Everyone is encouraged to say anything they feel like saying to their sapiens, that is how we identify bugs or other comprehension shortcoming in the system.
That being said, here are some known limitations, areas we already know need work. Since we already know about them there is nothing to be gained by revisiting them in your conversations and avoiding them could make your experience less frustrating.
- Sentence fragments. Other than responding to “Hello” and “Goodbye” we really don’t support comprehending one- or two-word inputs. Stay with complete sentences for now.
- Imperative Sentences. We support declarative and interrogative sentences but not imperatives. That is, you can’t tell you sapiens to do things – not even “Remind me to call home.” It might come back and say “I see” but it won’t do anything you tell it – yet. That is coming soon, however, so standby.
- Numbers, times and dates. These are essential and we are working on them right now, so again coming soon. Meanwhile telling your sapiens that you took your Metformin at 12:00 won’t work. (It should get “after lunch” though).
- Proper Nouns. most of the time, using proper nouns will work just fine, and “Otto is my cat” creates an instance of a cat with a name “Otto” and “Yosemite National Park” creates an instance of park with the right name but explicitly naming is definitely tricky, and answering questions about names is tricky. We knew there were problems with our routines that process names but now that we are viewing these dialogs, we see it more problematical that we thought. Pretty much everyone’s sapiens are now confused about what their own and their principles’ names are. Also we are not finished working on brand names. These are tricky because they are the only linguistic cases where a proper noun is used to identify a class of things and not a particular instance of a class. We need this asap for the Caregiver use case because people use Brand names and generic names interchangeably when talking about their medications.
- Negation. We don’t support negatives yet. So, it is pointless to say something is not something at this point. So once you teach your sapiens something that is not what you intended – you are stuck with it. Later when negatives are supported you will be able and clean things up. A direct way to clean things up now is to request to have your sapiens reset. This only takes a second but it will come back a clean slate like the moment you first met it. This is reasonable solution early on but obviously not after people have a lot of information entrusted to their sapiens. Down the road we will also build tools that will allow us to go in a particular sapiens memory with an editor in case something serious has gone wrong in the model structure and we need to straighten it out without losing stored information.
- Pronouns. Pronoun resolution is extremely difficult, we have invested much time on it will continue to do so. In the meantime keep pronoun usage to a minimum (he and she work pretty well if their referents in the conversation seem obvious)
- Why questions. Generally these are not supported except when your sapiens says “I have a problem with what you said” and similar variations, you can ask “Why” or “Why not?”