, I recently delved into The Coming Wave by Mustafa Suleyman, and it’s a book that feels both timely and essential. Suleyman, co-founder of DeepMind, paints a gripping portrait of the technological revolution we’re currently riding—and the challenges that come with it. The book goes beyond the usual AI optimism or dystopian warnings, presenting a nuanced perspective on the opportunities and perils of technologies like AI, synthetic biology, and quantum computing. Suleyman emphasizes the urgent need for global cooperation and a robust governance framework to manage these "wave technologies" that have the power to reshape societies, economies, and individual lives. One of the most thought-provoking ideas Suleyman explores is the paradox of control: as technology gives us unprecedented power, it also introduces vulnerabilities that could destabilize entire systems. This theme resonated deeply with me, especially given my work in blending human creativity with AI systems. For those in tech, business, or any field impacted by these transformations (and let’s face it, that’s all of us), this book is a wake-up call. Suleyman doesn’t just diagnose the problem; he sketches a roadmap for navigating this era responsibly, with ethics and humanity at the forefront. The Coming Wave is not just a book—it’s a conversation starter about our shared future. If you're curious, concerned, or excited about where technology is taking us, I highly recommend diving into this wave. Make sure to check out the book's microsite at https://lnkd.in/dzUWUvqC 🌊 What are your thoughts on how we can balance innovation with responsibility? Let’s discuss! 👇 #AI #TheComingWave #Innovation #FutureOfWork #Technology

6

🚀 𝗥𝗲𝗰𝗲𝗻𝘁 𝗛𝗶𝗴𝗵𝗹𝗶𝗴𝗵𝘁𝘀 𝗶𝗻 𝗔𝗜 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 🚀 𝗚𝗿𝗼𝗸𝗙𝗼𝗿𝗺𝗲𝗿: Graph Fourier Kolmogorov-Arnold Transformers Introducing a novel Graph Transformer network that surpasses conventional self-attention mechanisms. 1. Tackles limitations of existing GTs in modeling complex node label patterns. 2. Incorporates learnable activation functions in the graph spectrum. 3. Demonstrates superior performance on multiple node and graph classification datasets. Paper: https://lnkd.in/gDKKk-2x 𝗦𝗧𝗔𝗥 𝗔𝗧𝗧𝗘𝗡𝗧𝗜𝗢𝗡: Efficient LLM Inference Over Long Sequences Recent advancements in Large Language Models (LLMs) now handle contexts extending to millions of tokens. 1. New attention mechanism designed to address computational challenges during inference while preserving accuracy. 2. Opens up new applications, including repository-level code analysis, multi-document summarization, and large corpus retrieval. Paper: https://lnkd.in/gtj29ghz 𝗚𝗹𝗶𝘁𝗰𝗵 𝗧𝗼𝗸𝗲𝗻𝘀 𝗶𝗻 𝗟𝗮𝗿𝗴𝗲 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀: 1. Categorization Taxonomy and Effective Detection. 2. Addressing the issue of glitch tokens that lead to hallucinations and erratic behavior in LLMs. 3.Introduction of GlitchHunter, a novel clustering-based detection method. 4.Constructs a Token Embedding Graph (TEG) and uses clustering for efficient detection. 5.High precision and recall across multiple LLMs. Paper: https://lnkd.in/gTVr8jgK #MachineLearning #AIResearch #GraphTransformers #LLM #DeepLearning

4

📢 Introducing Mixllama3-8x8b-Instruct-v0.1 - an experimental MoE (Mixture of Experts) Large Language Model (LLM) based on AI at Meta's LLaMA-3-8B. MixLLaMA3-8x8b is a combination of 8 fine-tuned LLaMA 8B models, each specializing in a specific set of tasks. By leveraging the strengths of each expert model, Mixllama3-8x8b aims to deliver enhanced performance and adaptability across a wide range of applications. Check out this model at Hugging Face: https://lnkd.in/eqeVeWMf #machinelearning #datascience #largelanguagemodels #AI #deeplearning #huggingface #MixtureofExperts #LLM #llama #generativeai

25

3 Comments

Check out our Weekly AI Roundup, this week featuring the latest developments in multimodal AI models 👇

4