ReportWire

Tag: gpt

  • WTF Fun Fact 13718 – Recreating the Holodeck

    WTF Fun Fact 13718 – Recreating the Holodeck

    [ad_1]

    Engineers from the University of Pennsylvania have generated a tool inspired by Star Trek’s Holodeck. It uses advances in AI to transform how we interact with digital spaces.

    The Power of Language in Creating Virtual Worlds

    In Star Trek, the Holodeck was a revolutionary concept, a room that could simulate any environment based on verbal commands. Today, that concept has moved closer to reality. The UPenn team has developed a system where users describe the environment they need, and AI brings it to life. This system relies heavily on large language models (LLMs), like ChatGPT. These models understand and process human language to create detailed virtual scenes.

    For example, if a user requests a “1b1b apartment for a researcher with a cat,” the AI breaks this down into actionable items. It designs the space, selects appropriate objects from a digital library, and arranges them realistically within the environment. This method simplifies the creation of virtual spaces and opens up possibilities for training AI in scenarios that mimic real-world complexity.

    The Holodeck-Inspired System

    Traditionally, virtual environments for AI training were crafted by artists, a time-consuming and limited process. Now, with the Holodeck-inspired system, millions of diverse and complex environments can be generated quickly and efficiently. This abundance of training data is crucial for developing ’embodied AI’, robots that understand and navigate our world.

    Just think of the practical indications. For example, robots can be trained in these virtual worlds to perform tasks ranging from household chores to complex industrial jobs before they ever interact with the real world. This training ensures that AI behaves as expected in real-life situations, reducing errors and improving efficiency.

    A Leap Forward in AI Training and Functionality

    The University of Pennsylvania’s project goes beyond generating simple spaces. It tests these environments with real AI systems to refine their ability to interact with and navigate these spaces. For instance, an AI trained in a virtual music room was significantly better at locating a piano compared to traditional training methods. This shows that AI can learn much more effectively in these dynamically generated environments.

    The project also highlights a shift in AI research focus to varied environments like stores, public spaces, and offices. By broadening the scope of training environments, AI can adapt to more complex and varied tasks.

    The connection between this groundbreaking AI technology and Star Trek’s Holodeck lies in the core concept of creating immersive, interactive 3D environments on demand. Just as the Holodeck allowed the crew of the U.S.S. Enterprise to step into any scenario crafted by their commands, this new system enables users to generate detailed virtual worlds through simple linguistic prompts.

    This technology mimics the Holodeck’s ability to create and manipulate spaces that are not only visually accurate but also interactable, providing a seamless blend of fiction and functionality that was once only imaginable in the realm of sci-fi.

     WTF fun facts

    Source: “Star Trek’s Holodeck recreated using ChatGPT and video game assets” — ScienceDaily

    [ad_2]

    WTF

    Source link

  • Backed by Cresta founders, Trove's AI wants to make surveys fun again | TechCrunch

    Backed by Cresta founders, Trove's AI wants to make surveys fun again | TechCrunch

    [ad_1]

    Surveys have become an integral part of many aspects of our lives, but most of them are tedious, leading to ineffective responses and actions. Dinghan Shen and Yuan Xue, two software engineers working in Silicon Valley, recognized an opportunity to leverage the breakthroughs brought by large language models to make surveys more empathetic and engaging.

    Around six months ago, Shen and Xue, who had been friends since high school, started Trove AI, a SaaS platform that lets users create conversational surveys powered by GPT-4 and its own fine-tuned models. The idea has received backing already. Zayd Enam and Tim Shi, co-founders of Shen’s former employer Cresta, an a16z-backed unicorn empowering contact center agents with AI, invested an undisclosed amount in the startup’s pre-seed funding.

    Launched six weeks ago, Trove’s first version has amassed over 1,000 users who are mostly small and medium-sized businesses from around the world. “Dozens of” them have sent surveys at least twice since. Still free to use, the platform has attracted a wide range of users, including a London-based spa, a K-12 school in Boston, and a travel agency focused on Latin America.

    Applying conversational AI to surveys appears to be a low-hanging fruit in this era of ChatGPT frenzy. Enterprise-focused survey giant Qualtrics has adopted AI across its line of customer and employee feedback products. SurveyMonkey, the go-to survey provider for SMBs, is using AI to automate the creation process.

    To differentiate, Trove aims to ultimately become a “customer and employee experience management platform” for companies of all sizes, Shen said. The product is essentially experience management around customers, employees, products, and more. Following the recent management saga of OpenAI that briefly disrupted its chatbot service, applications that build on top of ChatGPT, or “wrapper products,” are rethinking their heavy dependence on third-party APIs.

    “We are 80% SaaS and 20% AI,” Shen told TechCrunch in an interview. As such, he reckoned Trove offers ample value in addition to its features powered by OpenAI. “We aim to do everything from survey creation, response, analytics, ticket creation to CRM integration… It’s an AI-generated feedback loop.”

    CRM integration, specifically, would allow Trove to create highly customized surveys upon which the system can create a ticket automatically and send a personalized follow-up email to thank the customer for giving feedback.

    “Fundamentally, we’re rethinking the experience management workflows from scratch in the context of the powerful large language model capabilities today,” said the founder.

     

    [ad_2]

    Rita Liao

    Source link