A new #1 AI model / AI creates 3D objects from only shadows / Using AI for standup comedy / AI news this week
Welcome to the AI Search newsletter. Here are the top highlights in AI this week.
New & Free AI that can Animate Any Photo
Hedra is a new AI tool that allows users to make any photo speak. It can generate highly realistic and expressive characters, with features like lip-syncing and movement. The tool is currently in its research preview phase and is free to use, with the company planning to add more advanced features in the future.
Introducing Claude 3.5 Sonnet: A Revolutionary AI Model that Beats All
Anthropic releases Claude 3.5 Sonnet, an AI model that outperforms its predecessors and competitors in various evaluations, including graduate-level reasoning, undergraduate-level knowledge, and coding proficiency. This new model operates at twice the speed of its predecessor, Claude 3 Opus, making it ideal for complex tasks like context-sensitive customer support and multi-step workflows. Additionally, Claude 3.5 Sonnet excels in vision tasks, such as interpreting charts and graphs, and can even transcribe text from imperfect images. The model is now available for free on Claude.ai and the Claude iOS app, with expanded features like Artifacts, which enables users to interact with Claude in a more collaborative and dynamic workspace.
OpenAI Co-Founder Ilya Sutskever Launches New AI Venture Focused on Safe Superintelligence
Ilya Sutskever, co-founder of OpenAI, has started a new AI company called Safe Superintelligence, aiming to address the risks associated with advanced AI systems. The company's mission is to develop AI that is not only highly intelligent but also safe and beneficial for humanity. Sutskever, a prominent figure in the AI research community, believes that current AI systems are not designed with safety in mind and that his new venture will focus on creating AI that is more transparent, explainable, and controllable.
ChatLLM by Abacus
ChatLLM by Abacus AI is an integrated platform that allows enterprises to use multiple LLMs, deploy custom agents, and collaborate with team members. Choose from state-of-the-art LLMs such as GPT-4o, Claude 3 Opus, and their new open source Smaug.
Nvidia Becomes World's Most Valuable Company with $3.34 Trillion Market Cap
Nvidia has surpassed Microsoft and Apple to become the world's most valuable public company, with a market capitalization of $3.34 trillion. The chip maker's stock price has surged over the past year, driven by the boom in AI and its dominant position in the market for computing used in AI systems. Here’s the full breakdown:
LLMs Struggle to Write Funny Jokes, Stand-up Comedians Find
Google's DeepMind tested the ability of LLMs to write jokes and entire stand-up comedy routines. They asked 20 professional stand-up comedians to use LLMs to generate jokes and then rate the results. While the LLMs were able to come up with many jokes, few of them were considered funny by the comedians. The jokes were often generic and lacked the surprise and edge that is typically needed to make a joke humorous. However, some comedians found the LLMs to be useful in generating a basic structure for their routines, which they could then build upon.
ALOGIC
ALOGIC offers a wide range of professional home/office hardware, including their Clarity series of professional 4K touchscreen monitors, wireless chargers, docking stations, and more. Enhance your productivity and elevate your digital lifestyle with ALOGIC!
Researchers Use Shadows to Create 3D Models of Hidden Objects
Researchers from MIT and Meta have developed a computer vision technique called PlatoNeRF that can create physically accurate 3D models of entire scenes, including areas blocked from view, using images from a single camera position. By combining lidar technology with machine learning, PlatoNeRF can generate more accurate reconstructions of 3D geometry than existing AI techniques. The system uses shadows to determine what lies in obstructed portions of the scene, allowing it to infer the geometry of hidden objects.
Humans struggle to distinguish themselves from AI in brief conversations
Researchers conducted a Turing test to evaluate whether people can distinguish between humans and AI models, specifically the GPT-4 model underpinning ChatGPT, in five-minute chat conversations. The results showed that humans found it difficult to tell apart humans from AI models, with GPT-4 successfully passing as human in approximately 50% of interactions. This raises concerns about the potential for AI systems to be used for fraud or misinformation, as people may become increasingly distrustful of others they interact with online.