AI Experts Fear Extinction, Mini LLM 'TinyLlama' Outperforms, Google's AI Image Generator, & More
Welcome to the AI Search newsletter. Here are the top updates in AI this week.
New Mini AI Model 'TinyLlama' Packs a Powerful Punch
Singapore University of Technology and Design (SUTD) has introduced a groundbreaking mini AI model called TinyLlama. TinyLlama is a small language model with 1.1 billion parameters that has outperformed other models of similar sizes in various benchmarks. Despite its compact size, TinyLlama has been trained on three trillion tokens of data and has shown exceptional performance in common-sense reasoning and problem-solving tasks. Unlike larger language models, TinyLlama requires fewer computational resources and can be deployed on mobile devices. Its open-source nature makes it accessible to smaller tech companies and research labs, allowing them to develop their own models for different applications.
Imagen 2 is now available across Bard, ImageFX, SGE and Vertex AI
Imagen 2, powered by Google DeepMind's latest advancements in text-to-image, offers high-quality and photorealistic image outputs. It improves on previous systems by rendering realistic hands and human faces and reducing visual artifacts. Imagen 2 generates more detailed images aligned with language prompts, making it capable of capturing nuance and delivering photorealistic images across various styles and applications. The technology is now available in Bard, ImageFX, Search Generative Experience (SGE), and Google Cloud's Vertex AI. Google has also introduced a new experiment called ImageFX, allowing users to create and modify images using text prompts and expressive chips.
Free AI Tool to Upscale, Unblur, Sharpen Images
Here's a 100% FREE AI that can upscale, sharpen, add details, and compress images. Check it out:
Survey Shows AI Experts Believe in Positive Advances, But Fear Catastrophic Consequences
A recent survey conducted by AI Impacts, which examines the long-term effects of artificial intelligence, asked nearly 3,000 machine learning experts about their predictions for how AI will shape our lives. The results revealed a majority of respondents believe that AI will lead to significant advancements, and that these advancements will happen sooner than previously thought. However, the survey also found that between 38% and 51% of respondents expressed concern over the potential of an AI-triggered extinction scenario, with nearly 60% believing there is at least a 1 in 20 chance of such an event occurring. Despite this worry, 70% of experts overall believe that AI will have more positive outcomes than negative as it becomes more advanced.
Italy Accuses OpenAI of Violating EU Laws
Italy accuses OpenAI of violating EU data protection law and gives the company 30 days to respond. Italy's data protection watchdog has notified OpenAI of breaches in data protection law related to ChatGPT. The watchdog had previously blocked the chatbot for several weeks, making it the first Western country to take such action. The watchdog concluded that there were breaches of the EU GDPR, and OpenAI has the opportunity to submit counterclaims in the next 30 days. The final determination in this case will take into account the work of a task force established by the EU's central data regulator. Last year, the Italian watchdog found that OpenAI had no legal basis to justify the collection and storage of personal data for training the algorithms of ChatGPT, and there was a lack of clarity on whose data was being collected.
Flexispot
Elevate your home office with the finest, ergonomic standing desks and chairs, designed to promote good posture and reduce the risk of back and neck pain.
New Soft Artificial Muscles for Robot Movement
Soft artificial muscles created for robot movement offer advantages over previous technologies by providing flexibility and improved sensitivity. Developed by researchers at ETH Zurich, the artificial muscles can be used in scenarios where robots require softness and adaptability, allowing them to interact with their environment in a different way. These muscles function similarly to human muscles, contracting in response to electrical impulses. The new artificial muscles, called HALVE actuators, feature a specialized shell made of high-permittivity ferroelectric material and electrodes, which reduces the required voltage for operation. The researchers have successfully built and tested these muscles in robotic grippers and fish-like robots.
💡Prompt Tip of the Week💡
For prompts involving multiple steps, specifying a “role” for ChatGPT can be very effective. For example: “You are BloggerGPT. Your role is to write high-quality, SEO-friendly blogs on [insert topic]. Do you understand?”
Promoting Sustainability through Advanced AI in Smart Cities
The article discusses the integration of Artificial Intelligence of Things (AIoT) technologies in smarter eco-cities to address environmental challenges. These cities leverage real-time data and optimize resource utilization to create sustainable and resilient urban ecosystems. The article presents insights and perspectives on the emergence of smarter eco-cities and provides valuable knowledge for policymakers, practitioners, and researchers.
Preventing Facial Recognition Using Cameras
Researchers at Zhejiang University have developed a new technique called CamPro that aims to address privacy concerns related to facial recognition. Unlike traditional methods that modify images after they are taken, CamPro operates at the camera sensor level, making it harder for malicious users to bypass. By leveraging the tunable parameters of the image signal processor, CamPro can protect users' facial privacy while still allowing for other visual recognition applications. The researchers aim to increase the privacy of users by obfuscating, synthesizing, or altering images using the field of research known as anti-facial recognition (AFR).