Finding ways to align superintelligent AI's goals with our objectives is important but difficult. Given the subgoal derivation problem, we should not solve superintelligence before resolving the alignment problem.
Computers have an architecture that can implement all functions independently of their structure. In contrast, the brain evolves with structures that are interlinked with their functions. Therefore, since the goal becomes parallel processing of information consciousness through structure and function itself, it is difficult to implement with serialized computers.
Omniscient LLM
Omniscient multimodal multi-task LLM
It is possible that we will have superintelligence in a few thousand days - Sam Altman 2024/9