Reflections on AI

Swisher spoke with Altman a good deal about his career as an entrepreneur and his path to becoming the CEO of OpenAI.

The main thing that sticks with me about this episode is that it was my first time hearing directly from Altman his reasons for taking the approach that OpenAI is taking to AI safety. He believes that releasing AI capabilities to the world gradually, as they are developed, is the safest approach because it gives society at large a chance to learn what AI is capable of and adjust. That adjustment takes many forms, e.g. developing new tools on top of AI, but also recognizing the dangers, creating regulations and forming agreements between nations to work together on protecting against the most catastrophic outcomes.

In response to why OpenAI changed from being a non-profit to a for-profit company, Altman stressed that it is a capped-profit company now, owned by a non-profit parent which has authority to restrict its activities. His explanation for this was that they simply needed far more capital (to fund the R&D necessary to build things like GPT-4 and beyond) than a non-profit can realistically ever have access to, unless they are funded by the U.S. government(which he says they tried). Capital like $10 billion from Microsoft, for example.

Swisher asked if Altman had spoken with Elon Musk about this. He said they had spoken, but they still didn’t see eye-to-eye. He did not reveal specifics of what he and Musk actually said to each other.