An Agnostic Test for Narrow, General, and Super Intelligence Based On the Principles of Recursive Compression and Algorithmic Probability

View the PDF file for the Superarc title: an unrelated test for narrow intelligence, general, and superior intelligence based on the principles of pressure and the possibility

PDF HTML (experimental) view

a summary:We provide an open test based on the possibility of a algorithm that can avoid standard pollution in the quantitative evaluation of the border models in the context of artificial public intelligence claims (AGI) and Superintebleign (ASI). Unlike other tests, this test does not depend on statistical pressure methods (such as GZIP or LZW), which are closely related to Evrobia Shannon from the complexity of Kolmogorov and cannot test behind the matching simple patterns. The test defies aspects of artificial intelligence, especially LLMS, related to the features of basic nature intelligence such as synthesis and the creation of the model in the context of reverse problems (generating new knowledge of observation). We affirm that the standards based on the abstraction of model and kidnapping (optimal bizer inference) for the “planning” of “planning” can provide a strong framework for an intelligence test, including natural intelligence (human and animal), narrow narrow intelligence, AGI, and ASI. We have found that LLM versions tend to be fragile and increasing as a result of the party only with progress that is likely to be driven by the size of the training data. The results were compared to the mixed nervous approach that theoretically guarantees global intelligence based on the principles of algorithm and the complexity of Kolmogorov. The method excels over LLMS to prove the concept over short bilateral sequences. We prove that the pressure is equivalent and is directly proportional to the predictive strength of the system and vice versa. That is, if the system can predict better that it can press better, and if it can press better, it can predict better. The results of the suspicion are enhanced by the basic restrictions of LLMS, and their exposure as improved systems to perceive the human language.