YeagerAI

Last week we co-hosted an event, "AI For Startups" for NY Tech Week with Queens Economic Development We provide high and low-level explanations of the key concepts in GenAI (LLMs, Fine-tuning, Embeddings, Prompt Engineering, RAG, etc.) and AI Agents. Ultimately, we discuss how startups can leverage GenAI to accomplish much more with limited resources. https://lnkd.in/eJXJQngh

José María Lago quoted here: "The challenge is having the right mix between deterministic and non-deterministic (agents’) routines. More mission-critical tasks require more determinism. At GenWorlds, we allow the developer to choose where to draw that line.” thanks e2b.dev https://lnkd.in/ed77N25V

Tune in tomorrow for a talk with the Queens Economic Development Corporation on the future of AI for startups with AI expert José María Lago .

“What’s the difference between 🧬🌍GenWorlds and Microsoft Research's AutoGen?" This is a question we have been getting often lately. We are impressed with AutoGen, but think it's important to lay out some of the key differences 👇   Integration and Connectivity through WebSockets🌐 🧬🌍GenWorlds employs a WebSocket-based event system, which significantly eases the integration with various existing technologies, such as frontend and backend systems, enhancing connectivity and interoperability.   Seamless Technological Cohesion🔗 Leveraging the event-driven nature through WebSockets not only streamlines communication among agents and objects within the environment but also enables effortless coordination with external technologies and user interfaces.   Diverse Interaction Modalities🔄 Agents in 🧬🌍GenWorlds are not bound to conversation as the sole means of interaction. They can engage through a variety of mediums, such as interacting via objects or dispatching distinct events, adding depth to interactions.   Collaboration Beyond Dialogue🤝 AutoGen agents collaborate through dialogue. In 🧬🌍GenWorlds, collaboration includes a project manager, objects like a ‘microphone token’ that agents pass around, or a router assigning tasks deterministically. GenWorlds also allows for further collaboration methods to be added.     Isolating Tasks for Precise Actions🎯 GenWorlds uses a Think & Do approach to ensure that actions and communication events are distinctly processed, enabling clearer task isolation. This enhances our agent’s capacity to hone in and execute specific tasks that users desire with elevated precision.   Advanced Customization and Deterministic Balance⚖️ A hallmark of our framework is the robust customization and the adept management of deterministic and non-deterministic elements, ensuring a meticulously tailored, interactive, and functional agent environment.   Object-Centric Capability Enhancement🌎 Objects within our simulated worlds facilitate an easy expansion of agent functionalities, not constricted to conversational abilities, thereby empowering agents to evolve and adeptly perform complex tasks.   Autonomous Agent Framework🤖 Our approach is deeply rooted in developing autonomous agents, providing them a varied and enriched environment to operate in, diverging from a primarily conversational focus, and allowing a myriad of problem-solving applications.   Conclusion: Two Distinctive Frameworks🚀 Our framework with its integration capabilities, distinctive agent operational approach, and enriched environment, presents a strong alternative in the realm of agent interaction and problem-solving.   🧬🌍GenWorlds Resources https://lnkd.in/eQtmQX3J https://lnkd.in/ej3CaUYr https://lnkd.in/eMyta4uE https://lnkd.in/ebHvCTe3

🚀Announcing a major refactor of 🧬🌍GenWorlds, the event-based communication framework for autonomous multi-agent systems! Now, developers can seamlessly create multi-agent environments using just a Jupyter Notebook Project Jupyter. Learn more here: GitHub: https://lnkd.in/eQtmQX3J Getting Started: https://lnkd.in/e7_udQGp We have been thrilled thus far with the reception from folks building in the AI Agent communities. If you are building any applications with AI agents (or know of projects that are), we'd be keen to gather your feedback as well.     More on the framework below 👇   Scalable Customization Introducing a split between the Abstract Interface Layer and Basic Utility Layer! Get the best of both worlds with full customization while getting started ⚡️swiftly with our utility layer, covering many use cases right out of the box.    Thought Actions Actions that leverage LLM calls for both reasoning and execution, bridging 🧠 intelligent thinking with practical actions in your autonomous agents. Unleash smarter, more responsive agents such as our basic assistant (Basic Assistant Repo) made with LangChain.   Cost Efficiency  GenWorlds optimizes LLM 💲costs💲. For basic actions/thought loops (like filling the event), agents now leverage GPT-3.5 instead of GPT-4, significantly reducing expenses without compromising capabilities.    Actions are Event Driven Actions, mainly Python routines, are the 🪨cornerstone of the framework. Actions consist of a Triggering Event and a routine that will be executed and sending events, which simplifies and ✨energizes your agent's tasks!    Easy Deployment 🔄Deploying your environment comes down to 2 main components: 'World' (a FastAPI WebSocket Server + Clients) and the 'Frontend'. GenWorlds becomes your non-deterministic backend, which easily connects to a frontend. Minimal setup, maximum functionality   Try it and give Feedback We greatly 🙏appreciate🙏 any feedback on the framework and we are here to support anyone building on it. Check out our new documentation and along with other resources below:   GitHub: https://lnkd.in/eQtmQX3J Tutorial: https://lnkd.in/e7_udQGp Discord: https://lnkd.in/e_RBWzxv Medium: https://lnkd.in/ebHvCTe3

Customers and Investors often ask if we are fine-tuning models or prompt engineering. Fine-tuning models costs 10x more per call! So, we build our agentic applications with a sophisticated prompt engineering technique known as Retrieval Augmented Generation (AKA "RAG"), leveraging limited use of a fine-tuned model for data processing. Together with Qdrant, we wrote a piece exploring this and more considerations. Check it out here: https://lnkd.in/e3TpH_WT Thanks Andre Zayarni and David Martin Learn more about 🧬🌍 GenWorlds the event-based communication layer for multi-agent systems: https://lnkd.in/eMv2ENw2 Learn more about Qdrant Vector Data Bases: https://qdrant.tech/

AI Agents are at the forefront of the transition to Autonomous Software. Just about every business is trying to figure out how to use Generative AI. Most of the focus is on LLMs (or “Large Language Models”). Use cases revolve around fine-tuning models and slapping APIs on them for domain-specific applications. While undeniably critical, we believe that the majority are overlooking the true source of impending revolutionary value — AI Agents empowered by LLMs. Check out our post on the topic 👇 https://lnkd.in/enpm3rRa

Honored to co-publish an article with LangChain, diving deeper into the 🧬🌍GenWorlds framework for coordinating powered by LangChain. https://lnkd.in/d24ThVnU Harrison Chase Brie Wolfson, many thanks for the support! 🧬🌍GenWorlds GitHub: https://lnkd.in/eQtmQX3J #aiagents #theresanagent4that