Octo-oLabs’ Post

Students taking the Hardware for Machine Learning course at @UCBerkeley will get to see some amazing demos and hear from our CEO & Founder, Jonathan Ross about his experience with ML acceleration.

Learn more next week about how Sigma Computing and Databricks are working together to advance analytics and machine learning goals. #DataAnalytics #DataAccessibility!!!

Within a keynote to a hackathon yesterday, Andrej Karpathy talked briefly about "LMOS" (Language Model Operating Systems). https://lnkd.in/eWqX2svF (starting at 6:22) His view as stated, in my opinion, ignores the benefit of optimized CPU code. In the realm of an LMOS, CPU code would be analogous to "muscle memory". When learning a new sport or body movement, the motion is fully consciously done, but after repetition over weeks, muscle memory develops where the motion is subconsciously done - and quickly and more accurately and more repeatably. An LMOS should therefore monitor itself, identify frequently recurring computations/prompts, and write some CPU code to execute for when it encounters a similar scenario in the future. If we use the colloquial concept of an OS as opposed to the technical one (which is, specifically, more of just the OS kernel), the UI of an OS is perceived by the end user as being the OS. So of course an LMOS would have natural language, speech, and gestures as input, but the back end would be handled, depending on the situation, either by the (likely agentic) language model directly or by identifying and executing relevant pre-generated CPU code.

Last weekend, together with Erdy Suryadarma, I had a chance to participate in a DOGE hackathon event hosted by lablab.ai. The challenge was to revolutionize government efficiency using xAI's Grok. We come up with an idea to streamline bureaucratic processes by: - providing citizens an intuitive platform to access government services and track the process, with an AI to assist in each step, - and workflow builder to help the government digitize and automate bureaucratic processes. Check out our submission here! we have the repo link, the demo link, and the pitch deck there! https://lnkd.in/gDxWesU3

4:00 PM today. Join my session at AI21 Labs booth #1827 to check when to use Jamba and how an AI Agent helps recruiting process. Let’s go!

This sounds amazing! 🚀 I'm excited to participate in the hackathon and explore the Mistral AI fine-tuning API. The optimized LoRA techniques and the option to deploy custom models through both open-source code and a managed API sound like game-changers. Looking forward to diving into the docs and getting started. Thanks for sharing! 🙌 #mistralai #mistral #llms #genai #artificialintelligence

Did you know that 90% of the world's data was generated in just the last two years? 🤯 While this data explosion offers immense potential, it also presents significant challenges. Many businesses struggle with data management, real-time insights, and compliance. At Dview, we're turning this challenge into an opportunity. We're developing innovative solutions to help businesses harness the power of their data. Want to learn more? Catch us at the Bengaluru Tech Summit on 19th & 20th November. Let's discuss how we can help you navigate this data-driven future. #DataScience #DataAnalytics #BigData #DataEngineering #DataPrivacy #BengaluruTechSummit Supratik Shankar, CFA | Kauts Shukla | Surjodoy Ghosh | Akansha Gupta

Day two of Data Innovation Summit starting! Stop by in C47 and learn how our federated learning software FEDn can help reduce data transmission for AI on the edge, and how it increases data privacy when training data is distributed and can't be pooled.

We love to see creative minds + technology = innovative applications in our daily works. HTX (Home Team Science & Technology Agency). The Teams have demonstrated such trait’s through their MVP with Streamlit in Snowflake & #CortexAI Thank you, Pan Yong Ng & Ivan Leong for organizing this hackathon. We hope to see brilliant minds creating more AI products which aide in productivity and delivering cool services for their users. Thank you Fayanne Foo Sasi Subramaniam ChangBoon Heng Patricia Leong, for your efforts in supporting the event! #ResponsibleAI #AIDrivenAnalytics #DatacentricAI

Last weekend, I competed in the Open Source for AI hackathon with Alexander Samuel, Hansel Lee, and Tien-Chern Woon at the Microsoft Reactor in Redmond. We had little to no experience with LLMs before this event, but in a matter of 6 hours, we were able to create a small project with the name of Project Cultural Awareness: a chat application where users chat AI of different ages and cultures. We utilized the LangChain framework to build an LLM, processing user input and displaying output based on bot’s age and culture. We then incorporated the PyAudio library so that users can speak to the AI bot of their choice. Finally, we integrated the Stable Diffusion AI image generation tool from OctoAI to generate images from the AI bot’s output. With this project, we were able to win 1st place in the exploration division of this hackathon! Thanks to Yujian Tang, OSS4AI, and OctoAI for a successful event! Here is the GitHub repo to our project: https://lnkd.in/gTdmkS5H