
OpenAI’s '12 Days of Innovations' Unveils Cutting-Edge Features | Image Source: www.zdnet.com
SAN FRANCISCO, Dec. 13, 2024 — OpenAI is captivating audiences with its ’12 Days of OpenAI’ event series, a unique holiday initiative aimed at showcasing groundbreaking advancements in artificial intelligence. The event, which started on Dec. 5, features daily live streams unveiling a mix of significant innovations and smaller enhancements, according to OpenAI CEO Sam Altman. The campaign has been a whirlwind of announcements, ranging from advanced voice capabilities to revolutionary video generation tools.
Highlights from the Event
OpenAI has structured the campaign to reveal one new feature or development each day. Among the most notable is the update to Advanced Voice Mode, which now includes screen-sharing and visual context capabilities. During a Dec. 12 live stream, OpenAI demonstrated how this feature allows users to integrate verbal instructions with visual inputs, such as guiding a coffee-making process using live feedback from the user’s phone camera or screen. In addition, a festive ‘Santa voice’ was introduced, providing an engaging seasonal touch. Users accessing the Santa voice for the first time also receive a usage limit reset, ensuring a seamless experience.
As per OpenAI, these features are gradually rolling out to mobile app users, with Pro and Plus subscribers receiving priority access. Enterprise and educational users can expect availability early next year. In Europe, rollout efforts are underway, with access coming “as soon as we can,” according to the company.
Innovations in Video and Visual Intelligence
The launch of Sora Turbo, OpenAI’s advanced video generation model, marked another major milestone in the event series. Announced on Dec. 9, the model enables users to create video-to-video and text-to-video content with enhanced efficiency and quality compared to earlier iterations. With resolutions of up to 720p for Pro users, Sora also introduces a community-driven ‘Explore’ page for sharing and reviewing video projects. Additionally, OpenAI showcased Storyboard, a tool that generates input sequences for video frames, further empowering creators.
Visual Intelligence took center stage during the Dec. 11 stream, highlighting the integration of ChatGPT with iOS 18.2. Apple users can now use Siri to access ChatGPT for complex queries, receive contextual assistance via their device’s camera, and utilize enhanced writing tools, including DALL-E-powered image generation. OpenAI emphasized user control and transparency, ensuring that all interactions are subject to explicit permissions before proceeding.
Canvas: A Game-Changer for Productivity
On Dec. 10, OpenAI unveiled Canvas, a productivity tool initially available in beta for ChatGPT Plus users. Integrated directly into GPT-4o, Canvas offers a streamlined interface for project management, featuring a split-screen view for question-and-answer exchanges alongside live project edits. Users can also run Python code directly within Canvas, a boon for developers tackling tasks such as debugging. OpenAI confirmed that Canvas is now accessible to all web users, expanding its reach and functionality.
Reinforcement Fine-Tuning Program
On the second day of the event, Dec. 6, OpenAI expanded its Reinforcement Fine-Tuning Research Program, inviting developers and researchers to customize AI models for domain-specific tasks. This technique leverages feedback loops to refine model outputs, improving accuracy and reasoning. OpenAI is prioritizing applications from organizations that tackle complex problems with objectively correct answers. The program is set to become publicly available in early 2025, further democratizing AI customization.
Launch of Pro Tier and o1 Model Updates
The event kicked off with significant upgrades to OpenAI’s product offerings. The new ChatGPT Pro tier, aimed at power users, grants unlimited access to premium features, including o1-mini, GPT-4o, and Advanced Mode. Priced at $200 per month, it is tailored for professionals requiring high-performance AI solutions. OpenAI also unveiled the full version of its o1 model, boasting faster reasoning and a 34% reduction in major errors compared to its preview version. These enhancements aim to broaden the model’s utility across various domains.
According to OpenAI, the live streams are hosted on its official website and YouTube channel, with updates shared on the company’s X account. Each session begins at 10 a.m. PT and includes a mix of live demonstrations and user testimonials, ensuring an interactive and informative experience.
The event has generated significant buzz, with features like Sora Turbo and Advanced Voice Mode garnering widespread acclaim. As OpenAI continues to unveil its innovations, the series underscores the company’s commitment to pushing the boundaries of AI while engaging users in meaningful ways.