sign in sign up
<<
Home , Charles Spearman, G factor (psychometrics)

Can the Play a Role in Artificial General Research?

Davide Serpico 1 and Marcello Frixione 2

Abstract. In recent years, a trend in AI research has ficial systems that behave intelligently in narrow started to pursue human-level, general artificial intel- domains, namely, “narrow AI”. These kinds of ligence (AGI). Although the AGI framework is char- artificial systems can carry out domain-specific acterised by different viewpoints on what intelli- intelligent behaviours in specific contexts and gence is and how to implement it in artificial sys- are, thus, unable to self-adapt to changes in the tems, it conceptualises intelligence as flexible, gen- context as general-intelligent systems can do [2- eral-purposed, and capable of self-adapting to differ- ent contexts and tasks. Two important questions re- 4]. main open: a) should AGI projects simulate the bio- The realisation of a human-level artificial logical, neural, and cognitive mechanisms realising intelligence has seemed unfeasible to many the human intelligent behaviour? and b) what is the scholars until recent years. However, in the last relationship, if any, between the concept of general two decades, the AI community has started to intelligence adopted by AGI and that adopted by pursue the goal of a “human-level” artificial psychometricians, i.e., the g factor? In this paper, we general intelligence (AGI). This is attested by address these questions and invite researchers in AI several conferences, publications, and projects to open a discussion on the theoretical conceptions on human-level intelligence and related topics 1 2 and practical purposes of the AGI approach. [4-5]. Although these projects point to many dif- ferent directions to be followed by AGI re- 1. INTRODUCTION: THE AGI search, they represent a new movement towards HYPOTHESIS the concrete realisation of the original dream of a “strong AI”. The dream of the first generation of AI research- Two important movements intertwined with ers was to build a computer system capable of AGI emphasise the importance of the simulation displaying a human-like intelligent behaviour in of the human mind. The first, known as Biologi- a wide range of domains. Since human intelli- cally Inspired Cognitive Architectures (BICA), gence is highly flexible with respect to different aims to integrate many research efforts involved tasks, goals, and contexts, making the dream in creating a computational equivalent of the come true would have required developing a human mind. The second, which has been ini- general-purposed thinking machine. tially proposed during the First Annual Confer- In spite of some initial success (e.g., Newell ence on Advances in Cognitive Systems (Palo and Simon’s General Problem Solver [1]), the Alto, 2012), aims to achieve the goals of the attempts of researchers did not result in a do- original AI and cognitive science, that is, ex- main-general AI. What they achieved was, ra- plaining the mind in computational terms and ther, the development of highly specialised arti- reproducing the entire range of human cognitive abilities in computational artefacts [3]. 1 School of Philosophy, Religion and History of Science, As we mentioned, the AGI community un- University of Leeds. Email: [email protected] derstands general intelligence as the ability, dis- 2 Dept. of Classics, Philosophy and History, University of played by humans, to solve a variety of cogni- Genoa. Email: [email protected]

Proceedings of the Society for the Study of Artificial Intelligence and Simulation of Behaviour 2018 D. Serpico & M. Frixione, 2018 tive problems in different contexts. Thus, nearly 2. EMULATING OR SIMULATING all AGI researchers converge on treating intelli- GENERAL INTELLIGENCE? gence as a whole: indeed, intelligence appears as a total system of which one cannot conceive one is defined by psychometri- part without bringing in all of it [5-6]. Ben cians as a domain-general cognitive ability, Goertzel [4] delineates the core AGI hypothesis namely, the g factor (see Section 3 for further as follows: the creation and study of a synthetic details). Wang and Goertzel [5] have rapidly intelligence with sufficiently broad scope and dismissed any connection between AGI research strong generalisation capability is qualitatively and the psychometric concept of general intelli- different from the creation and study of a syn- gence. According to them, projects in AI are not thetic intelligence with significantly narrower interested in the psychological description of scopes and weaker generalisation capability. human intelligence, if not in a weak sense. What is general intelligence? How can it be However, this conclusion seems to be, at implemented in artificial systems? In order to best, premature. Indeed, some attempts in AGI address these questions, it is necessary to open a research have encompassed a notion of intelli- discussion on both theoretical and practical is- gence that should be evaluated through the sues in AGI research. lenses of empirical findings. Since general intel- In this paper, we aim to clarify what rela- ligence represents to many psychologists, neuro- tionship exists, if any, between the concept of scientists, and geneticists the most important and human general intelligence and the AGI hypoth- well-studied aspect of human psychology [7, 8], esis. General intelligence was first conceptual- we cannot see any strong argument against the ised in the early twentieth century within psy- possible role of the g factor in (at least some) chometric research. Remarkably, as we shall research in AGI. Let us see why. show, is quite a different kind of An AGI project can aim to either emulate or psychological science than the one traditionally simulate human intelligence. In the case of emu- tied to AI, that is, cognitive science. We shall lation, an artificial system will display a human- argue that AGI researchers cannot safely rely on like intelligent behaviour regardless of details 3 the psychometric concept of general intelligence about its realisation or implementation. In the and should, rather, look at intelligence as emerg- case of simulation, instead, an artificial system ing from several distinct biological and cogni- will display general intelligence not only at the tive processes. behavioural level but also at the mechanistic and In Section 2, we analyse different view- processing level. In other words, human-level points on whether AGI research should emulate artificial intelligence is realised by underlying or simulate human intelligence. Since many AGI mechanisms which are analogue to those realis- projects are inspired by psychological, neurosci- ing human intelligence. The former case likely entific, and biological data about human intelli- represents the notion of AGI that Wang and gence, scholars in AI should care about the psy- Goertzel [5] have in mind. Our targets are, in- chometric theory of general intelligence, its stead, examples of AGI research characterised promises and perils. In Sections 3 and 4, we by the latter approach. summarise the fundamental aspects of such a Before analysing this approach in more de- theory by emphasising the widespread disa- tails, it is worth considering that whether AGI greement about the existence of general intelli- gence. In Section 5, we outline important impli- 3 Here, behavioural assessments, such as Turing’s test and cations for contemporary research on Artificial Nilsson’s employment test, can address whether we have General Intelligence. achieved a human-level artificial intelligence: in brief, systems with true human-level intelligence should be able to perform human-like tasks [9].

Proceedings of the Society for the Study of Artificial Intelligence and Simulation of Behaviour 2018 2

D. Serpico & M. Frixione, 2018

projects should emulate or simulate human intel- ing AI architectures by enlightening the func- ligence, and the relationship between BICA and tioning of central aspects of intelligence such as AGI, are controversial topics. Franklin and col- , attention, and . leagues [3, 10] agree that an AGI agent may be A dialogue between neuroscience and AI successfully developed by using an architecture research seems to be largely welcomed within that is not biologically inspired. However, they the AGI community. A survey conducted by argue, the goals of AGI and BICA are essential- Muller and Bostrom [13] highlights how, ac- ly equivalent. Indeed, AGI hopes to solve the cording to many researchers, a human-level AI problem already solved by biological , is to be achieved by means of research ap- namely, to generate adaptive behaviour on the proaches tying AI to neuroscience—e.g., Inte- basis of sensory input. Since biological minds grated Cognitive Architectures, Computational represent the sole examples of the sort of robust, Neuroscience, and Whole Brain Emulation. Of flexible, systems-level control architectures course, the commitment to the simulation of needed to achieve human-level intelligence, psychological, cognitive, and biological aspects copying after these biological examples—as of human intelligence is exerted in many ways. BICA projects do—represents a valuable strate- Let us see some examples. gy. 4 Some projects belonging to BICA and Wang [11] disagrees with this point of AGI’s agendas (e.g., SyNAPSE, HTM, SAL, view. According to him, in a broad sense, all AI ACT-R, ICARUS, LIDA, the ANNs, the Human projects take the human mind as the source of Brain Project, and the Large-Scale Brain Simu- inspiration. Nonetheless, few AI researchers lator) are interested in various aspects of the have proposed to duplicate a human cognitive human general intelligence and accept, though feature without providing a why this is to different degree, that simulating the human needed—consider that computers and human brain’s structure can be promising for AI re- beings are different from each other in many search [3, 14-20]. fundamental aspects. Therefore, the important Further, various researchers are inspired by decision for an AGI project is where to be simi- the ontogenetic and phylogenetic aspects of hu- lar to the human mind and why this similarity is man intelligence and suggest that we should desired. simulate the same facilities for learning that hu- Our aim is not to take a side in this contro- man infants have or the evolutionary trajectory versy, but rather to stress that some AGI projects of intelligence [9, 21, 22]. are, in fact, inspired by empirical data on human Lastly, Wang [6] suggests that an AGI sys- intelligence. Hassabis and colleagues’ review tem may require a single mechanism capable of [12] provides several examples of how neuro- reproducing the general-purpose, flexibility, and science has inspired both algorithms and artifi- integration of human intelligence. Accordingly, cial architectures. Moreover, neuroscience a general intelligent system should comprise seems to be able to provide validation of already both domain-specific and domain-general sub- existing AI techniques as well: if a known algo- systems: while the existing domain-specific AI rithm is found to be implemented in the brain, techniques are considered tools for solving spe- then this is strong support for its plausibility as cific problems, the integrating component is an integral component of an intelligent system. general, flexible, and can run the various do- In this view, brain studies have helped develop- main-specific programs. In the proposed archi- tecture, i.e., the NARS, reasoning, learning, and categorisation represent different aspects of the 4 For instance, since mind and brain are strictly related, same processes. This approach, which highlights the LIDA’s theoretical model proposed by the authors the relationship between general intelligence and seeks to reproduce it in silico.

Proceedings of the Society for the Study of Artificial Intelligence and Simulation of Behaviour 2018 3

D. Serpico & M. Frixione, 2018

a hypothetic domain-general mechanism, is par- there are two different ways of g: ticularly interesting to us. Indeed, as we shall on the one hand, the psychometric g; on the oth- explain shortly, the psychometric theory of hu- er hand, the neurocognitive g. Let us see them man intelligence draws on similar intuitions. one by one. Is human intelligence related to a single, From a psychometric point of view, the g general cognitive mechanism? How can general factor is related to the so-called positive mani- intelligence emerge from the complexity of the fold: individuals who show good performance human brain? To the extent that AGI researchers on a given task will tend to show good perfor- aim to reproduce human intelligence on a mech- mance also in other tasks. In other words, intel- anistic and processing level, they should care ligence measurements are positively intercorre- about these questions, which are typically ad- lated both in different cognitive domains and dressed by empirical research on human intelli- different individuals. Spearman understood that gence. In the next two sections, we briefly re- scores of a battery of tests tend to load on one view the psychometric theory of general intelli- major factor regardless of their domain. He em- gence and ask whether AGI and BICA projects ployed to identify this factor. The can safely rely on it. g factor, as Spearman called it, is a latent varia- ble which summarises the typical correlation matrix of intelligence test scores. 3. THE THEORY OF HUMAN What is the meaning of the psychometric g? GENERAL INTELLIGENCE Factor analysis can be understood as a procedure The concept of general intelligence was born in of “distillation” capable of identifying a factor the early twentieth century, in parallel with the reflecting the variance that different intellectual rise of the psychometric tradition. The central measures have in common. In this sense, the g aim of psychometricians is to develop method- factor explains ~40 percent of tests’ variance. ologies capable of assessing and quantifying in- Thus, it reflects individual differences in tellectual differences among people, i.e., IQ performance in intellectual tasks [7, 25]. This tests. Over the last century, these tests served a interpretation of g can hardly find room in variety of purposes, ranging from educational to neuroscientific research: indeed, the clinical ones, but they played a central role in psychometric g does not represent a “concrete” empirical research as well (e.g., in behavioural neurocognitive mechanism, but rather an and neuroscience). “abstract” entity or a property of a population of The concept of intelligence is generally re- individuals (see [26] for similar concerns]. lated to a wide range of psychological aspects From a neurocognitive point of view, the and adaptive capabilities (e.g., learning, story is different. In neuroscience, the g factor is , social skills, and ; see understood as a domain-general cognitive ability [23]). By contrast, psychometricians have main- that characterises the human brain [27]. In this ly focused on the cognitive abilities mostly in- respect, it represents the fundamental mecha- volved in the solution of the IQ test items (e.g., nism underlying general intelligence. However, mathematical, linguistic, logic, and visual- the meaning of neurocognitive g is still unclear. spatial abilities). Thus, general intelligence rep- When Spearman tried to clarify the nature of resents a theoretical construct related to these intelligence, he described g as a form of mental cognitive domains. energy. Successive researchers have tried to re- In order to clarify the nature of general in- duce g to some neurocognitive properties of the telligence, psychometricians generally refer to brain, e.g., , processing speed, what [24] called the general or neural efficiency (see Section 4 for details). factor of intelligence or g factor. Remarkably,

Proceedings of the Society for the Study of Artificial Intelligence and Simulation of Behaviour 2018 4

D. Serpico & M. Frixione, 2018

The reliability of the psychometric g is gen- Theory [30, 34], in turn, aims to locate the g fac- erally accepted as the positive manifold repre- tor into the human brain, i.e., in the parietal and sents a stable empirical phenomenon. By con- frontal regions. trast, several concerns have been raised on the Although these models represent interesting neurocognitive interpretation of g. In recent dec- attempts, many scholars believe there is no room ades, many neuroscientists have criticised the for general intelligence in contemporary neuro- concept of general intelligence and developed science. Indeed, most contemporary theories of non-generalist conceptions of human intelli- intelligence do not include the g factor within gence; others have interpreted g as a mere statis- the human cognitive architecture and do not tical artefact. In the next section, we analyse the identify a single general mechanism capable of controversial role of the general factor of intelli- summarising individual performances as a glob- gence in neuroscientific research. al score such as IQ. Rather, several aspects of biology and cognition are invoked. Renowned 4. A NEUROSCIENTIFIC VIEW ON examples are the theory of Multiple Intelligenc- es [35], the PASS model [36], and the Multiple THE g FACTOR Cognitive Mechanisms approach [37]. All these Is there any evidence of the existence of g in the theories appeal to the role of several distinct human brain? Is human intelligence general or cognitive processes to explain the human intelli- not? Since psychometrics and cognitive science gent behaviour. met a few decades ago, these questions divide If there is no general mechanism such as g scholars for both empirical and theoretical rea- in the human brain, why then the positive mani- sons. fold? Some scholars have recently provided val- From an empirical point of view, pro-g uable explanations of the empirical correlations scholars have tried to reduce g to neurocognitive among IQ tests performance without invoking a constructs, often assumed as reliable and, hence, general underlying mechanism. According to suitable to make sense of g in neuroscientific these proposals, the psychometric g is supported terms. Associations have been found, for in- by multiple, interacting mechanisms that be- stance, between IQ and processing speed, work- come associated with each other throughout the ing memory, problem-solving, meta-cognition, course of development. For instance, the mutu- attention, associative learning, glucose metabol- alist model, proposed by Van der Maas and col- ic rates, electrocortical activity, and leagues [37], recognises that the positive mani- [28-30]. However, to find reliable associations fold is a robust empirical phenomenon, but ad- between g and these variables has not been easy vances an explanation based on a developmental at all: replicability rates are often low and spuri- model involving the relationships between cog- ous correlations are ubiquitous. Moreover, the nitive processes. The mutual influence between associations between g and other aspects of the these processes gives rise to the positive mani- human brain are often considered to be theoreti- fold but rules out g as a single, . cally inconsistent or, at best, weak [31, 32]. According to the architects of this model, there From a theoretical point of view, pro-g is nothing wrong with using the g factor as a scholars have developed theories of intelligence summary index as long as we do not assume that aimed at reconciling neuroscientific and psy- this variable relates to a single underlying pro- chometric approaches. For instance, the Minimal cess. 5 Cognitive Architecture Theory [33] aims to To summarise, cognitive neuroscientists of- match the psychometric view with developmen- ten deny the existence of the neurocognitive g tal theories of intelligence and with the modular . The Parieto-Frontal Integration 5 See [2] for developmental approaches in AGI.

Proceedings of the Society for the Study of Artificial Intelligence and Simulation of Behaviour 2018 5

D. Serpico & M. Frixione, 2018

and, thus, suggest that general intelligence does relative in the human domain, the psychometric not represent a valuable posit for understanding one. Artificial systems inspired by such a theory human cognition (see also [38]). The disagree- can well turn out to be less human-oriented than ment about the existence of the g factor can be other, classical ones, such as the so-called nar- clarified by considering the theoretical gap be- row AI systems. Second, implementing some tween psychometrics and cognitive science. sort of general-purpose mechanism in artificial Since the birth of , cogni- systems to emulate the human intelligent behav- tive scientists have focused on the functional- iour—as Wang [6], for instance, suggests—may structural segmentation of the human mind. not be the right strategy. Thus, in a neurocognitive perspective, mental It is worth noting that, in general, a psy- abilities and cognitive processes cannot be con- chometric-like view of intelligence does not sidered properties of the brain taken as a whole: play a central role in AI. Indeed, most contem- rather, they are implemented by specific brain- porary artificial architectures do not assume that areas and populations of neurons (for instance, a human-level intelligence necessarily requires a the modularity of mind hypothesis relies on this single generative mechanism. Rather, intelli- assumption). This conclusion is sometimes gence is understood as emerging from many un- agreed by researchers in AGI as well. For in- derlying aspects—an interpretation with which, stance, Goertzel [4] has contrasted the concep- as we have shown, many cognitive neuroscien- tion of general intelligence with approaches tists agree. At the same , almost any scholar looking at the various competences that humans would agree that the classical narrow approach display (see the list of competences assembled at to AI is unsuccessful in shifting towards a hu- the 2009 AGI Roadmap Workshop [2]). man-level intelligence. In the last section, we explore some impli- So, what lies between specialised artificial cations for research in AI. systems and a single domain-general mecha- nism? Is there any intermediate level to work 5. CONCLUSIONS: IMPLICATIONS on? Essentially, these are the questions AGI re- searchers need to address (see [14, 39]). In other FOR AI RESEARCH words, AGI researchers are asked to develop The quest for the nature of human intelligence, lower-level, specific-purposed systems capable involving its generality and its architecture, re- of generating higher-level networks of processes mains open. However, it stands to reason that and interactions. These networks would argua- general intelligence cannot be safely understood bly realise general intelligence at the behaviour- as a real biological entity. Rather, we can de- al level. Indeed, intelligence represents a sys- scribe it as a behavioural, emergent phenomenon temic and dynamical property of complex sys- due to the causal interaction between many as- tems. pects of the neurocognitive development. Ac- Unfortunately, even complex cognitive ar- cordingly, intelligence seems to be a term im- chitectures, such as SOAR and ACT-R, are ported by everyday life that clusters together characterised by both technical and epistemolog- distinct cognitive processes, autonomous to a ical problems (see e.g., [14, 40, 41]). Neurosci- certain extent both in developmental and evolu- entific theories of intelligence can help AI by tionary terms. providing a meaningful explanation of human What does this imply for AI researchers neurocognitive development. Nevertheless, tak- who adopt a generalist view of intelligence? ing up the challenge ultimately depends on the Two things, at least. First, the generalist concep- ability of AI researchers to pick up the relevant tion of intelligence, if adopted in AGI and BI- conceptions of what an intelligent system is. In CA, threatens to inherit the weaknesses of its

Proceedings of the Society for the Study of Artificial Intelligence and Simulation of Behaviour 2018 6

D. Serpico & M. Frixione, 2018 this sense, a discussion on general intelligence REFERENCES in AI seems to us, at present, inevitable. [1] Newell, A., & Simon, H. (1976). Computer science as Can the g factor play a role in AGI research, empirical inquiry: symbols and search. The Tenth Tu- then? In light of our discussion, the answer can ring Lecture. of the Association for be either positive or negative depending on the Computing Machinery, 19 . very aim of AGI. A weak or instrumental notion [2] Adams, S., Arel, I., Bach, J., Coop, R., Furlan, R., of g, like the psychometric g, can play a role in Goertzel, B., ... & Shapiro, S. C. (2012). Mapping the landscape of human-level artificial general intelli- AGI projects characterised by an emulative ap- gence. AI magazine , 33 (1), 25-42. proach, where the goal is reproducing a human- [3] Franklin, S., Strain, S., McCall, R., & Baars, B. level intelligence regardless of details about its (2013). Conceptual commitments of the LIDA model neurocognitive or biological architecture. Here, of cognition. Journal of Artificial General Intelli- the psychometric g, as assessed by IQ tests, gence , 4(2), 1-22. [4] Goertzel, B. (2014). Artificial general intelligence: might help to evaluate the intelligent behaviour concept, state of the art, and future prospects. Journal of artificial systems besides other behavioural of Artificial General Intelligence , 5(1), 1-48. tests—e.g., Turing and Nilsson’s tests. [5] Wang, P., & Goertzel, B. (2007). Introduction: As- By contrast, a strong, neurocognitive notion pects of artificial general intelligence. In Proceedings of g is involved in the discussion about the com- of the 2007 conference on Advances in Artificial Gen- eral Intelligence: Concepts, Architectures and Algo- position of human intelligent systems, the causal rithms (pp. 1-16). IOS Press. interactions among parts, and how to artificially [6] Wang, P. (2004). Toward a unified artificial intelli- reproduce these aspects. In this respect, the pos- gence. In AAAI Fall Symposium on Achieving Human- sible role of the g factor in AGI research de- Level Intelligence through Integrated Research and pends on empirical data in neuroscience. As we Systems (pp. 83-90). [7] Jensen, A. R. (2002). Psychometric g: Definition and have argued in this paper, this role of g in AI is Substantiation. In R. Sternberg & E. Grigorenko dubious. (Eds.), The General Factor of Intelligence. How Gen- As we noticed, AGI research encompasses eral is it? (pp. 39-54). Mahwah: Lawrence Erlbaum. different viewpoints on what intelligence is and [8] Plomin, R., DeFries, J. C., Knopik, V. S., & Neiderhis- on what the purposes of a human-level AI are. er, J. N. (2013). Behavioral Genetics (6th ed.). New York: Worth Publishers. While many authors in AGI are cautious about [9] Nilsson, N. J. (2005). Human-level artificial intelli- their assumptions, others believe it is not enough gence? Be serious! AI magazine , 26 (4), 68-75. to merely emulate the intelligent behaviour. Ra- [10] Strein, S., & Franklin, S. (2013). Authors’ Response ther, in this view, artificial systems should simu- to Commentaries. In H., Dindo, J., Marshall, & G., late the mechanisms and processes that make Pezzulo (2013). Editorial: Conceptual Commitments of AGI Systems. Journal of Artificial General Intelli- humans intelligent the way they are. For these gence , 4(2), 48-58. approaches, where theories and data adopted by [11] Wang, P. (2013). Conceptual Commitments of AGI cognitive neuroscientists play an important role, Projects. In H., Dindo, J., Marshall, & G., Pezzulo we invite cautious about the commitment to the (2013). Editorial: Conceptual Commitments of AGI concept of general intelligence. As Goertzel [4] Systems. Journal of Artificial General Intelli- gence , 4(2), 39-42. notices, brain sciences are advancing rapidly, [12] Hassabis, D., Kumaran, D., Summerfield, C., & Bot- but our knowledge about the brain is extremely vinick, M. (2017). Neuroscience-inspired artificial in- incomplete. Seemingly, relying on a controver- telligence. Neuron , 95 (2), 245-258. sial theory of human intelligence, such as the [13] Müller, V. C., & Bostrom, N. (2016). Future progress psychometric one, can be perilous for AI re- in artificial intelligence: A survey of expert opinion. In V. C. Müller (Ed.), Fundamental issues of artificial search. intelligence (pp. 555-572). Berlin: Springer. [14] Anderson, J. R., Bothell, D., Byrne, M. D., Douglass, S., Lebiere, C., & Qin, Y. (2004). An Integrated Theo-

Proceedings of the Society for the Study of Artificial Intelligence and Simulation of Behaviour 2018 7

D. Serpico & M. Frixione, 2018

ry of the Mind. , 111 (4), 1036- [31] Kray, J., & Frensch, P. (2002). A View from Cogni- 1060. tive Psychology: g–(G)host in the Correlation Matrix? [15] Sun, R., & Naveh, I. (2004). Simulating organiza- In R. Sternberg & E. Grigorenko (Eds.), The General tional decision-making using a cognitively realistic Factor of Intelligence. How General is it? (pp. 183- agent model. Journal of Artificial Societies and Social 220). Mahwah: Lawrence Erlbaum. Simulation, 7 (3). [32] Ramus, F. (2017). General intelligence is an emerg- [16] Gunzelmann, G., Gluck, K. A., Van Dongen, H. P. ing property, not an evolutionary puzzle. Behavioral A., O’Connor, R. M., & Dinges, D. F. (2005). A Neu- and Brain Sciences , 40 , 43-44. robehaviorally Inspired ACT-R Model of Sleep Depri- [33] Anderson, M. (2005). Marrying intelligence and cog- vation: Decreased Performance in Psychomotor Vigi- nition. A Developmental View. In R. J. Sternberg & J. lance. In B. G. Bara, L. Barsalou & M. Bucciarelli E. Pretz (Eds.), Cognition and Intelligence (pp. 268- (Eds.), Proceedings of the Twenty-Seventh Annual 287). Cambridge, UK: CUP . Meeting of the Cognitive Science Society (pp. 857- [34] Jung, R. E., & Haier, R. J. (2007). The Parieto- 862). Mahwah: Lawrence Erlbaum. Frontal Integration Theory (P-FIT) of intelligence: [17] Markram, H. (2006). The Blue Brain Project. Nature converging neuroimaging evidence. Behavioral and Reviews Neuroscience 7 (2), 153-160. Brain Sciences , 30 (2), 135-154. [18] Chong, H. Q., Tan, A. H., & Ng, G. W. (2007). Inte- [35] Gardner, H. (1983). Frames of mind: The theory of grated cognitive architectures: a survey. Artificial In- multiple . New York: Basic books. telligence Review , 28 (2), 103-130. [36] Naglieri, J. A., & Das, J. P. (2002). Practical Implica- [19] Izhikevich, E. M., & Edelman, G. M. (2008). Large- tions of General Intelligence and PASS Cognitive Pro- Scale Model of Mammalian Thalamocortical Systems. cesses. In R. Sternberg & E. Grigorenko (Eds.), The PNAS, 105 , 3593-3598. General Factor of Intelligence. How General is it? [20] Jilk, D. J., & Lebiere, C. (2008). SAL: An explicitly (pp. 55-84). Mahwah: Lawrence Erlbaum. pluralistic cognitive architecture. Journal of Experi- [37] Van der Maas, H. J. L., Dolan, C. V., Grasman, R. P. mental and Theoretical Artificial Intelligence 20 , 197- P. P., et al. (2006). A dynamical model of general 218. intelligence: the positive manifold of intelligence by [21] Brooks, R. A. (1997). From Earwigs to Humans. Ro- mutualism. Psychological review, 113 (4), 842-861. botics and Autonomous Systems , 20 (2-4), 291-304. [38] Serpico, D. (2017). What Kind of Kind is Intelli- [22] Koza, J. R., et al. (2003). Genetic Programming IV: gence? Philosophical Psychology, 31 (2), 232-252. Routine Human-Competitive Machine Intelligence. [39] Lieto, A., Bhatt, M., Oltramari, A., & Vernon, D. Norwell, MA: Kluwer Academic Publishers. (2018a). The role of cognitive architectures in general [23] Legg, S., & Hutter, M. (2007). A collection of defini- artificial intelligence. Cognitive Systems Research, 48, tions of intelligence. Frontiers in Artificial Intelligence 1-3. and applications , 157 . [40] Lieto, A., Lebiere, C., & Oltramari, A. (2018b). The [24] Spearman, C. (1923). The Nature of Intelligence and knowledge level in cognitive architectures: Current the principles of cognition . : MacMillan. limitations and possible developments. Cognitive Sys- [25] Stankov, L. (2002). g: A Diminutive General. In R. tems Research , 48 , 39-55. Sternberg & E. Grigorenko (Eds.), The General Factor [41] Laird, J. E. (2012). The Soar cognitive architecture . of Intelligence. How General is it? (pp. 19-38). Mah- MIT press. wah: Lawrence Erlbaum. [26] Borsboom, D., & Dolan, C. V. (2006). Why g is not an adaptation: A comment on Kanazawa (2004). Psychological Review, 113 (2), 433-437. [27] Burkart, J. M., Schubiger, M. N., & van Schaik, C. P. (2017). The evolution of general intelli- gence. Behavioral and Brain Sciences , 40, 1-24 . [28] Pretz, J. E., & Sternberg, R. J. (2005). Unifying the Field: Cognition and Intelligence. In R. J. Sternberg & J. E. Pretz (Eds.), Cognition and Intelligence (pp. 306- 318). Cambridge, UK: CUP . [29] Deary, I. J., Penke, L., & Johnson, W. (2010). The neuroscience of human intelligence differ- ences. Nature Reviews Neuroscience , 11 (3), 201-211. [30] Haier, R. J. (2017). The neuroscience of intelligence . Cambridge University Press.

Proceedings of the Society for the Study of Artificial Intelligence and Simulation of Behaviour 2018 8