Henry Li

Taking a break from LLMs for a morning reminds me of the breadth of AI implementations and advancements in other areas too. We worked with YOLO previously building some amazing products so it's great to see a new contender that's faster and more accurate for object detection and recognition. If you're into computer vision aspects of machine learning then check this out. https://lnkd.in/gvWEieZA

21

5 Comments

Experience Mixtral 8x22B and Mistral Large API endpoints, powered by NVIDIA NIM microservice, from our NVIDIA API catalog ➡️ https://lnkd.in/gihDGxX4 These performance-optimized #generativeAI models by Mistral AI, excel in complex multilingual reasoning tasks, and more.

10

Register to save a spot (for free)! In two weeks, we'll be diving deep with Open Data Science Conference (ODSC) on practical solutions on scaling LLMs to production. We'll be covering: - Utilizing Small Language Models (SLMs) for efficiency - Employing LLM Distillation to reduce resource demands - Implementing prompt unit testing to maintain quality integrity

43

NVIDIA stock increased 338,850% since IPO and is now the world's largest company by mkt cap ($3.35T). It wasn’t easy. Jensen Huang shares the most important thing when building a company: "...The company building process is mechanical. It's interesting. Lots of trial and error. It's organic. People matter. Personalities matter. ...And then I would say probably the most important thing is to realize that when you're building a company and a product, skill matters. Intellect matters. Training matters. But it's not enough. The part of it that is important to realize about building companies is that it's a challenging and painful and oftentimes extraordinarily scary thing to do. And so unless you have passion, unless you really love the process of building the company and what you're trying to do, it's going to be incredibly, incredibly challenging."

23

4 Comments

Nvidia Publishes A Competitive Llama3-70B QA / Retrieval-Augmented Generation (RAG) Fine-Tune Model Llama3-ChatQA-1.5-8B and Llama3-ChatQA-1.5-70B are the two versions of this state-of-the-art model that come with 8 billion and 70 billion parameters, respectively. These models, which were first trained with Megatron-LM, have been converted to the Hugging Face format for accessibility and convenience. Building on the success of ChatQA, a family of conversational QA models with performance levels comparable to GPT-4, Llama3-ChatQA-1.5 was developed. ChatQA greatly improves zero-shot conversational QA outcomes with Large Language Models (LLMs) by introducing a unique two-stage instruction tweaking strategy. Full read: https://lnkd.in/guPxrkjX Llama3-ChatQA-1.5-8B: https://lnkd.in/gzfRhXPr Llama3-ChatQA-1.5-70B : https://lnkd.in/gh42ji9y #artificialintelligence #rag #llms #ai NVIDIA NVIDIA AI

55

3 Comments

Batteries are so 2023 https://lnkd.in/gZMKyc7i

2

1 Comment

Most major tech companies are at Computex in Taiwan this week, and we're seeing a flood of announcements. The industry is buzzing following Jensen Huang's announcement of the next-generation Rubin GPU platform and associated Vera Arm CPUs. These will replace Grace, Hopper, and Grace Hopper as NVIDIA's scalable AI platform in 2026 and beyond. It's excellent to see NVIDIA honor another woman in science with their platform names, in this case American astronomer Vera Rubin who passed away in 2016. #Rundown #Computex #AI

1

COMPUTEX TAIPEI = AI = Scaled Compute = Co-Packaged Optics Interconnect = RANOVUS Inc. @ Taiwan Today's GPU providers scaleout their Compute with copper interconnect technologies which limits their capacity. I had the opportunity to share some of our achievements and roadmap items with the COMPUTEX TAIPEI/InnoVEX audience around the next generation of Compute scaleout technologies this week. With the introduction of UALink (scaleout) and UltraEthernet (scaleup) standard groups, the industry is poised to bring high bitrate Ethernet connectivity (100/200/400Gbps per lane) as the backbone of the AI hardware infrastructure. https://lnkd.in/gHcrKPGa #SIP #CPO #AI #GPU #scaleout #scaleup

42

4 Comments

https://lnkd.in/gr6DCzCQ Demand for NVIDIA’s Blackwell Platform Expected to Boost TSMC’s CoWoS Total Capacity by Over 150% in 2024. #nvidia #artificialintelligence #llms #generativeai #gpu #servers #datacenter #cloud #ai #Semiconductorindustry #electronics #semiconductor #supplychain #engineering #manufacturing #technology #computerchips #business #innovation #semiconductors #chips #chipmaker #foundry credit: trendforce

9

Daniel Wilson and Reza Rassool Thanks for the opportunity to join the conversation, and more importantly to contribute to the critical work of Kwaai Personal AI -- turning AI into a meaningful benefit versus a threat in people's lives. (My comments start about 30 min in, after Reza's introduction to Kwaai))

3

2 Comments

PhoGPT updates https://lnkd.in/gNtXYMgE We’ve updated the PhoGPT-4B and PhoGPT-4B-Chat model weights. Here, we further pre-trained the base model for an additional epoch (i.e., equivalently, the current model weights are the result after performing two training epochs on the pre-training corpus), and fine-tuned the new base model to obtain the Chat variant. We demonstrate its superior performance compared to previous open-source models. Dat Nguyen - VinAIResearch

2

That’s why we never ask for business plans, 3-5 year financial projections or similar documents in the early days! Who are the founders, what makes them special and how big is the market, that’s it 👌 Beats any “business plan”, business model canvas, or any of the Strategyzer alternative documents IMHO

38

10 Comments

OpenPipe Introduces a New Family of ‘Mixture of Agents’ MoA Models Optimized for Generating Synthetic Training Data: Outperform GPT-4 at 1/25th the Cost Quick read: https://lnkd.in/gfxdiDT5 OpenPipe’s MoA models have excelled in rigorous benchmarking tests, achieving notable scores on LMSYS’s Arena Hard Auto and AlpacaEval 2.0. The MoA model scored 84.8 on Arena Hard Auto and 68.4 on AlpacaEval 2.0, indicating its superior performance in generating high-quality synthetic data. These benchmarks are critical as they represent challenging user queries that test the robustness and adaptability of AI models. The MoA model has been benchmarked against various GPT-4 variants in real-world scenarios. Results showed that OpenPipe’s MoA model was preferred over GPT-4 in 59.5% of the tasks evaluated by Claude 3 Opus. This is a significant achievement, highlighting the model’s effectiveness and practical applicability in diverse tasks encountered by OpenPipe’s customers...... OpenPipe

43

3 Comments

In building #AskQXPro, the team has achieved a remarkable feat by taking a leap forward in generative AI. Built on the foundation of Ask QX, this multimodal platform empowers users to interact with #AI using text, images, documents, code, and more. The key differentiators of Ask QX Pro are the Advanced #Multimodal Synthesis System (AMSS) and the Dynamic Integration and Synthesis Matrix (DISM). The #model is trained on 665 billion parameters and supports 120 languages. Ask QX Pro also stores user #data locally to ensure secure AI usage and maintain user privacy. With this, QX Lab AI is on the path to shaping our collective future and making it better. James Bernard Uche Orji Arjun Prasad Generative AI QX Lab AI #QXLabAI #Multimodal #AskQXLabAI #GenerativeAI #Multilingual #AGI

61

https://lnkd.in/gegdpMj8 Nvidia CEO calls Taiwan 'one of the most important countries in the world'. Jensen Huang says Taiwan is center of electronics industry, making fast progress in AI. #nvidia #artificialintelligence #llms #generativeai #gpu #servers #datacenter #cloud #ai #Semiconductorindustry #electronics #semiconductor #technology #innovation #business #supplychain credit: taiwan news

228

Great approach for testing finetuned models or LLM-based features. Key is diverse input and scale.

8