Global Readiness: An Exclusive Conversation with OpenAI’s Product Lead, Thibault Sottiaux

Thibault Sottiaux, a notable figure for users of OpenAI’s Codex as the person who manages token limits, is now spearheading a new initiative called ChatGPT Work. This platform aims to empower white-collar professionals by harnessing AI agents. I had the chance to discuss this product and its associated challenges with Sottiaux, and I’m excited to share a condensed version of our conversation, focusing on how to win over skeptics, the philosophy of product discovery, and the financial implications of AI.

Interviewer: As a key member of the technical team, can you clarify your role with Codex and ChatGPT Work?

Sottiaux: I oversee all core products, which include API services, agent infrastructure, and all aspects of ChatGPT, such as ChatGPT Work and the classic version, in addition to everything related to Codex.

Interviewer: Do you report directly to Greg Brockman?

Sottiaux: Yes, indeed. In essence, everyone reports to Greg!

Interviewer: Why is the launch of ChatGPT Work significant for OpenAI?

Sottiaux: We aim to democratize access to coding agents. Our goal is to transform a tool originally designed for technical users into something that is user-friendly and accessible on various platforms, whether mobile or web. That’s why we’ve made it part of the Plus plan, which is very affordable at $20 a month. The value it provides is substantial, and we want to extend this technology to as many people as possible.

Interviewer: Analysts suggest there are economic incentives for OpenAI to maintain a close relationship with users.

Sottiaux: Absolutely. The more value we create for users, the more likely they are to pay for that service. We’ve designed ChatGPT to offer significant utility, making it an easy decision for users to invest in it.

Interviewer: What role could this technology play in shaping public perception of AI?

Sottiaux: We see it as our responsibility to help everyone adapt to this technology. Initially, Codex was tailored for a more forgiving technical audience, but now we believe it’s the right time to share these advancements with a wider group. ChatGPT can manage complex tasks autonomously, enhancing usability while prioritizing safety. Our mission is to bring everyone along on this journey.

Interviewer: How do you navigate the balance between letting the model take charge versus providing user interface options?

Sottiaux: Our focus is on creating highly capable models and finding the simplest, most delightful ways for users to engage with them. This often means stepping back and allowing the model to showcase its capabilities without unnecessary friction. The introduction of ChatGPT Voice highlights this shift towards more natural interactions.

Interviewer: Is the model advanced enough to provide a seamless experience? Some users feel overwhelmed by the range of options available.

Sottiaux: We believe the market is ready for this kind of advanced technology. Our user base has expanded significantly, reaching 20 million users, indicating strong demand for a simple yet powerful tool.

Interviewer: When creating a versatile product, how do you prioritize features or identify essential problems to solve?

Sottiaux: Developing this product involves exploratory discovery. As we expand the capabilities of our models, we learn what they can achieve and focus on enhancing those aspects, creating exceptional tools around them. Continuous improvement through user feedback is integral to this process.

Interviewer: There seems to be a stark contrast between the subscription cost and token usage. Should users be concerned?

Sottiaux: We are committed to improving efficiency. We’ve recently implemented significant price reductions, ensuring that advanced capabilities will become increasingly affordable over time. We aim to deliver enhanced value without escalating costs for our users.

Interviewer: For those anxious about granting access to personal emails or messages, what reassurance can you provide?

Sottiaux: Choosing safe and aligned models is crucial. We invest heavily in safety measures and maintain transparency in benchmarks. Our models excel in these areas, providing a level of security that users can trust.