Search

This chapter contains sections titled: Background The Turing Test Simulation of Human Expertise Neural Networks Definition of Intelligence Intelligence, the Scientific Method, and Evolution Evolving Artificial Intelligence References Chapter 1 Exercises

Modern day computer vision tasks requires efficient solution to problems such as image recognition, natural language processing, object detection, object segmentation and language translation. Symbolic Artificial Intelligence with its hard coding rules is incapable of solving these complex problems resulting in the introduction of Deep Learning (DL) models such as Recurrent Neural Networks and Convolutional Neural Networks (CNN). However, CNNs require lots of training data and are incapable of recognizing pose and deformation of objects leading to the introduction of Capsule Networks. Capsule Networks are the new sensation in Deep Learning. They have lived to this expectation as their performance in relation to the above problems has been better than Convolutional Neural Networks. Even with this promise in performance, lack of architectural knowledge and inner workings of Capsules serves as a hindrance for researchers to take full advantage of this breakthrough. In this paper, we provide a comprehensive review of the state of the art architectures, tools and methodologies in existing implementations of capsule networks. We highlight the successes, failures and opportunities for further research to serve as a motivation to researchers and industry players to exploit the full potential of this new field. The main contribution of this survey article is that it explains and summarizes significant current state of the art Capsule Network architectures and implementations.

Interpretability, transparency, and auditability of machine learning (ML)-driven investment has become a key issue for investment managers as many look to enhance or replace traditional factor-based investing. The authors show that symbolic artificial intelligence (SAI) provides a solution to this conundrum, with superior return characteristics compared to traditional factor-based stock selection, while producing interpretable outcomes. Their SAI approach is a form of satisficing that systematically learns investment decision rules (symbols) for stock selection, using an a priori algorithm, avoiding the need for error-prone approaches for secondary explanations (known as XAI). The authors compare the empirical performance of an SAI approach with a traditional factor-based stock selection approach, in an emerging market equities universe. They show that SAI generates superior return characteristics and would provide a viable and interpretable alternative to factor-based stock selection. Their approach has significant implications for investment managers, providing an ML alternative to factor investing but with interpretable outcomes that could satisfy internal and external stakeholders. Key Findings ▪ Symbolic artificial intelligence (SAI) for stock selection, a form of satisficing, provides an alternative to factor investing and overcomes the interpretability issues of many machine learning (ML) approaches. ▪ An SAI that could be applied at scale is shown to produce superior return characteristics to traditional factor-based stock selection. ▪ SAI’s superior stock selection is examined using notional visualizations of its decision boundaries.

The RefPerSys open source software (see http://refpersys.org/ ….) is a symbolic artificial intelligence system (« expert system shell », GPLv3+ licensed, for Linux) inspired by the work of Jacques Pitrat, with emphasis on declarative rules, meta programming, C++ code generation, orthogonal persistence, reflection, Pthread parallelism, machine learning. Major themes of RefPerSys are : declarative knowledge, expert system rules, etc...It is developed by a small team of several passionate Linux developers (in France, India, etc.…) at home. In october 2022, it has about 31KLOC of C++ code on https://github.com/RefPerSys/RefPerSys/ , some (and more and more) of them self generated by RefPerSys from higher level formalisms. A rewrite in C has been attempted on in November 2021 https://github.com/RefPerSys/refpersys-in-c We are seeking partners, co-developers, applications and funding related to this software. Applicative topics could be in health-care software, computer assisted teaching, smart cities, smart grids, software engineering, chemistry and would require cooperation with domain experts. Funding could be thru ITEA (see https://itea .org/ ….) or HorizonEurope3 or ANR or others funding sources.If interested, contact Basile Starynkevitch by email to basile.starynkevitch@cea.fr (employed at CEA, LIST, see https://list.cea.fr/ ...) and basile@starynkevitch.net (or on my mobile +33 6 8501 2359 in France); Le logiciel RefPerSys (« reflexive persistent system », voir http://refpersys.org/ ….) est un logiciel libre d'intelligence artificielle symbolique (pour Linux, sous licence GPLv3+) en cours de développement, inspiré par les travaux de feu Jacques Pitrat 1 (1934-oct.2019), actuellement sans financement et développé sur le temps libre de ses principaux développeurs. Les idées directrices de RefPerSys sont : les connaissances déclaratives, moteur d'inférence, les métaconnaissances, la métaprogrammation, le raisonnement symbolique, les règles d'inférence (« systèmes experts »), la persistance orthogonale, la réflexivité, le parallélisme par threads POSIX, et l'apprentissage machine …. En octobre 2022, RefPerSys a environ 31000 lignes de code C++ (dont une partie, qu'on veut croissante, est générée par RefPerSys à partir de formalisations de plus haut niveau) sur https://github.com/RefPerSys/RefPerSys Une réecriture en C de RefPerSys a été tentée en octobre-novembre 2021 sur https://github.com/RefPerSys/refpersys-in-c Nous recherchons des applications (en médecine, « smart city », « smart grid », génie logiciel, chimie, ou dans les domaines de l'énergie ou de la culture …), des partenaires et un financement pour ce logiciel. Ce financement pourrait être par exemple fourni par l'ANR (agence nationale pour la recherche https://anr.fr/ ...) ou par ITEA (https://itea4.org/ ….) Si vous êtes intéressés, contactez-moi, Basile Starynkevitch 2 , par courriel vers

The main problem for AI consciousness is to operate within the right kind of AI. We distinguish between the traditional computing (GOFAI), and the computing based on stochastic pattern optimization. The latter will be called here computing at the edge of chaos. Optimization of learning patterns, which is the gist of its success, often happens between the areas of too much repetitive order and those of hard to predict and control stochastic processes. This is to change the focus from the opposition of symbolic versus sub-symbolic computing; symbols can appear at different granularities and the hedge between The Physical Symbol System Hypothesis and neural nets seems no longer the most productive cut to make. Computing at the edge of chaos is promising for AGI, especially for AGI consciousness. The second problem for AI consciousness is to work with the right definitions of consciousness.

International audience; A high utility itemset mining problem is the question of recognizing a set of items that have utility values greater than a given user utility threshold. This generalization of the classical problem of frequent itemset mining is a useful and well-known task in data analysis and data mining, since it is used in a wide range of real applications. In this paper, we first propose to use symbolic Artificial Intelligence for computing the set of all closed high utility itemsets from transaction databases. Our approach is based on reduction to enumeration problems of propositional satisfiability. Then, we enhance the efficiency of our SAT-based approach using the weighted clique cover problem. After that, in order to improve scalability, a decomposition technique is applied to derive smaller and independent sub-problems in order to capture all the closed high utility itemsets. Clearly, our SAT-based encoding can be constantly enhanced by integrating the last improvements in powerful SAT solvers and models enumeration algorithms. Finally, through empirical evaluations on different real-world datasets, we demonstrate that the proposed approach is very competitive with state-of-the-art specialized algorithms for high utility itemsets mining, while being sufficiently flexible to take into account additional constraints to finding closed high utility itemsets.

A smart home is a residence equipped with technologies that facilitate monitoring of residents, promote independence and increase the quality of life. In general, smart homes are used to control the operations of the home environment and, to automatically adapt it to its inhabitants’ needs. The smart home reasoning system (SHRS) is in charge of determining the automatic control and adaptation operations of the home system. Recently, there has been extensive research concerning different aspects of the SHRS. However, there is a clear lack of systematic investigation targeted at these systems. To close the gap, this paper explores the SHRS domain. For this reason, we applied the systematic literature review (SLR) method by conducting automatic and manual searches on six electronic databases, and in-depth analysis of 135 literature. From the SRL, this paper identifies about 43% of smart homes are designed to provide general home automation services. It also presents twelve major requirements of an SHRS. In addition, the study finds out that 55.5% of the research contributions in SHRS domain are conceptual and, 51.5% of them are based on symbolic artificial intelligence techniques. Further, it characterizes the usage and application trends of different reasoning techniques in smart home domain and, evaluates the major assumptions, strengths, and limitations of the proposed systems in the literature. Additionally, it discusses the challenges of reasoning in ambient assisted living environments. Finally, it underlines in the importance of utilizing hybrid reasoning approaches and the need to handle overlapping, simultaneous and conflicting multiple inhabitants’ activities and goals.

In good old-fashioned artificial intelligence (GOFAI), humans specified systems that solved problems. Much of the recent progress in AI has come from replacing human insights by learning. These learning systems are usually built by humans. Yet there is no reason to believe that humans are particularly good at defining such systems: we may expect learning to be better if we learn it. Recent research in machine learning has started to realize the benefits of that strategy. We should thus expect this to be relevant for neuroscience: how could the correct learning rules be acquired? Cognitive science has long shown that humans learn-to-learn. Here we discuss ideas across machine learning, neuroscience, and cognitive science that matter for the principle of learning-to-learn.

Artificial intelligence (AI) research enjoyed an initial period of enthusiasm in the 1970s and 80s. But this enthusiasm was tempered by a long interlude of frustration when genuinely useful AI applications failed to be forthcoming. Today, we are experiencing once again a period of enthusiasm, fired above all by the successes of the technology of deep neural networks or deep machine learning. In this paper we draw attention to what we take to be serious problems underlying current views of artificial intelligence encouraged by these successes, especially in the domain of language processing. We then show an alternative approach to language-centric AI, in which we identify a role for philosophy., Comment: 23 pages, 1 Table

It is clear from the implementation and analysis of the performance of the game Defend and Gather and the many other examples discussed in this chapter that neural-based agents have the ability to overcome some of the shortcomings associated with implementing classical AI techniques in computer game design. Neural networks can be used in many diverse ways in computer games ranging from agent control, environmental evolution, to content generation. As outlined in Section 3 of this chapter, by following the neural network development process, adding a neural network to a computer game can be a very rewarding process. Neural networks have proven themselves viable for agent design, but there are still many unexplored avenues that could prove to benefit from neural networks in computer games. The area of content generation has only briefly been discussed in recent research. The potential is that neural networks could generate entire worlds or even entire computer games based on human players' preferences. Neural networks have great potential for designing computer games and technology that will entertain players in terms of newly generated content and increasing challenge as the players learn the game.

International audience; In the context of the IA.TROMED project we intend to develop and evaluate original algorithmic methods that will rely on semantic enrichment of embeddings by combining new deep learning algorithms, such as models founded on transformers, and symbolic artificial intelligence. The documents’ embeddings, the graphs’ embeddings of biomedical concepts, and patients’ embeddings, all of them semantically enriched with aligned formal ontologies and semantic networks, will constitute a layer that will play the role of a queryable and searchable knowledge base that will supply the IA.TROMED’s clinical, predictive, and iatrogenic diagnosis support module.

Modern deep learning systems for medical image classification have demonstrated exceptional capabilities for distinguishing between image based medical categories. However, they are severely hindered by their ina-bility to explain the reasoning behind their decision making. This is partly due to the uninterpretable continuous latent representations of neural net-works. Emergent languages (EL) have recently been shown to enhance the capabilities of neural networks by equipping them with symbolic represen-tations in the framework of referential games. Symbolic representations are one of the cornerstones of highly explainable good old fashioned AI (GOFAI) systems. In this work, we demonstrate for the first time, the emer-gence of deep symbolic representations of emergent language in the frame-work of image classification. We show that EL based classification models can perform as well as, if not better than state of the art deep learning mod-els. In addition, they provide a symbolic representation that opens up an entire field of possibilities of interpretable GOFAI methods involving symbol manipulation. We demonstrate the EL classification framework on immune cell marker based cell classification and chest X-ray classification using the CheXpert dataset. Code is available online at https://github.com/AriChow/EL.

Generative models are constructs that, given proper input information and a program, can synthesize design alternatives. Earlier, computational approaches were based on techniques from symbolic artificial intelligence and operations research, primarily, to automate the generation of architectural layouts. Recently, a variety of computational techniques from different fields, such as agent-based modeling or deep learning, have been incorporated into the development of new generative models. We are reaching a point where we might not even have to explicitly program a generative model but, instead, have the model learn automatically from data or experience. However, the lack of consistency and rigor in the treatment of underlying ideas and terms is a barrier in the teaching of generative design. To address this, we introduce a taxonomy of the components of generative models, which are organized in three levels according to their function: expression, synthesis, or learning. Existing or novel generative models can be represented in the taxonomy as a combination of multiple building blocks. We introduced this taxonomy in computational design courses at Carnegie Mellon University, and it has resulted in interesting projects——some of which are highlighted in this chapter as examples.

Businesses are today moving from a data-driven focus for becoming ‘Insight-driven’ businesses. This paradigm shift hinges on the use of artificial intelligence technology. This shift needs accurate and smooth input of all pertinent details. This is facilitated by using artificial intelligence to provide intelligent detail extraction from documents and other sources. The techniques behind building an intelligent detail extraction product are discussed. The path from an initial solution through identifying and developing the characteristics that mark a product are described. A powerful synergy is created by appropriately using symbolic machine learning as well as deep learning techniques.

A well-studied trait of human reasoning and decision-making is the ability to not only make decisions in the presence of contradictions, but also to explain why a decision was made, in particular if a decision deviates from what is expected by an inquirer who requests the explanation. In this paper, we examine this phenomenon, which has been extensively explored by behavioral economics research, from the perspective of symbolic artificial intelligence. In particular, we introduce four levels of intelligent reasoning in face of contradictions, which we motivate from a microeconomics and behavioral economics perspective. We relate these principles to symbolic reasoning approaches, using abstract argumentation as an exemplary method. This allows us to ground the four levels in a body of related previous and ongoing research, which we use as a point of departure for outlining future research directions.

The pervasive use of artificial intelligence and neural networks in several different research fields has noticeably improved multiple aspects of human life. The application of these techniques to machines has made them progressively more “intelligent” and able to solve tasks considered extremely complex for a human being. This technological evolution has deeply influenced the way we interact with machines. Purely symbolic artificial intelligence and techniques like ontologies, have also been successfully used in the past applied to robotics, but have also shown some limitations and failings in the knowledge construction task. In fact, the exhibited “intelligence” is rarely the result of a real autonomous decision, but it is rather hard-encoded in the machine. While a number of approaches have already been proposed in literature concerning knowledge acquisition from the surrounding environment, they are either exclusively based on low-level features or they involve solely high-level semantics-based attributes. Moreover, they often don’t use a general high-level knowledge base for grounding the acquired knowledge. In this contexts, the use of semantics technologies, such as ontologies, is mostly employed for action-oriented tasks. In this article we propose an extension of a novel approach for knowledge acquisition based on a general semantic knowledge-base and the fusion of semantics and visual information by means of neural networks and ontologies. The proposed approach has been implemented on a humanoid robotic platform and the experimental results are shown and discussed.

Intelligent and autonomous agents is a subarea of symbolic artificial intelligence where these agents decide, either reactively or proactively, upon a course of action by reasoning about the information that is available about the world (including the environment, the agent itself, and other agents). It encompasses a multitude of techniques, such as negotiation protocols, agent simulation, multi-agent argumentation, multi-agent planning, and many others. In this paper, we focus on agent programming and we provide a systematic review of the literature in agent-based programming for multi-agent systems. In particular, we discuss both veteran (still maintained) and novel agent programming languages, their extensions, work on comparing some of these languages, and applications found in the literature that make use of agent programming.

Neuro-Symbolic Artificial Intelligence -- the combination of symbolic methods with methods that are based on artificial neural networks -- has a long-standing history. In this article, we provide a structured overview of current trends, by means of categorizing recent publications from key conferences. The article is meant to serve as a convenient starting point for research on the general topic., Comment: under review

Lean Manufacturing is well known as an effective means to improve productivity and decrease costs of operations, using a series of management practices developed first in Japan and then adapted to worldwide circumstances. Rapid technological progress has opened new business potentials and opportunities forcing companies to constantly introduce more and more advanced solutions to remain competitive. However, the potential of such technologies to support the implementation of Lean Manufacturing is not completely perceived yet. After defining Lean 4.0 which aims to provide useful insight for the integration of lean and Industry 4.0 in the manufacturing companies, AI is then briefly introduced, as a key enabler. This paper focuses on the integration and support of AI to Lean 4.0, mainly investigating the process and control issues through better use of data and knowledge.

The paper presents the implementation of an intelligent decision support system (IDSS) to solve a real manufacturing problem at JSC “Savushkin Product”. The proposed system is intended to control the quality of product labeling, based on neuro-symbolic artificial intelligence, namely integrating deep neural networks and semantic models. The system perform localization and recognition of images from a high-speed video stream and is based on several deep neural networks. Semantic networks fulfill intelligent processing of recognition results in order to generate final decision as regards the state of the production conveyor. We demonstrate the performance of the proposed technique in the real production process. The main contribution of this paper is a novel view at the creation of a real intelligent decision support system, which combines bio inspired approach, namely neural networks and conventional technique, based on a knowledge base.

Although Blockchain is still an emerging technology it has the potential to serve as a general purpose information technology platform. Already, smart contract / chaincode platforms, such as Ethereum and Hyperledger Fabric, provide support for the execution of arbitrary computations. However, the suitability of these platforms for specifying and enforcing data and service usage constraints (e.g., usage policies, regulatory obligations, societal norms) and providing guarantees with respect to conformance has yet to be determined. In order to address this gap, in this position paper we argue that symbolic artificial intelligence techniques in the form of semantic technology based policy languages and business process conformance tools and techniques, can together be used to provide guarantees with respect to the behaviour of autonomous smart contract / chaincode applications.

Reinforcement learning (RL) agents are often designed specifically for a particular problem and they generally have uninterpretable working processes. Statistical methods-based agent algorithms can be improved in terms of generalizability and interpretability using symbolic Artificial Intelligence (AI) tools such as logic programming. In this study, we present a model-free RL architecture that is supported with explicit relational representations of the environmental objects. For the first time, we use the PrediNet network architecture in a dynamic decision-making problem rather than image-based tasks, and Multi-Head Dot-Product Attention Network (MHDPA) as a baseline for performance comparisons. We tested two networks in two environments ---i.e., the baseline Box-World environment and our novel environment, Relational-Grid-World (RGW). With the procedurally generated RGW environment, which is complex in terms of visual perceptions and combinatorial selections, it is easy to measure the relational representation performance of the RL agents. The experiments were carried out using different configurations of the environment so that the presented module and the environment were compared with the baselines. We reached similar policy optimization performance results with the PrediNet architecture and MHDPA; additionally, we achieved to extract the propositional representation explicitly ---which makes the agent's statistical policy logic more interpretable and tractable. This flexibility in the agent's policy provides convenience for designing non-task-specific agent architectures. The main contributions of this study are two-fold ---an RL agent that can explicitly perform relational reasoning, and a new environment that measures the relational reasoning capabilities of RL agents.

Machine learning (ML) is a subfield of artificial intelligence, the science and engineering of making intelligent machines.2 One of the pioneers of artificial intelligence, Arthur Samuel, defined machine learning as a “field of study that gives computers the ability to learn without being explicitly programmed.”3 Figure 1-1 shows the relationship between artificial intelligence, machine learning, and deep learning. Artificial intelligence (AI) encompasses other fields, which means that while all machine learning is AI, not all AI is machine learning. Another branch of artificial intelligence, symbolic artificial intelligence, was the predominant AI research paradigm for much of the 20th century.4 Symbolic artificial intelligence implementations are referred to as expert systems or knowledge graphs which are in essence rules engines that use if-then statements to draw logical conclusions using deductive reasoning. As you can imagine, symbolic AI suffers from several key limitations; chief among them is the complexity of revising rules once they are defined in the rules engine. Adding more rules increases the knowledge in the rules engine, but it cannot alter existing knowledge.5 Machine learning models on the other hand are more flexible. They can be retrained on new data to learn something new or revise existing knowledge. Symbolic AI also involves significant human intervention. It relies on human knowledge and requires humans to hard-code the rules in the rules engine. On the other hand, machine learning is more dynamic, learning and recognizing patterns from input data to produce the desired output.

timestamp: Mon, 20 Jul 2020 12:38:52 +0200 biburl: https://dblp.org/rec/conf/ijcai/RaedtDMM20.bib bibsource: dblp computer science bibliography, https://dblp.org ispartof: pages:4943-4950 ispartof: Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI 2020 pages:4943-4950 ispartof: International Joint Conference on Artificial Intelligence, IJCAI 2020 location:Online date:7 Jan - 15 Jan 2021 status: published

There has been a rapid growth in the use of symbolic representations along with their applications in many important tasks. Symbolic representations, in the form of Knowledge Graphs (KGs), constitute large networks of real-world entities and their relationships. On the other hand, sub-symbolic artificial intelligence has also become a mainstream area of research. This workshop brought together researchers to discuss and foster collaborations on the intersection of these two areas.

Machine learning models have been successfully used in many scientific and engineering fields. However, it remains difficult for a model to simultaneously utilize domain knowledge and experimental observation data. The application of knowledge-based symbolic AI represented by an expert system is limited by the expressive ability of the model, and data-driven connectionism AI represented by neural networks is prone to produce predictions that violate physical mechanisms. In order to fully integrate domain knowledge with observations, and make full use of the prior information and the strong fitting ability of neural networks, this study proposes theory-guided hard constraint projection (HCP). This model converts physical constraints, such as governing equations, into a form that is easy to handle through discretization, and then implements hard constraint optimization through projection. Based on rigorous mathematical proofs, theory-guided HCP can ensure that model predictions strictly conform to physical mechanisms in the constraint patch. The performance of the theory-guided HCP is verified by experiments based on the heterogeneous subsurface flow problem. Due to the application of hard constraints, compared with fully connected neural networks and soft constraint models, such as theory-guided neural networks and physics-informed neural networks, theory-guided HCP requires fewer data, and achieves higher prediction accuracy and stronger robustness to noisy observations., Comment: 31 pages, 20 figures, 1 table

컴퓨터와 인간은 분명 다르지만 기본적으로 데이터를 저장하고 처리하는 개념적 측면에서는 서로 유사한 구조를 갖는다. 하지만 수집된 전체 데이터를 처리하고 분석하는 컴퓨터와는 달리 인간은 요약된 패턴 단위로 데이터를 처리한다. 즉 인간은 전체 데이터를 다루기보다는 요약된 정보를 통해 최적의 의사결정을 한다. 전체 데이터보다 요약된 정보만을 관리하면 시간과 비용 면에서 더 효율적인 시스템을 구축할 수 있다. 특히 빅데이터 환경에서 인공지능의 학습을 위한 대용량 데이터의 처리 및 분석을 위하여 요약된 정보에 기반 한 데이터학습에 대한 필요성이 제기되고 있다. 본 연구에서는 이와 같이 요약된 정보에 기반 한 심볼릭 인공지능 시스템의 효율적인 구축을 위하여 통계학의 심볼릭 데이터분석에 대하여 연구한다. 특히 대표적인 데이터언어인 R에서 제공하는 심볼릭 데이터분석 함수를 이용한 심볼릭 인공지능에 대한 방법을 소개한다. 제안방법의 성능평가를 위하여 객관적인 기계학습 데이터 사례를 이용하였다.

Artificial intelligence (AI) is an important technology that supports daily social life and economic activities. It contributes greatly to the sustainable growth of Japan's economy and solves various social problems. In recent years, AI has attracted attention as a key for growth in developed countries such as Europe and the United States and developing countries such as China and India. The attention has been focused mainly on developing new artificial intelligence information communication technology (ICT) and robot technology (RT). Although recently developed AI technology certainly excels in extracting certain patterns, there are many limitations. Most ICT models are overly dependent on big data, lack a self-idea function, and are complicated. In this paper, rather than merely developing next-generation artificial intelligence technology, we aim to develop a new concept of general-purpose intelligence cognition technology called Beyond AI. Specifically, we plan to develop an intelligent learning model called Brain Intelligence (BI) that generates new ideas about events without having experienced them by using artificial life with an imagine function. We will also conduct demonstrations of the developed BI intelligence learning model on automatic driving, precision medical care, and industrial robots., Comment: 15 pages, Mobile Networks and Applications, 2017

The presence of intelligence and rationality in Artificial Intelligence (AI) and the Internet requires a new context of analysis in which Herbert Simon's approach to the sciences of the artificial is surpassed in order to grasp the role of information in our contemporary setting. This new framework requires taking into account some relevant aspects. (i) In the historical endeavor of building up AI and the Internet, minds and machines have interacted over the years and in many ways through the interrelation between scientific creativity and technological innovation. (ii) Philosophically, minds and machines can have epistemological, methodological and ontological differences, which are based on the distinct configuration of human intelligence and artificial intelligence. Their comparison with rationality and its various forms is particularly relevant. (iii) Scientifically, AI and the Internet belong to the sciences of the artificial, because they work on designs that search for specific aims, following selected processes in order to achieve expected results. (iv) Technologically, AI and the Internet require the support of information and communication technologies (ICT). These have an instrumental role regarding the existence of AI and the Internet. Also ICT shape their diverse forms of configuration over the years. Within this framework, this paper offers a new context of analysis that goes beyond Simon's and follows four main steps: (i) the interaction between scientific creativity and technological innovation as the philosophico-methodological setting for Artificial Intelligence and the Internet; (ii) artificial intelligence and human intelligence as epistemological basis for machines and minds, where the differences between artificial intelligence and human intelligence are made explicit (under the consideration of "computational intelligence") and the analysis of minds and machines is made from the perspective of rationality ("symbolic rationality" and "adaptive rationality"); (iii) intention and its difference with design of machine learning are considered to distinguish human intelligence from artificial intelligence; and (iv) the internal and external aspects of artificial designs in contemporary society are considered through the perspective of rationality, which leads to the transition from intelligence to rationality in the Internet as well as to the historicity of information (how aims, processes, and results can be based on conceptual revolutions).

To achieve the artificial general intelligence (AGI), imitate the intelligence? or imitate the brain? This is the question! Most artificial intelligence (AI) approaches set the understanding of the intelligence principle as their premise. This may be correct to implement specific intelligence such as computing, symbolic logic, or what the AlphaGo could do. However, this is not correct for AGI, because to understand the principle of the brain intelligence is one of the most difficult challenges for our human beings. It is not wise to set such a question as the premise of the AGI mission. To achieve AGI, a practical approach is to build the so-called neurocomputer, which could be trained to produce autonomous intelligence and AGI. A neurocomputer imitates the biological neural network with neuromorphic devices which emulate the bio-neurons, synapses and other essential neural components. The neurocomputer could perceive the environment via sensors and interact with other entities via a physical body. The philosophy under the "new" approach, so-called as imitationalism in this paper, is the engineering methodology which has been practiced for thousands of years, and for many cases, such as the invention of the first airplane, succeeded. This paper compares the neurocomputer with the conventional computer. The major progress about neurocomputer is also reviewed.