Age Of Geeks

Dynamic RAG projects are gaining interest and they show an understanding of production LLM applications. 👀 Here's a useful thread on Langchain's subreddit that will help you build dynamic RAG projects in Google Colab using Langchain and Pathway. 📌 Make sure you upvote the thread and read the Colab link for getting started: https://www.reddit.com/r/LangChain/s/kvgGcGYRv5

OpenAI has confirmed the departure of its co-founder and chief scientist, Ilya Sutskever! 👀 Here are the specifics: - Sutskever expressed his belief that OpenAI is well-positioned to develop AGI that is safe and beneficial under its existing leadership. - Also exiting the company is Jan Leike, co-leader of Sutskever’s superalignment group, who shared his decision to leave through a mysterious resignation note. - These exits are part of a broader trend of recent departures from OpenAI’s superalignment and safety teams, sparking debates and speculation. - OpenAI's CEO Sam Altman has appointed Jakub Pachocki, a pivotal figure in the development of GPT-4, as the new chief scientist. The ongoing uncertainty about Ilya’s role since last year's shakeup has now been resolved. However, the continued exodus from the safety team raises ongoing concerns.

🚀 Just released! This 30-minute video is a gem. Łukasz Kaiser (now a Senior Researcher at OpenAI), co-author of Transformer & TensorFlow, and his former Google Brain colleague Jan Chorowski (now CTO of Pathway.com) reveal groundbreaking insights on the future of LLMs and deep learning. Watch now! 👇 https://youtu.be/_7VirEqCZ4g

The irony of this week is Stability AI displaying instability. 👀 According to a new report from The Information, Stability AI, creator of Stable Diffusion Models, is facing a cash crunch and is looking for a buyer.

Google I/O 2024 summary in 10 minutes. 👾 Video credits: Google.

Here are 12 spectacular announcements made at Google I/O 2024 that make us believe that Google is back in the AI race! 👀 1. Project Astra: This innovative AI agent has the ability to both see and hear in real-time, providing live contextual responses. 2. Veo: Google introduced Veo, a new video generation model capable of producing high-quality 1080P videos from text, image, and video inputs. 3. AI Teammates: New AI agents are now integrated within Google Workspace to assist with handling queries related to emails, meetings, and other data. 4. Search Enhancements: Google Search has been upgraded with expanded AI overviews and enhanced planning tools, organizing search results more effectively. It also includes the capability for users to pose questions through video, letting the system analyze visual content to offer useful insights. 5. Multi-step Reasoning in Search: Gemini's new functionality allows for breaking down complex queries into manageable steps, significantly accelerating the research process and enabling more efficient itinerary planning. 6. Gemini 1.5 Pro: The update boasts a significantly increased context window of 2M tokens and improved performance across code, logic, and image comprehension, with a preview feature reaching up to 10M tokens. 7. Imagen 3: An advancement of Google's text-to-image model, offering greater detail, text generation, and natural language understanding than its previous iterations. 8. VideoFX: This text-to-video tool introduces a 'storyboard' interface for detailed scene creation and supports adding music to video generations. It is currently in private preview in the U.S. for select creators. 9. Gemini Live: Enhancing user interaction, this feature allows real-time conversations with Gemini, mimicking live communication. 10. Gemini 1.5 Flash: A new, optimized version of the AI model designed for fast performance and efficiency, with a 1M token context window. 11. Gemma 2: Google's next-generation open-source models are now more accessible, and able to run efficiently on just a single TPU or GPU. 12. Gems: For Gemini Advanced subscribers, the introduction of 'Gems' allows the creation of customized personas from a simple text description, mirroring the functionality of personalized GPTs in ChatGPT.

In image: A user on X (/HistoryGPT) gave GPT-4o the following letter and asked for transcription. As we can see, GPT-4o understood the assignment. 💯

OpenAI announced GPT-4o, a GPT-4-level intelligence with capabilities across text, voice, and vision, to everyone for FREE. We just tried it, and it's surreal. 🤯 When using GPT-4o, ChatGPT Free users will now have access to features such as: - Experience GPT-4 level intelligence - Get responses from both the model and the web - Analyze data and create charts - Chat about photos you take - Upload files for assistance summarizing, writing, or analyzing - Discover and use GPTs and the GPT Store - Build a more helpful experience with Memory

📄 AI Paper Of The Week - More Agents Is All You Need. The study showcases how leveraging a larger number of basic agents, created by Large Language Models (LLMs), can enhance performance through a straightforward method of sampling and consensus. This discovery suggests that the complexity of tasks can be addressed by scaling up the number of agents, bypassing the necessity for elaborate Chain of Thought (CoT) processes or intricate frameworks for multi-agent collaboration. To us, it appears that the current utilisation of agent-based approaches is not fully tapped into. 🧐

Announcement! 📣For folks, who'd like to get AI updates via whatsapp, we have created a WhatsApp Channel. We're sharing a list of 10 AI companies, remotely hiring freshers and interns, on our WhatsApp channel tomorrow at 2:00 PM IST. Join in! 🙌🏼 https://whatsapp.com/channel/0029VaJDm6FEFeXi6GpWtT0u

Google's new announcement - Bard is now known as Gemini (free). 🤘🏼 Details: https://blog.google/products/gemini/bard-gemini-advanced-app/ Ultra 1.0, their most capable AI model, will be available in Gemini Advanced (paid), a new experience that helps with highly complex tasks like reasoning, following instructions, coding and creative collaboration.

Dear AOG family, for community discussions on AI [news, jobs, events, papers, and more], please join our brand-new Slack space! 🚀 🔗 https://join.slack.com/t/ageofgeeks/shared_invite/zt-2bxga5b5y-O_pqmV9Cfp9SNBN4XV9Sug

If you know, you know! 😂

Happy 2024 folks! 🎉 We're sure that staying at the forefront of all AI developments is a part of your list of resolutions for 2024. We've created a free-to-join AI community where we dive into whatever is latest in GenAI, ML, and data science. There's a lot of sharing and discussion among ~700 experts, beginners, and everyone in between. Given your profile and experience, we'd love to invite you there! 🦾 Join in: https://join.slack.com/t/markovmlcommunity/shared_invite/zt-29udk32lf-A6OaN5tQQF~nJ7mgra~DDA

🚨 BREAKING 🚨 Apple's new LLM has arrived. The name's - Ferret. Ferret is a Multimodal Large Language Model (MLLM) capable of understanding spatial referring of any shape or granularity within an image and accurately grounding open-vocabulary descriptions. Read the paper here: https://arxiv.org/abs/2310.07704 Github repo: https://github.com/apple/ml-ferret

IIT Guwahati is running this basic (3-week) free course on Generative AI and it's open for all. Coursework has just started and the registrations are open till this weekend, in case you're interested. 🔗 https://lu.ma/llm23 Course contributors include former researchers at Google Brain, Mila AI, and Microsoft Research. Join now.

Hello AOG fam! 🙌🏼 MarkovML is building this free-to-use, no-code platform that enables people to comprehensively analyse their datasets and get actionable insights with just a few clicks. Could you please give it a try? You can sign in here. Thanks a ton!

Ever wanted to see what was happening inside Large Language Models like ChatGPT? This new 3D LLM visualization tool shows all the steps to run a single token inference. 😍 Check it out: https://bbycroft.net/llm _________________________ We'd like to invite you to a Slack community, where we actively share and discuss state-of-the-art research papers, the latest updates, and other useful resources in AI, ML, and data science! 🦾 Join in through: https://join.slack.com/t/markovmlcommunity/shared_invite/zt-269j7xqgi-He4b2b5dur7K5924RG0DQw Please note that (since this is an exclusive community) the invite link expires in a few days.

Valuable video for all AI/ML developers! 🤘🏼 Streaming systems are crucial for processing data instantaneously and offering real-time insights beyond the scope of traditional batch processing used in AI/ML or any application for that matter. This 40-minute, FREE resource is a great place to learn about its practical aspects: https://youtu.be/MhSB5NqkylU. 💡