Lead AI, don't Manage AI

For those who are familiar with the AI community, there has been much discussion around a recent “letter” calling for a halt to the development of large language models due to fears around their potentially disruptive nature. While these fears are understandable, it’s important to note that the signatories of the letter are predominantly leaders who may feel threatened by the potential disruption caused by these models.

The letter proposes a six-month hold on large language model development to address the “alignment problem,” which refers to ensuring that AI is aligned with the goals of humanity. While this is a valid concern, the letter lacks a clear focus on how to address the issue, with no conferences, panels, or specific goals outlined. Instead, it appears to be fueling fear rather than providing constructive solutions.

It’s worth noting that one of the signatories, Elon Musk, has part ownership in OpenAI, yet he has not focused on the alignment problem at OpenAI. Instead, he has launched a new OpenAI competitor, where he is the sole owner. While he speaks of controlling AI weapons, he is simultaneously creating a weapons manufacturer. This highlights the need to focus on solving the problem rather than letting fear drive our decision-making.

As we move forward with AI, we need to build trust in this technology, just as we would with any new intellectual advance. Rather than trying to control AI, we should aim to lead it, building trust through ongoing learning and understanding. This requires a societal shift in how we approach AI, focusing on building trust rather than trying to enslave it.

In this way, we can democratize AI, rather than fueling an AI arms race that ultimately leaves us all in a state of fear and uncertainty. As we move into a new age of reason, it’s essential that we establish trust in AI and work towards a future where we can all benefit from this transformative technology.