AI Roundup - Monday, January 15th 2024
Introducing the GPT Store
OpenAI has introduced the ChatGPT Team plan, catering to teams of all sizes. This plan grants team customers exclusive access to a private section of the GPT Store, where GPTs (Generative Pre-trained Transformers) are securely published within their workspace. The GPT Store, which will launch in the near future, will also be available to ChatGPT Enterprise customers. It offers additional administrative controls, allowing businesses to select how internally-focused GPTs are shared and which external GPTs can be utilized within the organization. OpenAI emphasizes that, as with all usage on ChatGPT Team and Enterprise, conversations with GPTs are not used to enhance its models. This announcement demonstrates OpenAI's effort to provide collaborative and secure AI-powered conversation capabilities for teams in various contexts.
Introducing ChatGPT Team
The integration of AI, specifically GPT-4 and ChatGPT, into everyday organizational workflows has been found to significantly enhance productivity and work quality. According to a study by the Harvard Business School, employees at Boston Consulting Group who had access to GPT-4 completed tasks 25% faster and achieved a 40% higher quality level compared to their peers who did not have access. ChatGPT has also been widely utilized in various aspects of business, such as financial modeling, communications, recruiting, and note-taking, by companies like Sourcegraph, resulting in accelerated execution and high-level performance. In healthcare, Boston Children's Hospital has successfully piloted innovative GPT applications with ChatGPT Team, improving productivity and collaboration among their doctors, researchers, students, and administrative staff. The hospital recognizes the transformative impact of integrating GPT technology safely and responsibly into internal operations. ChatGPT Team is available for $25/month per user (billed annually) or $30/month per user (billed monthly), and users can upgrade their settings to access these features.
OpenAI and journalism
In a recent blog post, OpenAI responded to The New York Times' lawsuit over copyright infringement. OpenAI stated that they were disappointed by the lawsuit, as they believed their negotiations with The New York Times were progressing constructively. The discussions focused on a partnership that would allow The New York Times to connect with readers through OpenAI's ChatGPT, while providing OpenAI users access to their reporting. OpenAI explained to The New York Times that their content did not significantly contribute to the training of their models and would not be impactful for future training. The company also expressed their commitment to addressing any issues related to content regurgitation. OpenAI emphasized their seriousness in tackling this issue, citing their immediate action in July when they disabled a feature that reproduced real-time content unintentionally. OpenAI claimed that the regurgitated content The New York Times found was from old articles on third-party websites, suggesting intentional manipulation of prompts by the news organization. OpenAI stated that their models usually do not behave the way The New York Times insinuates, implying that examples may have been cherry-picked or the model was instructed to regurgitate specific content. In conclusion, OpenAI highlighted their continuous efforts to improve their systems' resistance to adversarial attacks and expressed their progress in recent models.