Welcome back to AI Hungry, where we dive straight into the latest developments that are shaping the future of artificial intelligence. Google is making waves with the release of Gemma, their new open source language models that promise to raise the bar in AI capabilities. Meanwhile, they've hit a snag with their Gemini AI model, as it grapples with historical inaccuracy and bias in image generation, sparking a heated debate and necessitating a pause on a key feature.
These stories not only highlight the rapid advancements in AI technology but also underscore the complex challenges that come with it. From technical breakthroughs to ethical considerations, join us as we explore the intricate landscape of AI through Google's latest endeavors.
Google Unveils Gemma: Cutting-Edge Open Source Language Models
Google has introduced Gemma, a new family of open source language models, challenging competitors like Meta's Llama 2 and Mistral with advanced capabilities. Gemma, derived from Google's powerful Gemini model, comes in two versions: a 2-billion parameter model designed for CPUs and a more robust 7-billion parameter model for GPUs and TPUs. Both models are touted to excel in dialogue, reasoning, math, and coding tasks, outperforming or matching their rivals.
In addition to their technical prowess, Google's Gemma models are accessible for commercial use and offer integration with various platforms and tools. The company is also committed to responsible AI use, providing a toolkit for safer applications and embedding safety measures within Gemma. Google encourages a collaborative approach to managing the risks and benefits of open source models, despite concerns about potential misuse by bad actors.
Google Pauses AI Image Feature Amid Criticism of Historical Inaccuracy and Bias
Google's Gemini AI model has faced backlash for producing images with perceived anti-white bias, prompting the tech giant to halt its people-generating feature. Users reported that the AI was creating images of diverse ethnicities in historically inaccurate contexts, such as depicting Vikings as black individuals and the founding fathers as Indigenous people. Critics, particularly from right-wing circles, have accused Google of harboring an anti-white bias. However, experts argue that the issue reflects the limitations of AI rather than intentional bias, with AI struggling to understand historical context.
Google, which launched Gemini as a competitor to OpenAI's GPT model, is working on improving the model's accuracy. The AI's inability to distinguish between historical and contemporary prompts has led to errors, highlighting the challenges AI faces in balancing representation and diversity without bias. The company is committed to refining its AI to better handle nuanced historical contexts while maintaining a diverse representation for its global user base.
💡 SoftBank's Masayoshi Son Eyes $100 Billion for New AI Chip Venture. SoftBank's Masayoshi Son is planning to raise $100 billion for a new AI chip company, aiming to compete with NVIDIA. The project, named after a creation deity, seeks to address chip shortages and advance AI, but details remain uncertain. (Link)
🧠 Neuralink's First Human Patient Successfully Controls Computer Mouse with Mind. Elon Musk's Neuralink has made a breakthrough with its first human patient, who can now control a computer mouse using thought alone after recovering from a chip implant surgery. (Link)
🚀 Stability AI Unveils Stable Diffusion 3 with Enhanced Image and Text Generation Capabilities. Stability AI has announced Stable Diffusion 3, boasting improved image quality and complex prompt handling. The model ranges from 800 million to 8 billion parameters, integrating new research like Diffusion Transformer Architecture. Safety measures have been implemented to prevent misuse. (Link)
🤖 Reddit Inks $60M Deal to Train AI, Revealing Google as Mystery Partner. Reddit has entered into a $60 million deal with Google to use its content for AI training, highlighting the growing value of social media data for AI development and the potential rise in licensing costs for such data. (Link)
🌐 Google Unveils 'Gemini Business' for Enhanced AI in Workspace. Google launches 'Gemini Business' for Workspace, offering AI-powered tools like writing assistance and image generation for businesses. A new AI Premium Plan for consumers includes advanced AI features across Google apps for $19.99/month. (Link)
🚀 Worldcoin Cryptocurrency Surges Amid AI Hype and Crypto Market Growth. Worldcoin's value has nearly tripled this week, reaching close to a $1 billion market cap, following OpenAI's release of the AI tool Sora and the World App's milestone of 1 million daily users. (Link)
