AI Roundup - Monday, January 22nd 2024
Democratic inputs to AI grant program: lessons learned and implementation plans
OpenAI has announced the completion of its first pilot program, AI for Good, which aimed to support projects that utilize artificial intelligence to promote democratic governance and inclusion. The program received nearly 1,000 applications from 113 countries but selected only 10 teams. The chosen teams encompass a diverse range of backgrounds and expertise, including law, journalism, peace-building, machine learning, and social science research. Throughout the program, the teams received guidance and support from OpenAI. They were encouraged to document their processes and collaborate with other teams. The teams showcased their concepts at a Demo Day in September, where they presented projects such as video deliberation interfaces, crowdsourced audits of AI models, mathematical formulations of representation guarantees, and methods to fine-tune model behavior. The teams' work, including the code they developed, is now available for the public to access.
How OpenAI is approaching 2024 worldwide elections
Ensuring accurate voting information: Google aims to elevate reliable information about voting, such as registration and voting location details, to combat any misinformation or confusion that may arise during elections. They will work to make this information easily accessible to users. Enforcing measured policies: Google will continue to enforce its policies to prevent any abuse or manipulation of its platforms during the election process. This includes taking action against any content or ads that violate their policies, such as spreading misinformation or attempting to suppress voter turnout. Improving transparency: The company recognizes the importance of transparency in maintaining trust during elections. They will provide more visibility into their political ads library and disclose who is paying for ads related to elections, forming a key part of their effort to prevent foreign interference. Cross-functional effort: Google has formed a dedicated team with expertise from various departments to address potential abuse and threats during elections. This team includes members from safety systems, threat intelligence, legal, engineering, and policy teams, who will collaborate to quickly investigate and tackle any issues that may arise. As elections in 2024 approach, Google is committed to continuously evolving its approach to safeguard the democratic process and ensure the integrity of elections worldwide.
Introducing the GPT Store
OpenAI has unveiled its latest offering, the ChatGPT Team plan, catering to teams of all sizes. The plan provides team customers with access to a private section of the GPT Store, where securely published GPTs are available within the workspace. Although the GPT Store will soon be accessible to ChatGPT Enterprise customers, it will come with enhanced administrative controls. These controls will enable businesses to choose how internal-only GPTs are shared and determine which external GPTs can be utilized within their organization. Importantly, similar to other usage on ChatGPT Team and Enterprise, OpenAI assures that user conversations with GPTs will not be used to improve the AI models. With this new plan, OpenAI aims to empower teams with the collaborative capabilities of ChatGPT while maintaining privacy and control over the AI interactions.