## On the computational measurement of intelligence factors (2000)

Venue: | National Institute of Standards and Technology |

Citations: | 14 - 8 self |

### BibTeX

@INPROCEEDINGS{Hernández-orallo00onthe,

author = {José Hernández-orallo and Departament De Sistemes Informàtics},

title = {On the computational measurement of intelligence factors},

booktitle = {National Institute of Standards and Technology},

year = {2000},

pages = {1--8}

}

### OpenURL

### Abstract

In this paper we develop a computational framework for the measurement of different factors or abilities which are usually found in intelligent behaviours. For this, we first develop a scale for measuring the complexity of an instance of a problem, depending on the descriptional complexity (Levin LT variant) of the ‘explanation ’ of the answer to the problem. We centre on the establishment of either deductive and inductive abilities, and we show that their evaluation settings are special cases of the general framework. Some classical dependencies between them are shown and a way to separate these dependencies is developed. Finally, some variants of the previous factors and other possible ones to be taken into account are investigated. In the end, the application of these measurements for the evaluation of AI progress is discussed.

### Citations

1681 | An Introduction to Kolmogorov Complexity and its Applications
- Li, Vitányi
- 1993
(Show Context)
Citation Context ...problems it is better (shorter) to give a more general solution than the specific solution for an instance of the problem. This has been formalised under the notion of “instance complexity” (see e.g. =-=[17]-=-), which gives the shortest solution to an instance of a problem provided it does not give a contradictory solution for other instances of the same problem. However, instance complexity is only of int... |

1155 | On computable numbers, with an application to the Entscheidungsproblem - Turing - 1936 |

404 |
A formal theory of inductive inference
- Solomonoff
- 1964
(Show Context)
Citation Context ...ism used. In [13] and [11] this difficult problem is addressed, according to a comprehensive criterion, a variant of the simplicity criterion based on Kolmogorov Complexity in the style of Solomonoff =-=[20]-=-, but ensuring that the data is covered comprehensively, i.e. without exceptions. Accordingly, the simplest explanatory description, denoted by SED(x|y), is defined in [11] as the simplest (in LT term... |

259 | Toward a mathematical theory of inductive inference - Blum, Blum - 1975 |

118 | Computing Machinery and - Turing - 1950 |

43 |
Universal search problems
- Levin
- 1973
(Show Context)
Citation Context ...me execution of a program, the formula LTβ(px) = l(px) + logτβ(px), where τ is the number of steps the machine β has taken until x is printed by py, was introduced by Levin in the seventies (see e.g. =-=[16]-=-). Intuitively, every algorithm must invest some effort either in time or demanding/essaying new information, in a relation which approximates the function LT. The corresponding complexity, denoted by... |

31 | The Universal Turing Machine A Half-Century Survey - Herken - 1988 |

30 | A formal definition of intelligence based on an intensional variant of Kolmogorov complexity
- Hernández-Orallo, Minaya-Collado
- 1998
(Show Context)
Citation Context ...re, in which aspects are they more intelligent? In this paper we investigate a framework for the evaluation of such a progress in different factors, extending in a natural way the work endeavoured in =-=[13]-=- and [11], which was specific developed for only some inductive factors. The main aim of the extension should be to develop the less number of factors as possible, by proposing general factors instead... |

21 | Eye on the prize
- Nilsson
- 1995
(Show Context)
Citation Context ...genceskillsofdifferent systems devised for quite different goals. We share the view that “it is time to begin to distinguish between general, intelligent programs and the special performance systems” =-=[19]-=-. How to define general and absolute characteristics of intelligence computationally is more difficult but also more attractive. In any way, in our opinion, the real progress inthe ‘intelligence’ of ... |

18 |
Three Approaches to the Quantitative Definition of
- Kolmogorov
- 1965
(Show Context)
Citation Context ...he term l(x) denotes the length or size of x in bits and logn will always denote the binary logarithm of n. The complexity of an object can be measured in many ways,oneofthembeingitsdegreeofrandomness=-=[15]-=-,which turns out to be equal to its shortest description. Descriptional Complexity, Algorithmic Information or Kolmogorov Complexity was independently introduced by Solomonoff, Kolmogorov and Chaitin ... |

9 | A model for intelligence - Eysenck - 1982 |

9 | Thesis: Computational measures of information gain and reinforcement in inference processes - Hernández-Orallo |

9 | Two kinds of probabilistic induction
- Solomonoff
- 1999
(Show Context)
Citation Context ...s which might be members of the set, although only one of them is really in it. Solomonoff formalised deterministic (sequential) prediction [20] and recently, has formalised non-sequential prediction =-=[22]-=-. This problem is similar to the inductive problem of learning a Boolean classifier and can be extended to the case of a general classifier. To eliminate the deductive contamination of the measurement... |

3 | A Formalization of the Turing Test (The Turing Test as an Interactive Proof System - Bradford, Wollowski - 1995 |

3 | Gödel’s Theorem and - Chaitin - 1982 |

2 |
How to Measure the
- Bien, Kim, et al.
- 1998
(Show Context)
Citation Context ... where their complexity (or learnability) has been established, there is no formal framework for giving a scale for concrete instances. In this new and beneficial interest in measurement, Bien et al. =-=[1]-=- have defined a ‘Machine Intelligence Quotient’ (MIQ), or, more precisely, two MIQs, from ontological and phenomenological (comparative) views. Any comparison needs a reference, and the only reference... |

2 | Complexity-based induction sytems: comparisons and convergence theorems - Solomonoff - 1978 |

1 | The inference to the best explanation”, Philos - Harman - 1965 |

1 | Computational Gain and Inference - Hernández-Orallo - 2000 |

1 |
Beyond the Turing Test”, to appear in
- Hernández-Orallo
(Show Context)
Citation Context ...ich aspects are they more intelligent? In this paper we investigate a framework for the evaluation of such a progress in different factors, extending in a natural way the work endeavoured in [13] and =-=[11]-=-, which was specific developed for only some inductive factors. The main aim of the extension should be to develop the less number of factors as possible, by proposing general factors instead of speci... |

1 | Needed: A New Test - Johnson - 1992 |