A debate on CNN, MSNBC, FOX News, NBC, ABC, CBS, NBC, FOX that features an all AI system (no human hybrid) that in a timely manner paraphrases and assigns truth judgments (it can be of any type/scale) to (a number of) claims (made by candidates in a presidential primary or presidential general election) that are overlayed on the screen will resolve to Yes; the AI must necessarily make a certain number of truth judgments for it to resolve Yes (at least 10 total for the entire debate); the AI's paraphrasing and judgments need not be judged correct or accurate by humans ( it just has to show it tried so even a confabulation/hallucination would be counted ); the AI judgments must be part of the broadcast production televised stream to count (if the AI fact checking is offered only through a particular streaming service app or device it is not counted - in other words everyone able to, must see it). If truth judgments are displayed more than 30s after the candidate finishes the respective claim, then that jugment is not considered timely and not counted towards the total. To repeat, if the truth claims are made available as optional, opt-out/opt-in or as a separate device or app that is not counted.
I am using the following definition to satisfy what is a truth judgment: "Truth judgments reflect inferences drawn from three types of information: base rates, feelings, and consistency with information retrieved from memory." (https://www.annualreviews.org/doi/10.1146/annurev-psych-010419-050807)
Related questions
I want to clarify the spirit of this question: if all the AI system is doing is a transcription of the speech, doing semantic comparison to a database of generated fact-checking claims, and then outputting a result based on some threshold of similarity, this WILL count as a truth judgment. I understand some people think that this is not a "judgment" since it 'seems' to be a simple look-up, but I am thinking of this as the AI system using such a database of claims as an authoritative reference for facts; this might seem a stretch if not for the context of what it is going to be used against, which is most likely to be politicians repeating claims they've made many times before.
I don't anticipate this to be problematic, as the first definition I retrieved for "truth judgment" is: "Truth judgments reflect inferences drawn from three types of information: base rates, feelings, and consistency with information retrieved from memory." (https://www.annualreviews.org/doi/10.1146/annurev-psych-010419-050807)
I believe the above classifies as an inference (semantic similarity & threshold) from memory (database). I have updated the description to use this definition of truth judgment.
@Nikola I want there to be some uncertainty to make it more interesting as we near the debates (I will resolve it by the end of 2028). It's possible the committees for the debate will allow it, but then the networks might pull the rug out at the last moment if the candidates get cold feet and pressure them by threatening to pull out.