Many of you are familiar with Patient HM, the famously co-operative epilepsy patient whose seizures were alleviated with a radical new technique – removing the hippocampus and parahippocampus. They managed to stop the wild propagation of nonsensical signalling cascading through his neurons. This man, Henry Molaison, is responsible for much of our understanding of the neuroscience of the hippocampus as the seat of memory, as well as hemisphere specialization and plasticity after traumatic brain injury. (His brain worked like PrintWriter – old experiences weren’t saved, they were simply constantly overwritten with the new.)
But he did not need to live in an MRI machine for us to learn from him. Most of the experiments he engaged in were observational. Some researchers worked with him for years. And the interesting thing is, the people who worked with him all describe him as warm and friendly, and they were very fond of him. Explicit memory was unnecessary for people to feel connected to him.
What are computers great at? Remembering things for us. But they aren’t great with complete experiences, or analog experiences. So the things they can remember are fairly limited. And unlike the brain, they do not strengthen connections over time – there is no “implicit” memory. Computers are explicit or bust. But we want to assign implicit memory to them. We want our computers to remember “us”. We want computers to be part of the rest of “us” – not just a tool but an extension of ourselves.
Is there a way to do that? Can we make computers remember us somehow? Would it need to be a learning algorithm running constantly? And if we do that, will it become impossible to treat them objectively?