116 Progress toward Cyberimmortality links such as that “gives” requires someone in a giver relation- ship and someone in a receiver relationship. Nobody knows how many episodes an adult remembers, but the number 50,000 is sometimes mentioned in Cognitive Science discussions. If Shastri is right, then a computer model of the hippocampal system might not need to be very big. If concepts are addressed efficiently in other parts of the brain, then an episode like “John gave Mary a book” could be stored in fewer than a hundred bytes. This would imply the entire hippocampal system could get by with only 5 megabytes of memory, a tiny fraction of the memory of today’s pocket com- puters. Shastri notes that the human mind imaginatively fills in the missing details of memories, and each episodic memory really has very little information in it. We are not generally conscious of the yawning gaps in our memory, any more than we are conscious of the blind spot in the vision of each of our eyes. Shastri also notes that memories are generally stored in multiple copies, perhaps to guard against losing them through the death of any single neuron, but the redundancy in memory may have other functions as well, such as helping us combine facts from different sources by placing some copies of them nearer to each other. Computerized memories might not need that redundancy. This very quick summary of the current state of AI suggests that we really cannot predict how soon computer and infor- mation scientists will be able to simulate real human minds. Rapid progress is going on in other directions, and in a few years  a  renewed  interest  in  duplicating  human  intelligence could plausibly move forward very quickly by exploiting all the discoveries and inventions that are being made now for other purposes. Once we know how to duplicate a mind in a computer, then we will know far better how much informa- tion of what kinds we will need. We do not currently have this knowledge, but such rapid progress is being made in several