Several times, I have been contacted about signing petitions for AI safety. I’ve visited websties such as safe.ai and read their mission statement. However, I refuse to sign such petitions. For example, the organization Center for AI Safety says on their website,
We believe that artificial intelligence (AI) has the potential to profoundly benefit the world, provided that we can develop and use it safely.
Simply put, I don’t believe in safe AI and managing AI risk nor do would I ever support a cause that gives any legitimacy to the idea that AI has benefits that we could reap if only we could manage the risks.
AI is simply too dangerous. No amount of laws will remove the serious dangers of AI. In fact, laws, regulations, and management are the worst sort of pursuit because they actively help AI integrate better with society. They take the development of AI as a foregone conclusion and are built to make AI just a little less unpleasant. I’d rather have AI be more unpleasant so that people outright reject it.
One can argue that AI safety is the next best thing to no AI at all, and that AI management at least makes conditions tolerable. I reject that as well. I’ll never accept that AI is necessary. Perhaps taking it down would be very difficult, but it is not impossible. If AI must be taken down by a revolution, then so be it.
Destroying AI is the only cause I support when it comes to AI. Some say that is unrealistic. I say that realism should not be a consideration for those that fight for what is right. And the unrealistic has a tendency to become real the moment someone does something unexpected. Moreover, it is better to fight for something unrealistic than to help the enemy, and the true enemy here is advanced technological society. There is simply no point in helping AI grow by making it more safe.