Google I/O 2023: Announcements leaked!
Plus: RasaGPT, MultiModal-GPT, Autodesk's AI-based features in Forma
Hello, Engineering Leaders and AI enthusiasts,
Welcome to the 15th edition of The AI Edge newsletter. In today’s edition, we bring you the expected big reveals at Google I/O 2023. Thank you everyone who is reading this.
In today’s edition:
📢Google I/O 2023: Announcements leaked
🤖RasaGPT: The first headless LLM chatbot platform
🧠MultiModal-GPT: The smartest assistant you'll ever need
💡Autodesk unveils AI-powered insights with Forma
💬The magic behind ChatGPT: RLHF by
Let’s go!
Google I/O 2023 predictions: What to expect on the AI front
Google is set to make several AI-related announcements at its annual developer conference, Google I/O 2023. This includes updates to existing AI technologies and products, as well as a new general-use large language model called PaLM-2.
PaLM-2 (Pathways Language Model) is the next-generation language model expected to outperform its predecessor, PaLM. Similar to Google Bard and ChatGPT, PaLM-2 is a language model that can understand and generate human-like language. PaLM-2 has been operating under the codename "Unified Language Model. Has a capacity of more than 100 languages.
Why does this matter?
PaLM-2 could have significant implications for language translation, making it easier to communicate with people who speak different languages. It could also accelerate the development of new applications that require advanced language capabilities, such as language translation and language-based search.
RasaGPT: The first headless LLM chatbot platform
RasaGPT is the first headless chatbot platform built on top of Rasa and Langchain, which provides a boilerplate and reference implementation of Rasa and Telegram using an LLM library for indexing, retrieval, and context injection.
It is designed to work on:
➡ Creating your own proprietary bot end-point using FastAPI, document upload, and “training” pipeline included
➡ How to integrate Langchain/LlamaIndex and Rasa
➡ Library conflicts with LLM libraries and passing metadata
➡ Dockerized support on MacOS for running Rasa
➡ Reverse proxy with chatbots via ngrok
➡ Implementing PGVector with your own custom schema instead of using Langchain’s highly opinionated PGVector class
➡ Adding multi-tenancy (Rasa doesn't natively support this), sessions, and metadata between Rasa and your own backend/application
Why does this matter?
RasaGPT eliminates many of the common implementation headaches and allows developers to focus on building the chatbot's logic and functionality. This can save significant time and resources and enable developers to create chatbots more efficiently.
Overall, RasaGPT can help democratize chatbot development and enable more developers to create powerful and effective chatbot solutions
MultiModal-GPT: The smartest assistant you'll ever need
MultiModal-GPT is a vision and language model designed to conduct multi-round dialogue with humans. It can follow instructions, generate captions, count objects, and answer questions.
It is a parameter-efficient model fine-tuned from OpenFlamingo. It uses a Low-rank Adapter (LoRA) and is Trained using a unified instruction template for both language and visual instruction data.
Why does this matter?
It can potentially improve human-machine interactions, For example, it can be used in customer service chatbots to provide personalized and helpful responses to customer inquiries. It can also be helpful in various domains, such as robotics and automation, where machines need to perform specific tasks based on human commands.
Overall, MultiModal-GPT can help to bridge the gap between humans and machines and make communication more seamless and effective.
Autodesk introduces AI-powered insights in Forma’s first set of features
Autodesk unveiled its latest release, Forma, with the first set of capabilities. A key feature is AI-powered analysis in real-time to uncover greater insights into a building site and its surroundings for key factors such as sunlight, daylight potential, wind, and microclimate. This will also simplify design concept exploration and offload repetitive tasks, giving architects time to focus on creative solutions.
Why does this matter?
The use of AI in AECO (Architecture, Engineering, Construction, and Operations) is rapidly gaining momentum. As the industry continues to embrace AI, we can expect to see more innovative solutions that will improve the efficiency and accuracy of AECO workflows and augment them in a data-rich world.
Knowledge Nugget: Edge 266: The magic behind ChatGPT: Reinforcement Learning with Human Feedback
Did you know the core idea enabling ChatGPT comes from a 2017 research paper?
This insightful piece by
explains RLHF (Reinforcement Learning from Human Feedback) and how OpenAI utilized it in the development of ChatGPT.
The article also discusses how the use of RLHF in ChatGPT was pioneered by a previous model called InstructGPT. And it concludes by noting that some of the core ideas that differentiated ChatGPT from GPT-3 dated back to 2017 and how they were created with a completely different purpose.
Why does this matter?
Shedding light on the limitations and advancements in developing and training language models, this knowledge provides understanding and insights for future research and development. Plus, it can help researchers and developers create more effective and versatile models.
What Else Is Happening
💭 Create your dream world in seconds with Lovelace Studio. (Link)
🤖 Compare Language Models with Chatbot Arena by LMSYS. (Link)
🖌️ Google’s New Online Game, Spot A.I.-Generated Art. (Link)
🔮 An AI can predict Pancreatic Cancer. (Link)
💻 Create Vector Representation With Ai-Powered Algolia Neuralsearch. (Link)
💰 Cut Cloud Costs with Antimetal's AI-Powered Solution. (Link)
🤝 KPMG and Google Cloud Partner to Boost Enterprise Adoption of Generative AI. (Link)
🤟 Yolo-Nas - A New Open-Source Object Detection Architecture (Link)
Trending Tools
Cometcore: Meet CometCore, the ultimate AI platform! Multimedia, coding & content with AI agents, smart editing & automation.
Bugasura: Bugasura AI is a smart add-on to the issue tracker that unlocks bug impact and reduces redundancy by detecting similar issues.
Flagright: Flagright AI automates repetitive tasks for fincrime ops teams and brings GPT to AML compliance teams globally!
Briefly: Transcribes calls, generates summaries, assigns action items. Transform conversations into emails and docs.
Bito: AI-driven platform for apps, websites, NFTs & more. Describe your project and get a custom design.
Dataherald: Supercharges GPT with structured data for content generation. Technical preview offers limited functionality.
Nightfall: AI-powered data leak prevention for Zendesk. Discover sensitive data in tickets and files.
Whismer: Private AI assistant to create a customizable knowledge base. Add files, web links, and notes.
That's all for now!
If you are new to ‘The AI Edge’ newsletter. Subscribe to receive the ‘Ultimate AI tools and ChatGPT Prompt guide’ specifically designed for Engineering Leaders and AI enthusiasts.
Thanks for reading, and see you tomorrow.