Let's get things rolling!
Welcome to the crass.ai demo.
Below you see the so called "premise". It describes an event.
Below you see the so called "premise". It describes an event.
There are dozens of ways to describe resp. syntactically construe an event. crass.ai focusses in its first itteration on rather simple event descriptions, connecting two nouns by a verb with a corresponding preposition which
denote an event consisting of an (mostly physical) interaction of those two bodies, objects or actors.
Take a look at the premise to get an understanding of the event in question. Then click on "show counterfactual" to see the next step.
{{startairunconnection.data.unglaublich2.premise}}
The counterfactual or to be more precise, the "counterfactual conditional" is a notorious construct of many languages, which describes a situation by employing fake tense. That is the respective verb is conjugated to past
tense, although it's not about time but hypotheticality.
Most languages do not have a grammatical constructs to denote hypotheticality, but it is of paramount importance in order to understand if a speaker of a language actually masters the concepts used in that language, as he or
she needs a deep understanding in order to formulate the correct alternatives to a hypothetical situation.
So, look at the shown counterfactual and imagine this hypothetical scenario. Then click on "show response" to continue.
{{startairunconnection.data.unglaublich2.counterfactual}}
Finally, we are at the heart of our endeavour. The premises and counterfactuals were given to so called LLMs (large language models) with the task to state an admissable alternative to the counterfactual.
By inspecting those answers we get a good picture if an AI actually understood the situation at hand, or just took some syntactically correct and maybe semantically related notion it scrapped of off some training
data in order to respond.
It is here where the rubber meets the road.
It is here where the rubber meets the road.
Take a second to look at the answer and then decide if you think, that that is a correct answer a human would have uttered in response to this counterfactual question. Followinlgy, classify it as "correct" or "inncorrect" to
continue.