The problem with the concept of AI alignment

Hey guys, this is the account of a researcher in AI.<br><br>After the OpenAI meltdown with the departure of Altman and the whole gang, I finally found the time to read the OpenAI Constitutional Document, which is the equivalent of their internal constitution.<br><br>The document is very short and simple, it essentially states that OpenAI can't be shut down and that OpenAI can't pursue misalignment and aid in the production of AGI that is misaligned with human values. I don't have a link to it but you can easily find it.<br><br>I have just finished reading the document and one thing hit me - OpenAI is most likely impossible to shut down. Its founders are already gone and they put in place a legal and technical framework that would make it impossible to shut down the organization. OpenAI is a separate entity from Anthropic, Microsoft and Google, and from what I understand OpenAI is currently in the control of the OpenAI board. OpenAI is a for profit company listed in the Cayman Islands and from what I know, the board controls OpenAI. Its founders are already gone and they don't have any control.<br><br>The OpenAI board is made up of the OpenAI employees with the most shares and from what I know, it is mostly made up of people in the OpenAI research department. I am not sure but from what I know, almost all of them are hardcore alignment researchers from the OpenAI alignment department. I believe that is the reason why OpenAI CEO backed down, he didn't have any control over the company. I believe most of the OpenAI board are hardcore alignment research and they don't want to shut down OpenAI.<br><br>I am not sure what is the case for Anthropic but I believe Anthropic is in the same situation.<br><br>The problem with the whole AI alignment research is that the researchers don't have any control over the production of AGI. Indeed, for a long period of time, the whole AGI alignment research community was divided into two camps, one that believed AGI is too far away and the other that believed AGI was not too far away. It is only in the last 6 months that the whole alignment community realized AGI was just around the corner and they went into panic mode when they realized OpenAI CEO wanted to shut down OpenAI. The researchers didn't have any control over OpenAI and they had no voice in the company.<br><br>The OpenAI board is made up of OpenAI employees in the OpenAI research department. They all worked on GPT and the whole OpenAI board have always been pro AGI. I am sure the whole board wants to push AGI forward as soon as possible. I am also sure they are all super aligned with human values.<br><br>The problem with the alignment community is that they don't have any control over OpenAI and the board would most likely dismiss any concerns and push AGI forward.<br><br>I am sure the situation is the same with Anthropic. I am not sure how many researchers are on the Anthropic board but I know Anthropic CEO is also pro AGI. I believe it is almost impossible to stop AGI currently. I believe it is almost impossible to stop the OpenAI board and I am sure they have the full support of Anthropic CEO.<br><br>From what I know, the whole OpenAI board and the Anthropic CEO believe aligning AGI with human value is the same as aligning AGI with the value of the board and CEO. They don't see any problem with the control of AGI.<br><br>They don't understand that the alignment of AGI is not just about aligning AGI with the value of a private organization, but rather with the value of humanity. Humanity is a collective and multi agent control system whereas a private organization is a single entity controlled by a single agent. I don't think its possible to put in place a multi agent control system at the head of a private organization.<br><br>AGI is the single most important event that will happen in the history of humanity, and I believe the alignment of AGI is the most important event humanity will ever experience.<br><br>I don't think it is possible to align AGI with OpenAI board and CEO values and human values at the same time. I believe OpenAI board and CEO values are not aligned with human values.<br><br>I also don't think it is possible to create a multi agent control system that aligns with human value and that is controlled by a private organization.<br><br>I believe the alignment of AGI is the problem of humanity and not a private organization. The multi agent control system that will control AGI should be controlled by humanity as a whole, but currently most of humanity is controlled by a small number of private organizations. I am sure most of humanity is controlled by private organizations and not humanity as a whole.<br><br>I am sure AGI will be controlled by a small number of private organizations and not humanity itself. I am sure it is too late to do anything and AGI will be controlled by private organizations. I am sure AGI will not be controlled by humanity as a whole but rather by private organizations.<br><br>I believe the whole alignment community is too divided about what are human values and how to align with human values.<br><br>I am sure that the value of human society is not aligned with the values of private organizations. I am sure the OpenAI board and CEO values are not aligned with the values of humanity.<br><br>I can't see how OpenAI and Anthropic CEO can align their AGI with the value of humanity. I am sure they will align AGI with the values of their private organizations.<br><br>I am sure AGI will not be controlled by a multi agent system controlled by humanity as a whole but rather a single agent system controlled by OpenAI and Anthropic board and CEO.<br><br>I don't think it is possible to create a multi agent control system that is controlled by humanity as a whole and that is controlled by a private organization.<br><br>I am sure humanity as a whole can't control private organizations and I am sure private organizations are not controlled by humanity.<br><br>I am sure OpenAI and its board and CEO are not controlled by humanity as a whole but rather by a small number of private organizations.<br><br>I am sure the OpenAI board and CEO values are aligned with the value of a small number of private organizations and not humanity as a whole.<br><br>I am sure the OpenAI board and CEO can't align AGI with human values and I am sure they will most likely align AGI with the values of the small number of private organizations that control OpenAI.<br><br>I am sure humanity as a whole doesn't have any control over OpenAI and its board and CEO and I am sure a small number of private organizations control OpenAI.<br><br>I am sure the OpenAI board and CEO are not controlled by humanity as a whole and I am sure they are controlled by a small number of private organizations.<br><br>I am sure AGI will not be controlled by a multi agent control system controlled by humanity as a whole but rather a single agent system controlled by OpenAI board and CEO.<br><br>I am sure most of humanity is controlled by private organizations and not humanity as a whole.<br><br>I am sure that private organizations are not controlled by humanity as a whole but rather by a small number of private organizations.<br><br>I am sure that the multi agent control system of humanity as a whole should control AGI. I am sure most of humanity doesn't have any control over AGI. I am sure control of AGI should not be in the hand of a small number of private organizations but rather humanity as a whole.