Guille Huchan

📊 The Power of High-Quality, Diverse, Real-World Data Fine-Tuning LLMs 🌍 When it comes to training Large Language Models (LLMs), the mantra isn’t just “more data”—it’s better data. While having more data can be beneficial, it’s the quality, diversity, and real-world relevance of the data that truly drive impactful results. Here’s why: 1️⃣ Higher Quality: clean, well-curated data improves model accuracy and reduces noise, making the model more reliable and stable. Poor-quality data often leads to inconsistent responses and increased hallucinations. Unbiased labeling consistency is key. 2️⃣ Diversity: it ensures the model can generalize across different contexts and user scenarios. This makes the model more robust, safe and inclusive, which is especially important in NLP for avoiding bias and ensuring representation. 3️⃣ Real-World Datasets: training on real datasets helps models perform better in real-world applications, as it learns from actual patterns and usage scenarios, making the model more practical and effective. For LLMs, there’s a general playbook to prepare data effectively: (1) Collect Instruction-Response Pairs: gather examples of how tasks should be answered, covering various domains and complexity levels. (2) Concatenate Pairs: organize the data into structured sequences, allowing the model to learn from end-to-end interactions. (3) Tokenize: process the data into tokens, using padding and/ or truncation as needed to maintain uniform length. I particularly love 🤗 transformers library's AutoTokenizer to identify which tokenizer to use on every model. (4) Split into Train/Dev/Test Sets: partition your data into training, validation, and testing sets to ensure proper model evaluation and tuning. Ultimately, data quality prevails over quantity, specially as we live in the LLMs era when models have been pre-trained with almost all the public data available in the 🌎 . A well-prepared dataset tailored to your application will make all the difference in the performance and reliability of your model. #AI #MachineLearning #DataQuality #LLM #NLP #ModelTraining #DataPreparation

5

Crafting effective prompts is an art itself, but getting it right can mean the difference between a great output and total garbage! Today, we’ll explore the 5 prompt engineering techniques that every AI engineer should know to get the most out of LLMs. #llms #llmops #chatgpt4 #rag #gemini #ai https://lnkd.in/d3an85KE

3

Excited to announce that Zapier Central is now available in public beta! 🤖 Central lets you create and customize your own AI assistant to work alongside you. Your AI assistant can analyze data, provide answers, and help complete tasks across any of Zapier’s 6,000+ apps. Here’s what new: • 🤝 New integrations to chat with live data from knowledge sources like HubSpot, Airtable, Notion, Google Sheets, Google Docs and Zapier Tables. Just ask, and Central will reference the most up-to-date info when completing tasks. • ⚡Add instant actions to have your bot instantly complete tasks in any of Zapier’s 6,000+ app integrations.  • 🌐 Bring Google to you by browsing the web via chat. Ask your bot for the latest news articles and recommendations, and it’ll surface the latest results—right in Zapier Central.  • 👀 Coming soon: Central's Chrome extension makes research a breeze. As you browse blogs, videos, or other sites, your bot can instantly condense that info. Then, ask your helpful bot to use it to draft emails to a sales contact, summarize a blog post, or add data to a comparison table. Try Central today for free. I’d love to hear what you think: https://zapier.com/central

283

27 Comments

#PET #OWNERSHIP? Reminder of proxy variables Looking at map of pet dog ownership tells… … easily mistaken story. The clear geographical delineation in graph (matching the cold war Iron curtain line) can easily sway you into cultural root cause. 🤦🏻‍♂️ And that would be a great example of the correlation fellacy. Because 25 years after the Iron curtain is down is 2-3 generations of the dogs, long enough time for the political factors to wear off, if they ever had effect. 📖 The real learning of this graph is that there might be hidden factor that is not obvious and still it has profound effect on the actual “target variable”. 🐕 In our dog situation the underlying hidden root cause for the pet ownership is the share of agriculture on total labor market 1.5-2 generations back. Because if you had a farm, you were more likely to get used to living with animals or even needed some (shepherd dogs anybody?). The 2 generations lag is important, because you don’t make pet purchase decisions until you live in your own household and that is roughly, when you start your family (hence when you pass on the generation baton). 🔗 This phenomena is also known as #proxy #parameter and has significant impact also (in positive means) in #MachineLearning. Because proxy parameters are strong predictors, but are insights that markets tend to oversee. So if you know them, proxy parameters give you edge. 💡 Now you know the phenomenon I have a #puzzle for you: It is proven by data in many insurance companies (I worked on one of such studies) that #Zodiac sign is very strong predictor of the #car #accident rate. Yes, really true! Can you guess what it is proxy parameter of?

3

📬 In collaboration with Vu Trinh, we present a deep dive into MapReduce. MapReduce may be old and not really needed today, but I would still encourage folks to understand the technology as it's one of the game changing and core data processing tech which inspired a lot of innovation that we see today. 👉https://lnkd.in/evWP2uh8 #dataengineering

95

4 Comments

In this article we are providing the major architectural building blocks and blueprint for building end-to-end GenAI-based applications that are ready for production. I’d like to call out several key considerations as you go about implementing and designing these llm-based applications. #llms #llmops #chatgpt4 #aiml #gemini #mlops #data https://lnkd.in/dMev5ncq

8

Still time to come see us in SF on Nov 7 for Data Science Salon. I'll be talking about my favorite topic- AI for Automation in the non-digitally native world. The same rules do not apply. If nothing else, a fun day at GoogleHQ? #AI #machinelearning

54

7 Comments

Come and build a smart shopping assistant designed to answer customer questions, guide product discovery, and tailor search results. Combine the power of AlloyDB, Gemini (Large Language Model) and Vertex AI Agent Builder to quickly bootstrap an intelligent conversational assistant.

7

Learn how to build an autonomous research assistant using LangGraph with a team of specialized AI agents #llms #chatgpt4 #gemini #aiml #llmops https://lnkd.in/d6TUbZNj

2

Real-Time #Analytics News from the Past Week: 1. Confluent announced AI Model Inference, an upcoming feature on Confluent Cloud for Apache Flink, to enable teams to easily incorporate machine learning into #DataPipelines. 2. Crunchy Data announced a new offering, Crunchy Bridge for Analytics. Crunchy Bridge for Analytics gives users the ability to retrieve and interact with their data lake using #PostgreSQL commands, through extensions and a vectorized, parallel query engine. 3. The United States Department of Homeland Security (DHS) announced the establishment of the Artificial Intelligence Safety and Security Board (the Board). The Board will advise the DHS Secretary, the critical infrastructure community, other private sector stakeholders, and the broader public on the safe and secure development and deployment of #AI technology in the nation’s critical infrastructure. 4. LinkedIn announced the launch of LakeChime, a new data trigger service designed to streamline #data management within modern data lakes. LakeChime addresses significant infrastructure challenges associated with managing large-scale data by offering a unified solution that simplifies data triggers across both traditional and modern table formats. 5. Nokod Security announced the general availability of the Nokod Security Platform. This platform enables organizations to protect against #security threats, vulnerabilities, compliance issues, and misconfigurations introduced by low-code/no-code (LCNC) applications and robotic process automations (RPAs). Read the rest on RTInsights ➡️ https://lnkd.in/edMmS-3h

16

2 Comments

Meta's move to go Open-Source: understanding the impacts and what lies ahead If you’ve been keeping up with the tech world, you’ve likely heard about Meta’s move: the open-source release of Llama, a large language model (LLM) trained on a massive amount of text data. Unlike its competitors, like OpenAI and Google DeepMind, Meta is taking on a different strategy by making part of its AI technology widely available. 🧠 Digging deeper into this topic, we discovered interesting parallels between Meta’s approach and the economic theory of "commoditize your complement." To put it simple, the principle suggests that a company can drive demand for its core products by lowering the cost and barriers of its complementary products. 💡 We’ve seen this before. A notable example is when Google launched Android as an open-source operating system, leading to the commoditization of the mobile OS market. This move allowed Google to dominate the mobile ecosystem, ensuring its search engine, its primary revenue driver, thrived across devices. By reducing platform risks, Google secured its position as the mobile market transitioned from PCs to smartphones. (Dixon, C., 2024. Read Write Own: Building the Next Era of the Internet. United Kingdom: Cornerstone). 📈 For Meta, whose apps rely heavily on content creation, generative AI is the complement to its core social media business. By open-sourcing Llama, they are reducing the barriers for others to build AI-powered tools and services, which in turn directly and indirectly facilitates content creation and sharing on its platforms on a massive scale. In essence, they're commoditizing AI to solidify their competitive edge. 🌍 To present a different take too, some view Meta’s move as a democratization of technology. Zuckerberg himself has framed Meta's strategy as a democratizing force in AI development. By open-sourcing AI, he argues in his Manifesto published on July 23, Meta will “ensure that more people around the world have access to the benefits and opportunities of AI (…) It will make the world more prosperous and safe.” Whether you view this as a bold strategic move or a genuine effort to empower society, one thing is certain: Meta's approach has sparked an important debate in the AI community, and it will be interesting to follow its developments. 

99

8 Comments

🚀 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

Do you want to learn the biggest secret in accessing the most urgent product manager openings from the best companies directly from the hiring managers themselves? See you in Skool. https://lnkd.in/gCsMZ-Ab This was posted 28 minutes ago. #productmanagement

3

Graph databases are a game-changer in how we handle complex data relationships. Instead of rows and tables, you get nodes, edges, and properties—a structure that's built for speed and flexibility when dealing with connected data. Here’s a real-world example: In retail, we used graph databases to power our "Frequently Bought Together" recommendations. Every product is a node, and every co-purchase is an edge. This lets us quickly identify patterns—like how eyeliners in the beauty category are often bought with mascara. The result? Instant, data-driven product recommendations that boost cross-sells and enhance the shopping experience. On the tech side, this setup scales efficiently even with millions of transactions. Queries that would be heavy in a relational database—like finding all products frequently bought together—are lightning-fast with a graph. Graph databases aren’t just for social networks or fraud detection; they’re a tool we can leverage for smarter, more personalized retail experiences. If you’re dealing with interconnected data, it’s time to look into graph tech. #GraphDatabases #RetailTech #DataEngineering #AI #Personalization

17

We just released a blog on whether the use of multimodal feature transfer can detect deception (phenomena like sarcasm, irony, and condescension) in conversation. The short answer... Yes. This particular write-up profiles a subset of our work at Symbl.ai on multimodal large language models, and is research led by one of our summer interns, Benjamin R. (from Prof. Larry Heck's AVA Lab at Georgia Institute of Technology). We have a paper currently under submission on this work, and we can't wait to share the full details with all of you soon. Blog Link: https://lnkd.in/g6m_wFDi #multimodal #featuretransfer #LLMs #deception #conversation #sarcasm #irony #language #AI #ML #NLP #LargeLanguageModels #SummerInternship #ResearchPapers #ACL #Llama2 #SymblDotAI

16

Cerebras and Neural Magic have reached a significant milestone in the realm of large language models (LLMs). Through innovative techniques and purpose-built hardware, we achieved up to 70% parameter reduction in LLMs without sacrificing accuracy. This breakthrough opens doors for more efficient training and deployment, democratizing access to LLMs across various sectors. #AI #MachineLearning #Innovation

9

1 Comment