3/6/2024: OpenAI Planned to Close Source AI Models Long Ago

Their openness claim is dubious

The OpenAI and Elon drama continues. OpenAI published a blog post yesterday recounting its history with Elon. OpenAI said their current peculiar capped profit structure was blessed by Elon. Staying non-profit will be impractical because they needed the money to develop the powerful AI models to counter Google. Elon at the time claimed OpenAI's chance of success would be zero so he let them do whatever they wanted. But now he is regretful because of OpenAI’s amazing success.

Let’s not forget the Anthropic founders left OpenAI because of OpenAI’s pivot from non-profit to capped profit to raise billions from Microsoft. I share Anthropic founders’ concerns about the viability of the new structure and the potential mission drift.  If you tell the world that you are doing this to benefit all humanity but practically you are benefitting Microsofts and other investors first, at what point your claim about benefiting all humanity is just an empty promise. OpenAI also said they can benefit all humanity by not open-sourcing. Open API calls are good enough. Well, in this regard, how is OpenAI different from Google and Anthropic? What exactly is the point of being a non-profit if you have investors who want returns and you are charging users market rates to use your “open” API calls?

One thing I found shocking from the blog post is that OpenAI planned to close source their models long ago due to safety concerns. Ilya thinks AI could become unsafe and OpenAI itself should be a gatekeeper to decide what they want to share and who to share with. I understand his good intentions. At the same time, practically what they are doing is to concentrate the power in their own hands. How does OpenAI convince us that the power will not be abused to benefit its own leaders and the institution itself? OpenAI is also run by people who lean left. How do they ensure free speech across the whole political spectrum? Power corrupts. Absolute power corrupts absolutely. I am glad that OpenAI is not the only game in town. OpenAI has successfully become another Google in terms of AI prowess. But I hope there are at least dozens of OpenAI like AI companies so the world is not controlled by a small number of organizations. An organization can claim they are doing everything to benefit all humanity but talk is cheap. What really matters is what they actually do. So far, I can’t tell the difference between OpenAI and Google Gemini. Google at least never claimed to be a savior of humanity while OpenAI’s virtue signaling sounds grating and hypocritical.