GPT4.5 Turbo Leaked, EU Passes AI Regulation Act, GPT Gets a Body, & More!
Welcome to the AI Search newsletter. Here are the top updates in AI this week.
GPT-4.5 Turbo Leaked on Search Engines
The teaser text reveals that GPT-4.5 Turbo is the fastest, most accurate, and most scalable model from OpenAI to date, boasting a context window of 256,000 tokens - double that of GPT-4 Turbo. It is expected to be up-to-date until June 2024, hinting at a potential launch date. Speculations about the new model, including possible video or 3D capabilities, have been circulating since December.
GPT Gets a Body: Humanoid Robot Unveiled
Figure 01 is a humanoid robot created by robotics company Figure AI and OpenAI, allowing it to engage in detailed conversations, plan actions, and carry them out independently. The robot's capabilities include describing its surroundings, planning future actions, reflecting on memories, and explaining its conclusions verbally. Using a multimodal model developed by OpenAI, the robot processes its entire conversation history to generate spoken responses and determine the appropriate learned behavior to execute a command.
EU Passes First AI Regulation Act
The act, approved with a resounding majority in the EU parliament, aims to position Europe as a global leader in shaping ethical and operational standards for AI technologies. The EU Act divides AI technologies into different risk levels, ensuring alignment with societal values and ethical standards. Prohibited AI systems under the act include exploitation of vulnerabilities, biometric categorization, manipulative practices, real-time biometric identification, and more. Civil rights advocates and data rights activists have lauded the act for addressing major concerns and ushering in a new era of AI regulation.
Murf AI
Introducing Murf, the most versatile AI text-to-speech generator. Create studio-quality voice overs in minutes using lifelike AI voices suitable for podcasts, videos, presentations, and more. Choose from 120+ text to speech voices in 20+ languages, with the ability to add video, music, or images and sync them to the voiceover.
MOSAIC, Your AI Cooking Assistant
Researchers at Cornell University have developed MOSAIC, a modular system that enables home robots to engage in interactive cooking and assist humans with complex household tasks. This system was tested in real-world experiments using different robotic systems and successfully completed a variety of recipes in collaboration with a human user.
Will This AI Replace Software Engineers?
Introducing Devin, the AI software engineer created by Cognition Labs. Devin can code, create websites, and software with just one prompt. The AI tool has passed engineering interviews from top AI companies and can be accessed early by submitting a form on Cognition's official website. Devin's standout features include its ability to plan complex tasks, make decisions, learn from mistakes, and improve over time.
Luminar NEO
Luminar Neo is a new AI-driven creative image editor developed by Skylum Software. It’s designed to make complex editing quick and easy for all levels of photographers, from beginners to pros. Luminar Neo leverages artificial intelligence and 3D depth mapping, offering innovative tools such as the new Portrait Background Removal AI, Mask AI, and the Relight AI tool with 3D Depth Mapping. It can function as a standalone application for macOS and Windows, and can also integrate with Lightroom Classic, Photoshop, Photos for macOS, and Microsoft Photos as a plugin and extension. With Luminar Neo, you can transform your photos into the images you imagined, making creative image editing accessible and fun.
Introducing SIMA, a Scalable Multiworld Agent
SIMA, from Google DeepMind, has been trained on a diverse range of video games to demonstrate its ability to follow instructions and complete tasks within these virtual worlds. By utilizing language and image inputs, SIMA can interact with different gaming environments without needing access to the game's source code. The current version of SIMA has mastered 600 basic skills and is continually being developed to tackle more complex tasks, ultimately aiming to become a helpful and capable AI system for a wide range of applications.
LLMs Think in English, Even When Prompted in Another Language
Researchers at EPFL have discovered that even when large language models are prompted in languages other than English, they primarily use English internally. This finding raises concerns about linguistic and cultural bias as AI becomes more prevalent in our daily lives. The study focused on the Llama-2 model and found that the model's internal processing heavily relies on English, even when translating between other languages. This English dominance has implications for how language models perceive and interact with the world, potentially leading to biases in their responses.