Here are a few of the headlines that caught our eye in the world of AI in March.
Nvidia’s share price fell after DeepSeek’s rise but soon rebounded, backed by $39.33 billion in revenue, exceeding forecasts. Nevertheless, new chip tech from Meta, Google, and Microsoft, along with efficient algorithms like DeepSeek’s R1, pose a threat. Nvidia also launched “Isaac GROOT N1,” an open-source AI model for more adaptable humanoid robots. CEO Jensen Huang calls it the start of “generalist robotics.”
Amazon’s upcoming ‘AlexaPlus’ will feature built-in AI. The demo showed it assisting with grocery orders, controlling smart home devices, and checking home activity via security camera footage.
OpenAI’s new AI model, GPT-4.5 (Orion), has been released. Reports tell us that it excels in factual accuracy but struggles with reasoning. GPT-5 aims to enhance both general AI and reasoning skills. We also expect ChatGPT users will soon get access to OpenAI’s AI video tool Sora.
Microsoft have launched Dragon Copilot, an AI assistant for doctors, based on Nuance technology acquired in 2021. It offers:
- Real-time transcription: Summarises clinical visits.
- Task automation: Manages referral letters, visit summaries, and documentation.
- Medical insights: Provides information from trusted sources.
Fact-checking initiatives such as “YesNoError” and the “Black Spatula Project” are in progress. “YesNoError” has scanned over 37,000 papers in two months, identifying potential mistakes that have yet to be confirmed. The “Black Spatula Project,” which has analysed 500 papers so far, notifies authors directly of possible errors rather than making them public.
Google has unveiled a new AI feature for Gemini. It uses your phone’s camera in real-time to answer questions about anything it sees, from selecting paint colours to identifying items on your screen. Powered by “Project Astra,” which debuted nearly a year ago, these updates are now available for Gemini Advanced users on the Google One AI Premium plan.
Also look out for “AI Overviews” in Google Search. You should start to see a tab on the search page called ’AI-Mode” although its only currently available to subscribers of Google One AI premium.
Introducing 4o Image Generation OpenAI has just introduced a powerful new image generation capability natively integrated into ChatGPT, leveraging the advanced GPT-4o model. This upgrade allows users to generate precise, accurate, and photorealistic images directly within the ChatGPT interface. Unlike previous versions that relied on external models like DALL-E, the new native generator ensures a more consistent and seamless experience. It excels in rendering text, maintaining character consistency, and creating detailed diagrams, making it a practical tool for various applications. This enhancement is available to all ChatGPT users, including those on free plans, though with some usage limitations.