This is to introduce my github project (MIT license):
Taxonomy of common concepts based on matrices of concepts, for AI improvement
Compelling though it is, in most cases, the GPT model answers questions like these incorrectly:
With which concept is blissful optimism in the same type of relationship as subjectivity and objectivity?
The answer is: '... perception.'
We propose here a set of training data that allow for a response of the type:
'On the one hand, blissful optimism is a negative concept. On the other hand, subjectivity is a negative one and objectivity is a positive concept. Thus, the missing concept is a positive one. Hence, blissful optimism and awareness of problems are in the same relationship as subjectivity and objectivity.'
Top comments (0)