OpenAI Takes Aim at Google with SearchGPT
OpenAI has launched SearchGPT, an AI-powered search engine designed to provide direct answers to queries using information drawn from the internet. Similar in format to ChatGPT, SearchGPT presents information, images, and source links, allowing for follow-up questions and related searches. The platform incorporates location data for localized results and is powered by OpenAI’s advanced language models.
While SearchGPT launches in beta for a limited audience, OpenAI aims to integrate some features into ChatGPT eventually. This move positions OpenAI as a direct competitor to established search engines like Google, particularly with its emphasis on responsible AI practices, including clear attribution and publisher collaboration. However, the launch comes amidst concerns about plagiarism, inaccuracies, and potential harm to content creators, issues that have plagued other AI search tools.
Google Expands Access and Capabilities of Gemini Chatbot
In a bid to compete with other AI chatbots, Google is significantly upgrading Gemini, its free AI chatbot. Now available in 40 languages and nearly 230 countries, Gemini boasts faster performance, particularly in reasoning and image understanding. Users will benefit from an expanded context window, allowing Gemini to process and recall more information. Additionally, all users will soon be able to upload files for analysis, a feature previously reserved for paying subscribers. To combat inaccuracies, Gemini will now display links to related web content, offering transparency into the chatbot’s information sources. Google is also expanding Gemini’s availability in its Messages app and launching a teen-focused experience globally, accompanied by enhanced safety measures and an AI literacy guide.
Microsoft Challenges OpenAI with Serverless Fine-Tuning for Phi-3
Microsoft is stepping up its AI game with a new feature for its Phi-3 small language model: serverless fine-tuning. This allows developers to customize Phi-3 for specific tasks without managing their own servers, a process that is initially free. Microsoft highlights Phi-3’s affordability and suitability for various applications, particularly in enterprise settings where safety and customization are crucial. This move directly challenges OpenAI’s dominance in the AI development sphere, especially as both companies court developers with competitive pricing for fine-tuning capabilities. With the emergence of other models like Meta’s Llama 3.1 and Mistral’s Large 2, the competition to provide the most compelling and accessible AI tools is intensifying.
Airtable Reimagines App Development with AI-Powered Cobuilder
Airtable’s new tool, Cobuilder, uses AI to build custom business applications in seconds, potentially revolutionizing enterprise software development. By inputting natural language prompts, even non-technical users can create apps, eliminating the need for traditional coding and bridging the gap between business needs and technical execution. While currently reliant on public data, Airtable plans to integrate company-specific data and AI automations in future updates, signaling a bold vision for the future of no-code app development.
Runway Under Fire for Allegedly Using YouTube Videos to Train AI
Runway, an AI video startup, is facing criticism after a report alleged the company used thousands of YouTube videos, including content from major studios and popular YouTubers, to train its AI model without permission. While Runway hasn’t confirmed the authenticity of the leaked spreadsheet detailing the videos, the news has sparked outrage among creators, some of whom are already engaged in lawsuits against AI companies over similar data-scraping practices. This incident highlights the growing tension between the advancement of AI technology and artists’ rights in the digital age.
