Sam Altman removed TikTok because he couldn't resist scrolling. Similarly, OpenAI discontinued Sora for the same reason.

      Sam Altman mentioned that he became “addicted” to TikTok while researching it for Sora. He ultimately deleted the app after spending three hours on it one Saturday. OpenAI discontinued Sora in March, stating that ChatGPT is created to avoid trapping users.

      Altman revealed that he removed TikTok because he couldn't limit his usage. On the “Relentless” podcast, the OpenAI CEO explained that he compelled himself to learn the app while OpenAI was developing Sora, its AI-driven video social network. “I forced myself to get hooked on TikTok because I wanted to understand it,” he recounted. His use began with 10 minutes before bedtime, then increased to an hour, and finally, three hours one Saturday afternoon. “I was enjoying it in the moment like a drug,” he admitted, leading him to delete the app.

      OpenAI also discontinued Sora, shutting it down in March after approximately six months. The decision stemmed from the high computational demands of video generation, particularly when competitors like Anthropic were advancing in the AI race. Sora had secured a partnership with Disney and gained significant attention, but the financial viability was not there. The comparison is stark: Altman found TikTok too engaging to keep on his phone, just as OpenAI deemed Sora too resource-intensive to maintain.

      Altman reflected on this experience in relation to the AI device being developed by Jony Ive, which OpenAI plans to reveal before the year's end. He expressed a desire for the device to alleviate the constant barrage of notifications that make phones feel chaotic, like “walking through Times Square.” In contrast, Google revamped its Search to include AI summaries that encourage longer user engagement. Altman advocates for OpenAI to pursue the opposite approach: creating a product that users engage with briefly and then leave. “We focus on whether users leave having accomplished what they set out to do, rather than measuring success by time spent or clicks,” OpenAI stated last August.

      This admission is significant as it comes from someone who is developing the AI systems that will shape future recommendation algorithms. Altman personally experienced how TikTok’s algorithm is, in his words, “too powerful,” and acknowledged that even he could not resist it. Research from Wharton indicated that people tend to accept AI outputs 80% of the time, even when incorrect. TikTok's algorithm, however, is not incorrect; it expertly identifies what keeps users engaged. Even the CEO of the world's leading AI firm could not outwit it, marking it as one of the most candid product reviews TikTok has received.