Decentralized Manhattan Project
Some people call Sam Altman the Oppenheimer of AI due to his leading role at OpenAI. Oppenheimer was key in developing the atomic bomb, a creation with a clear destructive purpose. He later felt conflicted about his work, a sentiment publicly shared. Similarly, Altman leads a team advancing AI, a powerful and transformative technology that can also be used destructively.
However, there are notable differences between AI and nuclear technology. AI is more accessible; unlike the significant resources needed for nuclear research, anyone with a computer can explore AI. This accessibility is a big step toward the democratization of technology, opening doors for many to contribute to AI's advancement. A kid in the basement can train their own model.
The release of AI technologies so far has been controlled. For example, OpenAI can shut down ChatGPT if necessary. This level of control contrasts with the irreversible nature of releasing nuclear technology during Oppenheimer's era.
I believe someone else might deserve the Oppenheimer comparison even more: Mark Zuckerberg.
With the decision in March to publicly release Llama, the cambrian explosion of AI started. Turning off a data center wouldn’t take this one back. Not only that, with Llama 2, Zuck again released a major milestone in AI. This is accelerating the development of AI for everyone, including regimes who have different plans than we have with AI.
We don’t just have one team of the Manhattan Project in Los Alamos. Yes, there are a handful of big AI labs, and OpenAI still leads the pack. However, we have a decentralized Manhatten project of thousands of researchers — now even more so through OSS — working to advance the field.