Superpower Daily
Posts
OpenAI Announced Next DevDay

OpenAI Announced Next DevDay

OpenAI won’t watermark ChatGPT text because its users could get caught

Saeed Ezzati
August 06, 2024

In partnership with

In today’s email:

🔥 Figure 02, the most advanced humanoid robot on the planet is coming today
🤔 What do people really ask chatbots? It’s a lot of sex and homework
👀 Tesla Dojo: Elon Musk’s big plan to build an AI supercomputer, explained
🧰 12 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.

OpenAI Announced Next DevDay

This fall, OpenAI is taking DevDay on the road to bring the experience closer to developers worldwide. With events scheduled in San Francisco on October 1, London on October 30, and Singapore on November 21, OpenAI is responding to last year's feedback by expanding their reach and offering more opportunities for developers to connect and learn from one another. These developer events aim to dive deep into new AI features and products, providing a platform for innovation and collaboration.

At OpenAI DevDay, attendees can expect hands-on workshops designed to enhance their skills with the OpenAI platform and API tools. Breakout sessions will cover a wide range of topics, such as model customization, steerability, scaling, and best practices, all led by AI experts and community members. Participants will also have the chance to witness live demos from the OpenAI product and engineering teams, showcasing the latest models and platform capabilities. Developer spotlights will feature groundbreaking projects from the community, and the day will conclude with an evening reception offering dinner, drinks, and entertainment.

As the community eagerly awaits OpenAI's next big model, these events will focus on the latest advancements in the API and developer tools, highlighting the incredible projects built by developers. OpenAI invites interested developers to apply for the limited-capacity in-person events by August 15. Selected applicants will be required to pay a registration fee of $450, with scholarships available for some. OpenAI looks forward to welcoming developers and exploring the future of AI together.

Understand AI’s Risks and Opportunities

With the rapid growth of AI, it's crucial for businesses to understand both the risks and opportunities it presents. Discover how you can utilize AI to drive success while minimizing potential pitfalls.

In the MIT Artificial Intelligence: Implications for Business Strategy online short course you’ll gain:

Practical knowledge and a foundational understanding of AI's current state
The ability to identify and leverage AI opportunities for organizational growth
A focus on the managerial rather than technical aspects of AI to prepare you for strategic decision making

LEARN MORE

OpenAI won’t watermark ChatGPT text because its users could get caught

Some ChatGPT users said they’d use ChatGPT less if it included watermarks.

OpenAI has developed a tool capable of detecting AI-generated text, such as essays and research papers created by ChatGPT, with a remarkable 99.9% accuracy. However, the company has refrained from releasing it due to internal debates over potential impacts, such as its effect on non-native English speakers and the desire to maintain user retention. The watermarking method works by subtly altering the text's token patterns, creating a detectable watermark. Despite its effectiveness, OpenAI is concerned about the watermark being easily bypassed and the balance between providing the tool to educators and preventing misuse by malicious actors.

The decision to release the watermarking tool has been debated within OpenAI for nearly two years, even before ChatGPT's launch. The tool's potential benefits, such as assisting educators in combating AI misuse in academic settings, are weighed against concerns about negative user reactions and possible false accusations of AI use. A survey revealed that 69% of ChatGPT users worry about wrongful accusations, with nearly 30% stating they would use the AI less if watermarking were implemented and competitors did not adopt similar measures.

OpenAI has acknowledged the public's support for an AI detection tool but remains cautious due to mixed user reactions. Despite concerns about degrading ChatGPT's writing quality, internal tests showed that watermarking did not affect performance. OpenAI is exploring alternative approaches and aims to shape public opinion on AI transparency and related laws. The company recognizes that releasing the tool could bolster its credibility as a responsible AI developer, especially amid rising concerns about AI misuse in educational contexts.

Trio of OpenAI Leaders Depart, Take Leave of Absence

John Schulman, a co-founder of OpenAI, has left the company to join Anthropic, an AI startup focused on alignment research. Schulman, who played a key role in developing OpenAI's reinforcement training techniques for models like ChatGPT, expressed his desire to focus more on AI alignment, ensuring AI systems behave as intended. He mentioned that his decision was personal and not due to a lack of support at OpenAI, emphasizing his confidence that OpenAI would continue to thrive without him. Following the departure of Jan Leike, another AI safety researcher who also joined Anthropic, Schulman headed OpenAI's alignment science efforts, contributing significantly to the company's mission of developing safe AI.

In addition to Schulman's departure, OpenAI confirmed that Greg Brockman, the company's president, will take an extended leave until the end of the year. Brockman, who has been with OpenAI for nine years since its inception, stated that he plans to use this time to "relax and recharge" after years of hard work toward building a safe AGI. OpenAI has not yet announced a replacement for Schulman's roles on the alignment science team and the newly formed safety committee. Peter Deng, a product manager who joined OpenAI last year after working at Meta, Uber, and Airtable, also quietly left the company recently, further marking a period of transition for the AI giant.

OpenAI CEO Sam Altman praised Schulman's contributions in a post on X, calling him a brilliant researcher and a great friend while expressing hope that Schulman would continue to be proud of the work being done at OpenAI. Despite recent controversies around OpenAI's approach to AI safety research, the company reaffirmed its commitment to alignment research, highlighting the strong foundation Schulman helped establish. As Schulman moves to Anthropic, only three of OpenAI's 11 original founders remain: Sam Altman, Greg Brockman, and Wojciech Zaremba, who leads language and code generation efforts at the company.