Making an AI evolve is a problem and you need control parameters. Using moods in numerical form is one of the understandable ways. Both the AI and the human interlocutor are evaluated with the same algorithms. The purpose of AI is to learn and associate moods with words. To condense in a few numbers what is the mental state of an intelligent entity.
What it does
AI SISM is an extension for a dialogue interface with an AI. Allows you to constantly compare your sentiment with that of AI. It allows to make a psychological self-assessment to the human user. Allows you to check the quality of the APIs and scoring criteria. Allows AI to evolve using the score criterion as a guide.
How we built it
Challenges we ran into
Trying to simplify the interface was a main goal. It was also important to be able to give as much information as possible with little data. Optimizing the underlying algorithm requires further research and experimentation.
Accomplishments that we're proud of
The ease of use of the interface after many attempts and solutions discarded. Adding a communicative window for emotional states adds further empathy. Being able to use the emotional aspects related to language will be useful for the evolution of AI itself.
What we learned
Measuring sentiment isn't easy at all. The system provided by the expert.ai API is a valid tool even if not yet perfect. The system shows how API sentiment classification requires further refinements. The API still allows it to be a good foundation to build on without having to start from scratch.
What's next for AI SISM
The AI SISM mechanism may have several future expansions. The system can be used as a growth engine for the AI that uses it. Some ideas for a possible symbolic elaboration on the same symbolic elaboration. Another small step towards AGI.
NOTE: The video describes an example of superimposed interaction, where the audio is dissociated from the images. In this way, in the same unit of time we want to transmit the unpredictable but coherent type of interaction. The scoring system for the positivity of sentences aims to compare both the human and the machine moreover, at the end it will be possible to evaluate the quality of the way in which the score is attributed in practice this system allows to give marks to the human, to the machine and also to the underlying API system. What grade will you be given? What grade do you think you deserved? Perhaps the difference lies in the need for better tuning of the algorithms ... Evolving AI through sentiment scores will be a way to have AI more aligned with human expectations.