On the first day of its 12-day event, OpenAI introduced two major upgrades for its chatbot, ChatGPT: a new subscription tier called ChatGPT Pro and the full version of its o1 model. Priced at USD 200 per month, ChatGPT Pro offers unlimited access to OpenAI o1, o1-mini, GPT-4o, advanced voice mode, and an enhanced version of o1 known as o1 Pro Mode. The o1 Pro Mode uses more compute to tackle harder problems, providing superior answers, according to OpenAI.
“As AI becomes more advanced, it will solve increasingly complex and critical problems. It also takes significantly more compute to power these capabilities,” ChatGPT shared when introducing the Pro tier.
“OpenAI o1 is more concise in its thinking, resulting in faster response times than o1-preview. Our testing shows that o1 outperforms o1-preview, reducing major errors on difficult real-world questions by 34 percent,” OpenAI announced on X.
Additionally, OpenAI awarded 10 grants of ChatGPT Pro subscriptions to medical researchers at leading US institutions, with plans to expand Pro grants to other regions and research areas in the future.
The company expects to add more powerful, compute-intensive productivity features to this plan in the future.
Day 2: Reinforcement Fine-Tuning
On December 6, the company announced that OpenAI o1 had been fully rolled out to 100 percent of ChatGPT Plus, Team, and Pro users. OpenAI also provided an update on its model customisation offerings, allowing users to fine-tune OpenAI’s models using their own data. The company introduced a new fine-tuning method called reinforcement fine-tuning.
So, day 2 focused on developers, with the expansion of the “Reinforcement Fine-Tuning Research Program.” This enables the creation of domain-specific expert models with minimal training data.
This is a preview release, with a public launch expected in 2025. However, OpenAI expanded alpha access to Reinforcement Fine-Tuning, inviting researchers, universities, and enterprises with complex tasks to apply for early access to the technology.
Also Read: OpenAI and Anduril Partner to Develop AI Solutions for US National Security
Day 3: Sora
On December 9, 2024, OpenAI released Sora, its text-to-video generation model first introduced in February. This model allows users to generate realistic 1080p resolution videos, up to 20 seconds long, in widescreen, vertical, or square formats. Users can upload assets to blend, remix, or create entirely new content from text prompts.
“Sora serves as a foundation for AI that understands and simulates reality—an important step towards developing models that can interact with the physical world,” OpenAI said.
Sora, OpenAI’s video product, is the company’s holiday gift to users. The company also developed new interfaces to make it easier to prompt Sora with text, images, and videos. This includes a storyboard tool for precise frame-by-frame input and integration of text, images, and videos.
OpenAI said Sora is available to users in the US and abroad (with the exception of the United Kingdom, Switzerland, and the European Economic Area) to ChatGPT Pro and Plus subscribers starting immediately. OpenAI also confirmed a significantly faster version of the model called Sora Turbo, released as a standalone product for ChatGPT Plus and Pro subscribers.
Sora is included as part of Plus accounts at no additional cost. Users can generate up to 50 videos at 480p resolution or fewer videos at 720p each month. OpenAI confirmed that the version of Sora currently being deployed has several limitations, such as generating unrealistic physics and struggling with complex actions over long durations.
Additionally, all Sora-generated videos come with C2PA metadata, which will identify a video as coming from Sora to provide transparency and can be used to verify its origin.
On December 10, Sam Altman posted on X that sign-ups would be temporarily disabled, and video generations would be slower due to higher-than-expected demand.
Also Read: OpenAI Raises USD 6.6 Billion to Accelerate AI Research and Expansion
Day 4: Canvas
On Day 4, OpenAI announced several updates to Canvas, a collaborative tool integrated into ChatGPT that provides an editable side panel for writing and coding tasks.
Canvas is a new ChatGPT interface for collaborative writing and coding projects, offering tools for editing, highlighting, and inline feedback. It enables targeted suggestions, version control, and customisable adjustments like changing tone or length, making complex projects easier to refine.
On December 10, OpenAI made Canvas available by default in GPT-4 for all users, both free and paid. The company also announced new capabilities, such as the ability to integrate Python code within Canvas, bring Canvas to custom GPTs, and support shortcuts.
Canvas is available to all ChatGPT users on web and Windows platforms, with upcoming availability on Mac and mobile platforms (iOS, Android, and mobile web).
Day 5: ChatGPT in Apple Intelligence
On Day 5, the live stream featured integration with Apple Intelligence, part of several features launched in iOS 18.2, which enhances Siri with ChatGPT responses across Apple’s platforms. Demos showcased how users can ask ChatGPT through Siri to plan holiday parties, create Christmas playlists, and more. Apple iPhone owners do not need a ChatGPT subscription or account to use the technology.
Also Read: OpenAI Targets 1 Billion Users by 2025 with AI Innovations and Apple Partnership: Report
Day 6: Advanced Voice with Video and Santa Mode
On Day 6, OpenAI announced a major update to its advanced voice mode, a feature enabling users to have natural conversations with a selection of AI voices that sound remarkably human.
Additionally, video and screen sharing, along with image upload capabilities, began rolling out on December 12 to iOS and Android apps, with full availability expected within the next week for all Teams and most Plus subscribers. These features are already available to nearly all Pro subscribers, and OpenAI plans to make them available soon to Pro subscribers in the EU, OpenAI said. Usage of video and screen-sharing capabilities is limited for eligible plans on a daily basis, while image uploads count toward the plan’s usage limits.
OpenAI also announced that Santa Mode is available to all users who have access to voice chats on ChatGPT. For the month of December, users can click on a snowflake icon in the ChatGPT app to begin a conversation with Jolly Old Santa. Now you can chat with ChatGPT over video and voice in real-time and have fun conversations.
Day 7: Projects in ChatGPT
On Day 7, OpenAI introduced Projects, a new interface for ChatGPT that makes it easier to organise chats with shared topics or context in GPT-4. A project keeps chats, files, and custom instructions together in one place, simplifying the process of returning to ongoing work. Users can upload files, set instructions specific to a project, and tailor multiple chats to the project itself. OpenAI confirmed that features like Search, Canvas, Advanced Data Analysis, and DALL-E also work within Projects.
ChatGPT Projects are available to all Plus, Team, and Pro users. Projects will roll out to ChatGPT Enterprise/Edu early next year.
Also Read: Microsoft AI Solutions Drive Transformation for Over 200 Businesses: December 2024 Edition
Day 8: ChatGPT Search
On Day 8, OpenAI brought ChatGPT’s search abilities to all logged-in users in regions where ChatGPT is available. ChatGPT will automatically choose to search the web based on your queries, or users can manually select the option to search by clicking the web search icon. Search is also integrated into advanced voice mode, so users can ask the AI voices to search the web on their behalf.
OpenAI also added maps to ChatGPT in mobile apps, so users can search for and chat about local restaurants and businesses with up-to-date information. ChatGPT’s enhancements, including faster searches and maps on mobile, are available to all users in ChatGPT’s paid tiers.
Users can now search while having a voice conversation with ChatGPT, and during voice chats, ChatGPT can provide answers by searching the web.
Day 9: OpenAI o1 and New Tools for Developers
On Day 9, December 17, OpenAI introduced developer-focused updates. The o1 reasoning model became available through OpenAI’s API, replacing the previously available preview version. Pricing for the company’s real-time API, including simple WebRTC integration for creating high-speed conversational experiences, was reduced by 60 percent. OpenAI also added GPT-4o mini (10x cheaper than previous prices), improved voice quality, and enhanced input reliability.
Preference Fine-Tuning, a new model customisation technique that tailors models based on user and developer preferences, was introduced, along with two new official SDKs: Go and Java SDKs in beta for developers.
Day 10: 1-800-CHATGPT
On Day 10, OpenAI announced an experimental new launch with a toll-free number that anyone can call to interact with ChatGPT. By dialing 1-800-CHAT-GPT (1-800-242-8478), anyone with a US phone number can call the AI model, engage with the advanced voice mode, and discuss anything for up to 15 minutes per month, free of charge. Users can also now start a conversation with ChatGPT through WhatsApp by texting the same number, though daily limits on WhatsApp messages apply.
Day 11: Work with Apps on macOS
On Day 11, December 19, OpenAI announced an update to its computer desktop apps. Users can now direct the ChatGPT app to control other apps on their computers to complete tasks like debugging in terminals, working through documents, analysing data repositories, or getting feedback on speaker notes. Support for additional note-taking and coding apps, including Apple Notes, Notion, Quip, and Warp, was also added.
Also Read: EU Selects Seven Sites for First AI Factories, Marking EUR 1.5 Billion Investment
Day 12: o3 Preview
On the final day of OpenAI’s 12-day series, OpenAI announced its o3 and o3-mini models, the successors to its o1 and o1-mini “reasoning” models.
Sam Altman explained, “Out of respect to our friends at Telefonica, and in the grand tradition of OpenAI being truly bad at names, we decided to skip o2 and name the new model o3.”
“On several of the most challenging frontier evaluations, OpenAI o3 sets new milestones for what’s possible in coding, math, and scientific reasoning. It also makes significant progress on the ARC-AGI evaluation for the first time,” OpenAI shared on X.
The first version of o3 is expected to launch publicly in early 2025. However, OpenAI opened early access applications for safety and security researchers to test these frontier models starting December 21, 2024.
Day 13 of Shipmas
On December 21, Sam Altman announced a special Sora bonus on X, stating, “Our GPUs get a little less busy during late December as people take a break from work, so we are giving all Plus users unlimited Sora access via the relaxed queue over the holidays!”
Also Read: Microsoft CEO Highlights the Human Impact of AI Across Sectors in 2024
OpenAI’s Corporate Structure
OpenAI announced on December 27 that its Board of Directors is evaluating its corporate structure in order to best support the mission of ensuring artificial general intelligence (AGI) benefits all of humanity.
The plan involves transforming the for-profit entity into a Delaware Public Benefit Corporation (PBC), which will allow for conventional capital-raising while balancing shareholder and public interests. This change aims to make the non-profit more sustainable, with its significant interest in the PBC amplifying resources for charitable initiatives. The goal is to evolve OpenAI into a long-term, enduring organisation that continues advancing AGI safely and ethically for the benefit of all.
The PBC will run and control OpenAI’s operations and business, while the non-profit will hire a leadership team and staff to pursue charitable initiatives in sectors such as health care, education, and science, the company said.
“We’ve learned to think of the mission as a continuous objective rather than just building any single system. The world is moving to build out a new infrastructure of energy, land use, chips, datacenters, data, AI models, and AI systems for the 21st century economy. We seek to evolve in order to take the next step in our mission, helping to build the AGI economy and ensuring it benefits humanity,” OpenAI said in a blog post.
Some people read for free. A few choose to support. If you found TelecomTalk useful, you can help keep us running.
Support Telecomtalk
