OpenAI Enhances ChatGPT with Web Search Feature
OpenAI has recently introduced a web search capability to its AI-driven chatbot, ChatGPT. This addition enables ChatGPT to bridge the gap with competitors such as Microsoft Copilot and Google Gemini. Users now have the option to manually activate the search function or have it triggered by specific queries, allowing for real-time access to web-based information during conversations.
The search feature within ChatGPT is powered by a blend of technologies, including Microsoft's Bing, and is seamlessly integrated across all ChatGPT platforms. It is underpinned by a refined version of GPT-4o. Despite this enhancement, OpenAI remains committed to consistently updating its training data to ensure users benefit from the latest advancements in AI technology.
Physical Intelligence Pushes AI Boundaries with Household Chores
Physical Intelligence, a startup based in San Francisco, is pioneering the development of an artificial intelligence model designed to handle a wide array of household tasks. By leveraging extensive data sets for AI model training, the company aims to transition the concept of domestic robots from the realm of science fiction to practical reality.
The approach taken by Physical Intelligence mirrors the methodology used in creating large language models (LLMs) for chatbots. The ultimate objective is to devise a versatile learning algorithm tailored for real-world applications, capable of executing diverse tasks across various robotic platforms. CEO Karol Hausman draws parallels between training language models and their approach for developing a general-purpose AI algorithm, emphasizing its adaptability and capacity to leverage data from different robotic frameworks.
Decart's Oasis: A Revolutionary "Open-World" AI Model
Israeli AI firm Decart has launched Oasis, an innovative "open-world" AI model that simulates a dynamic, playable rendition of Minecraft in real-time. The model, trained on actual Minecraft gameplay videos, generates frames instantaneously based on user keyboard and mouse inputs, thereby recreating the game's physics, regulations, and visual elements.
Despite current limitations such as low resolution and occasional lapses in level retention, Decart is actively working on enhancements for Oasis. This includes the ability to construct custom "worlds" from uploaded images. Future iterations of Oasis, optimized for Etched's forthcoming AI accelerator chips, have the potential to produce gameplay in up to 4K resolution. However, questions regarding copyright implications have surfaced due to Decart's omission of details regarding permission from Microsoft for training the model on Minecraft content.
