Unum’s Post

⚗ Introducing the Multilingual AI with Cohere and LanceDB (YC W22) . 👬 Leverage Cohere's new multilingual model. 💭 Creating and Ingesting data made easy with LanceDB (YC W22)'s embeddings API. 👉 Checkout the blog here. https://lnkd.in/gi3v3NYy 💻 Jump straight to code. https://lnkd.in/gxZpq9yS 👨💻 Follow the GitHub repo for more such examples and tutorials. https://lnkd.in/gU7yQWYa Do give the repo a star ⭐ #vectorsearch #vectordatabase #vectordb #vectordatabases #llms #multilingual #generatieveai

Revolutionize your LLM's capability without the hefty fine-tuning costs! 🚀💡 The power of LoRA adapters lets you boost your language model's versatility, combining different task-specific adapters into a multi-functional AI powerhouse—be it for translation or chatting. Using Llama 2 as a base, this innovative approach unlocks multi-tasking without the memory overload. Say goodbye to task-specific limitations; welcome a new era of adaptable, efficient AI. 🌐🔧 🔑 Takeaways: - Merge LoRA adapters to form a multi-tasking AI. - Customize your AI's skills without extensive retraining. - Efficient, cost-effective method to expand your AI's horizons. 🔗 Check out the hands-on guide and provided notebook for all the details! #ArtificialIntelligence #MachineLearning #LoRAAdapters #LanguageModels #AI #Innovation #MultiTasking #Efficiency #Adaptability #DataScience

🤖Agent Types 🔗 As incredible as the agents are, it is equally important to choose the appropriate agent type for our use case. 1️⃣ Zero-shot ReAct Without being trained on sepecific data Zero-shot ReAct Agent has the ability to create realistic contexts. It can be used for tasks involving generation of creative text formats, language translation, and generating different types of creative content. agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION 2️⃣ Conversational ReAct Designed for conversational use, it incorporates React framework to determine which tool to use couple with memory to remember the context of the previous interactions (conversation buffer memory). agent="conversational-react-description" 3️⃣ ReAct Docstore These Agents utilizes React framework to communicate with a docstore (cloud-based digital document archive and retrieval system). Uses a combination of lookup tool and search tool agent="react-docstore" 4️⃣ Self-ask with Search This Agent utilizes the Intermediate Answer tool (which can do wikipedia searches) for self asking questions. agent="self-ask-with-search" 🍀Pro Tip: Make use of - Verbose= True to see the "Chain of thought" series the agent goes through.  #Agents #autonomousagents #AgentTypes #AI #artificialintelligence #Langchain #LangChainAgents

How do the top LLMs compare? We partnered with Surge AI for a systematic evaluation of each across four attributes: Reasoning, Writing, Creativity, and Non-English Fluency, and the results are in. 👇 ▶ GPT-4 wins the competition on all dimensions (not counting speed and price) but there are some notable callouts in 2nd place. ▶ Google’s PaLM beats all other contenders on non-English language proficiency, supporting the broadest range of languages of the online population effectively. This seems to have not been previously reported. ▶ Claude 2 takes 2nd place on reasoning, encompassing tasks like solving math questions and programming challenges. ▶ Llama 2 70b manages 3rd place on writing and creativity, assessed in our expert human evaluation conducted with Surge AI. For an in-depth look into our evaluation methodology, visit our new bot ranking page with an interactive leaderboard and example prompts across all the measured bots: https://lnkd.in/eywN2R7Q And of course, Poe provides access to any of these official bots from OpenAI, Anthropic, Google, or Meta, along with millions of user-created bots built on top of them.

It's TODAY! 🤩 🎉 Calling all language professionals and tech-savvy linguists! Are you ready to take your translation and interpreting skills to the next level? We have the perfect solution for you. Join today's FREE webinar and learn to harness the power of Bing AI to turbocharge your translation and interpreting! Here's what you can expect from our webinar: 🌐 Search the net using Artificial Intelligence  📚 Deepen your understanding by exploring sources 🔌 Optimize your browser with the best Bing plugin 📃 Harness Bing chat to summarize web pages and extract terms from long documents Gear up to explore my top three hacks for mastering AI-powered search with Bing 🦾 tfw.rocks/freewebinar 👈  #xl8 #1nt #AI

Evaluate anything you want | Creating advanced evaluators with LLMs 1. The Importance of Evaluating Language Models 1.1. Understanding the capabilities and limitations of language models is crucial for aligning with business objectives. 1.2. Standard metrics such as perplexity, BLEU scores, and sentence distance often fail to capture the subtle nuances in real-world applications. 1.3. LLMs use 'black box' metrics to assess the quality of generated text using large language models themselves. 2. Building and Running Evaluation Examples 2.1. Build chatbots and RAG using the LangChain framework, as it incorporates simple evaluation functionalities. 2.2. Demonstrate how to create custom evaluators through detailed code examples. 2.3. Discuss the implementation of translation quality assessment and context relevance evaluation. 3. Strategies for Creating Assessment Prompts 3.1. Establishing assessment criteria and defining a numerical scoring scale are key to success. 3.2. Require reasoning behind scores to gain deeper insight into the assessment logic. 3.3. Provide queries and context for reference, and demand a strict response format for ease of parsing. 4. Implementation and Optimization of Evaluation Chains 4.1. Implement a basic evaluation chain class for parsing output scores and reasons. 4.2. Consider the randomness of evaluations by running asynchronously and averaging scores. 4.3. Integrate frameworks to leverage their maximum benefits (optional). 5. Case Studies in Practical Assessment 5.1. Assessment of translation chains from English to French, identifying and reasonably explaining issues. 5.2. Contextual relevance assessment effectively identifies information unrelated to the query. 5.3. Results visualization and experimental tracking through the Langsmith platform. 6. Conclusions and Practical Applications 6.1. Customizing control model performance allows companies to build AI systems that align with their unique business objectives. 6.2. Encourage experimentation and the creation of custom evaluators for specific use cases. 6.3. All code is available on GitHub, facilitating practical application and further development. #LLMPerformance #CustomEvaluators #LangChainTech #TranslationQuality #RealTimeFeedback

Are you curious about how to give instructions to #Bing GPT in Filipino? Sometimes it's freaky how #AI can really communicate using our own tone and humor. :) Do you want to learn how to communicate with an AI that can understand and respond in your own language and tone? If so, you might be interested in this post. Bing GPT is a powerful artificial intelligence model that can generate natural language texts based on various inputs and prompts. It can also understand and communicate fluently in different languages, such as English, 中文, 日本語, Español, Français, Deutsch, and Filipino. To give instructions to Bing GPT in Filipino, you need to follow some simple steps: • First, you need to specify the language of your input and output by using the hashtag #Filipino at the beginning of your message. This will tell Bing GPT to use Filipino as the default language for the conversation. • Second, you need to use the appropriate format and syntax for your instructions. For example, if you want Bing GPT to create a poem, you need to use the format #poem("title") followed by the content of your poem. If you want Bing GPT to search the web, you need to use the format #search_web("query") followed by your query. You can also use other predefined tools such as #graphic_art("prompt") or #code("language") to generate different types of content. • Third, you need to end your message with a period (.) or a question mark (?) to indicate that you are done with your instructions. This will trigger Bing GPT to process your input and generate a response.

🚀 I recently dived into Anoop Kunchukuttan’s survey presentation on Extending Large Language Models (LLMs) to other languages, AI4Bhārat it’s nothing short of a playbook for building customized LLMs. 🌐🛠️ 🔍 Highlights: • An E2E crisp overview on Extending the power of English-based LLMs to Indic and other languages.🌍 • Technical Breakdown: Clear intros to canonical LLMs and design patterns – essential for both beginners and experts. 🧩 • In-depth Process Guide: From extending vocabularies 📚 to embedding new words , initialisation methodologies 🧬, the survey meticulously outlines each step. • Practical Insights: Details on pretraining corpuses, continual pretraining, instruction fine-tuning, and the importance of dataset preparation by augmenting/distilling from existing models/translations. 💡 - Insights and details have been given on various data mixing strategies like romanized / code switching/ using cross lingual datases • Cross-Lingual Innovation: Discusses cross-lingual prompting methods and instruction alignment . 🤖⚙️ • Empirical Evidence: Ablation studies to back the theories and provide concrete data to easily understand what worked better on these activities 📊 This survey is not just an academic read but a solid reference for anyone looking to adapt LLMs to specific domains or languages. It underscores the critical steps in making AI accessible and functional across linguistic barriers. A massive shoutout to Anoop Kunchukuttan ! #AI #LanguageModels #MachineLearning #InclusiveAI #AI4Bharat #generativeai #deelearning

This Bengali Sentence Error Correction Model is trained on 1.3 M Bengali incorrect-correct sentence pairs and archives a score of BLEU: 0.443, CER: 0.159, WER: 0.406, and Meteor: 0.655 when tested on unseen data. For a closer look at the model in action, check out the live demo here: https://lnkd.in/gSsh7u6x The full training process and methodology can be reviewed on GitHub: https://lnkd.in/gzeDC_TK The model was trained on Google Colab for a single epoch, spanning 6500 steps with a batch size of 128. I believe that expanding the dataset and extending the training duration could significantly enhance the model’s performance and utility. That would be the next step for this model. Try this model and let me know your experience. https://lnkd.in/gb6iPqwQ #BengaliNLP #MachineLearning #NaturalLanguageProcessing #Finetune #LLM #DataScience #AI #DeepLearning #HuggingFace #ErrorCorrectionModel

📅 July 10, 2024 AIBuzzWorld Daily Newsletter! Dive into the fascinating world of Artificial Intelligence and be the first to learn about the latest AI news: 1. **LLMs Struggle with Book-Length Text**      • Current long-context LLMs struggle to understand and reason over book-length texts.   • Researchers created NOCHA, a dataset to test LLM's comprehension of lengthy narratives.   • Even advanced LLMs like GPT-4o achieved only 55.8% accuracy.   • Readmore: https://lnkd.in/g2tFiH5g 2. **Quora’s Poe Introduces Artifacts-Like Feature**      • Quora’s Poe now has a feature to create custom web apps within the chat.   • This feature works well with LLMs that excel at coding.   • Users can create interactive experiences like games and data visualizations.   • Readmore: https://lnkd.in/gSJffbB9 3. **New Features in Ollama 0.2**      • Ollama 0.2 now supports multiple chat sessions and running various models simultaneously.   • Users can load different models for tasks like RAG and running agents.   • This update improves efficiency and multitasking.   • Readmore: https://lnkd.in/g-BCN4zH 4. **Wheebot for Quick Landing Page Creation**      • Wheebot allows users to create and edit landing pages via WhatsApp.   • Users can describe their requirements in plain English.   • Wheebot generates and updates sites instantly through encrypted chats.   • Readmore: https://lnkd.in/gsRtzGrd #AI #ArtificialIntelligence #MachineLearning #LLM #Microsoft #Quora #ClaudeAI #Wheebot #TechNews #Innovation #AIUpdates #FutureTech