ThirdAI retweeted
A Rice-developed tool called PRISM (Preventive RISk Monitoring) uses generative AI to enhance research security. PRISM is the product of a collaboration between Rice’s Office of Research Security and ThirdAI, a @RiceCompSci-launched startup. news.rice.edu/news/2025/how-…
ThirdAI retweeted
Privacy Preserving GenAI at Scale: Anonymize Your Text without GPUs for a hundredth the cost. --- Free your GPU cycles for other tasks.
medium.com/thirdai-blog/priv…
#privacy #security #GenAI #scale #Anonymize #LLMs #GPU #latency
ThirdAI retweeted
NEW: In an exclusive interview, I chatted with @headinthebox about his next steps now that he's left @Meta—which comes as other top Meta AI talent such as @abhshkdz and @deviparikh have also recently departed.
fortune.com/2024/04/02/mark-…
ThirdAI retweeted
NEW: In an exclusive interview, I chatted with @headinthebox about his next steps now that he's left @Meta—which comes as other top Meta AI talent such as @abhshkdz and @deviparikh have also recently departed.
fortune.com/2024/04/02/mark-…
ThirdAI retweeted
🚀 Unleash the Power of AI with @ThirdAILab PocketLLM - Your Personal Knowledge Search Engine! 🧠
Discover the future of document search with PocketLLM, a revolutionary tool powered by AI and LLMs. It’s like having a personal knowledge search engine that can memorize thousands of pages, PDFs, documents, and even scrape URLs for you to search through.
🔒 What sets PocketLLM apart is its commitment to privacy. All files and models are stored locally on your device, ensuring that only you have access to them. You can easily add, update, or delete models and files at any time.
🎯 Whether you’re a legal firm, a researcher, or someone who wants to build an internal knowledge base from documents for fast search and retrieval, PocketLLM is for you. Plus, it’s capable of learning from feedback, allowing you to fine-tune results and teach it new associations.
⚡ Experience hyperfast training and retrieval with PocketLLM. Train up to a billion parameter model on your laptop in a few minutes and get instant results from search. Get started for free today! 💻
Here are some key features:
🔸𝗡𝗮𝘁𝗶𝘃𝗲 𝗚𝗺𝗮𝗶𝗹 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻: Made email management smarter and more private. With native Gmail support, PocketLLM allows you to sync your emails directly to your laptop, enabling AI-assisted searches and responses. It’s email, redefined. 📧
🔸𝗣𝗲𝗿𝘀𝗼𝗻𝗮𝗹 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗦𝗲𝗮𝗿𝗰𝗵 𝗘𝗻𝗴𝗶𝗻𝗲: Memorizes thousands of pages, PDFs, documents, and even scrape URLs for you to search through.
🔸𝗣𝗿𝗶𝘃𝗮𝗰𝘆 𝗙𝗼𝗰𝘂𝘀𝗲𝗱: All files and models are stored locally on your device.
🔸𝗨𝘀𝗲𝗿 𝗙𝗿𝗶𝗲𝗻𝗱𝗹𝘆: Easily add, update, or delete models and files at any time.
🔸𝗔𝗱𝗮𝗽𝘁𝗮𝗯𝗹𝗲: Capable of learning from feedback, allowing you to fine-tune results and teach it new associations.
PocketLLM, your personal AI-powered knowledge search engine, is now live on Product Hunt! It’s privacy-focused, user-friendly, and adaptable.
📌𝗗𝗼𝘄𝗻𝗹𝗼𝗮𝗱 𝗣𝗼𝗰𝗸𝗲𝘁𝗟𝗟𝗠 𝗵𝗲𝗿𝗲: thirdai.com/pocketllm/
Experience hyperfast training and retrieval, and get started for free today. We appreciate your support.
👉 𝗦𝘂𝗽𝗽𝗼𝗿𝘁 𝘂𝘀 𝗵𝗲𝗿𝗲: producthunt.com/posts/thirda…
#PocketLLM #AI #KnowledgeSearch #PrivacyFirst #HyperfastTraining
ThirdAI retweeted
Cross the Chasm with #RAG: Implicit Feedback and Click-Through Data
medium.com/thirdai-blog/cros…
#GenAI #search #LLM #ChatGPT4 #AcademicChatter
ThirdAI retweeted
We’re helping #AI innovators deliver best-in-class solutions to drive value to their customers by leveraging Intel AI technologies. Learn how Intel® Gaudi®2 accelerators and Intel® AMX are helping a new generation of disruptors. intel.ly/3uYMoNV
ThirdAI retweeted
ThirdAI has <30 people and they are crushing it. @ThirdAILab @Anshumali_
Houston has the STEM talent. :)
It's only 160x more efficient than @OpenAI.
Looks like Scam Altman has some competition.
ThirdAI retweeted
Watad Partners with ThirdAI to Boost AI Integration in Saudi Arabia
ow.ly/my1550PLFtN @watadenergy @ThirdAILab #TCIwire #AIwire
ThirdAI retweeted
"Are You Pre-training your #RAG Models on Your Raw Text?" The blog and the experiments will likely change our understanding about pre-training & foundation models capabilities
#LLM #chatgpt #vectordatabase #openai #AI #GenAI #foundationmodels #Llama2
medium.com/thirdai-blog/are-…
ThirdAI retweeted
New post: ThirdAI Training of Billion-Parameter Neural Networks on Commodity CPUs cryptogon.com/?p=67616
ThirdAI retweeted
anyscale.com/blog/how-thirda…
🚀 ThirdAI democratizes AI with training on standard CPUs.
🧠 BOLT engine sometimes outperforms GPU-based models in speed and latency.
📊 Experimental results show near-linear scaling on terabyte-scale datasets.
ThirdAI retweeted
Replying to @ThirdAILab
@ThirdAILab leverages @raydistributed w/ BOLT engine to train a billion-parameter neural network model using only CPUs yielding:
☑️ reduced memory by 2x w/ Ray Train & BoltTrainer for distributed training
☑️ simplified the developer experience w/ Ray Train APIs & BoltTrainer
👇
ThirdAI retweeted
Worried about #GPU shortages or costs? Locally sensitive hash functions can supercharge #artificalintelligence performance on #CPUs Computer scientists from @RiceUniversity and @ThirdAILab make the case for pocket #LLMs Backgounder 👉 techhq.com/2023/08/is-runnin…
ThirdAI retweeted
Rice CS faculty member @Anshumali_ Shrivastava is featured in this @InnoMapHou article. He discusses the new @ThirdAILab PocketLLM app, a multilingual ChatGPT-like AI training tool. bit.ly/3QpX3JP
ThirdAI retweeted
PocketLLM Beta from @ThirdAILab: The Ultimate #GenAI Hub for @Microsoft @Windows and @Apple #macOS Users. A #Private On-Device Neural #Database for Personalized Retrieval-Augmented Generation (#RAG) and Optional Integration with @OpenAI #chatGPT
medium.com/thirdai-blog/intr…
ThirdAI retweeted
Do you know embedding are heavier than raw text itself? Don't blow you data out of proportion!
Neural Database: Next Generation Context Retrieval System for Building Specialized #AI-Agents with #ChatGPT link.medium.com/SLafs35QEAb
ThirdAI retweeted
Rice CS faculty member @Anshumali_ Shrivastava is quoted in @IEEESpectrum. In this article, he discusses democratizing AI with CPUs.
Until recently, AI researchers were all in on GPUs. A small but growing number are arguing that simpler CPUs deserve another chance to prove their worth. #AI spectrum.ieee.org/ai-cpu?sha…
ThirdAI retweeted
Until recently, AI researchers were all in on GPUs. A small but growing number are arguing that simpler CPUs deserve another chance to prove their worth. #AI spectrum.ieee.org/ai-cpu?sha…
ThirdAI retweeted
As @ThirdAILab CEO and Founder @Anshumali_ explains to Neeraj Hablani, with ThirdAI sparsity-based training and inference, you can get the best AI out there and meet your latency demands. Learn how ThirdAI is making generative AI accessible to all: https://nitter.gq/t.co/h4Puzc17ae



















