I noticed this when I ask Chat GPT questions. It somehow gets stuck in cycles of answers and does not really integrate prior history to a great degree. I thought it would be great if he remembered all the answers it gave throughout a history of months and years and, for example, in terms of a medical condition, if you gave it clues and hints over years, it suddenly connected data point 79 from March with data point 889 from September and it gave a helpful directive in terms of that.

Of course this information would have to be strictly monitored and kept within the boundaries of a single safe account.

How far are we away from such a chat ai and what version in the closes we have now ?