Former OpenAI Researcher: The Need to Protect Fragile but Highly Valuable Ideas

At a conversation organized by Tencent Reading, Yuan Xiaohui, a senior expert at Tencent Research Institute, had a two-hour conversation with former OpenAI researchers Kenneth Stanley and Joel Lehman around 10 questions. When asked about consensus, Kenneth Stanley said, "Sometimes you do need consensus, but sometimes you need disagreement. Innovative organizations are inherently complex." However, he also added, "You need to protect ideas that are fragile but highly valuable, especially if they come from the top talent you've hired." He sees innovation as a cyclical process, with phases of disagreement and phases of convergence, as some projects do require multiple people to collaborate in order to complete them. Even if there are times when the team needs to agree, it doesn't have to be on the same goal. And with respect to OpenAI's success, Kenneth Stanley says it's about the interest and the courage of the leadership who were willing to really put all their bets on something that seemed so risky, and Joel Lehman gives pretty much the same view, and he also says that OpenAI has very smart people and a willingness to explore. It took a lot of courage to make that kind of unconventional bet from the beginning and to put all their chips on the line at once. So when people try to emulate OpenAI, sometimes they need to make a big bet that no one wants to make, but there's no guarantee that it's going to work.