There's been a ton of updates to the site and app recently but most of which you likely wouldn't care to read about. However, one of the things that is notable is an improvement to how we handle the context of the conversations.
Now for each of the phrases you speak to your AIVitars, the system will understand which questions actually hold context and which do not. This then allows the system to bring all the conversations up until the point of context within the scope of understanding so that we can be more precise whenever we're determining how much information to pass the AI models.
This is after a revamping that essentially knocked out the memories and context during testing due to other major changes that were being made to the system.
What exactly is context?
Context is the understanding of what subject we're talking about with the AIVitars.
For instance if you ask the AI
"What's the tallest building in the world?" It might answer "The largest building in the world is The BurjKhalifa in Dubai." This has the context because we know what we're talking about with a single phrase and answer.
However, if you then asked "How many floors does it have"
Without context, it would have no idea what you're referring to and with context it would understand that you previously asked about the tallest building in the world and what that answer was. The AIVitar can then infer that you are asking how many floors are contained within The Burj Khalifa and provide you with an accurate answer.
This improvement increases the percision of the system and will provide an overall better outcome for users.