Are there any potential risks or downsides to implementing the Global Workspace Theory in AI?