Session Detail

Back

Talk Session 2-1: AI, Robotics and Philosophy

Sep. 2, 2017 08:30 AM - 10:00 AM

Room: Archimedes Room
Session chair: Kevin Kimble
Nvmsim: a Computer-aided-design Tool for Non-volatile Memory Based Cognitive Computing Hardware

Presentation Number:211.01Time:08:30 - 08:45Abstract Number:0176
Darsen D. Lu1, Huai-Kuan Zeng1, Yi-Ci Wang1, Fu-Xiang Liang1
Institute of Microelectronics, National Cheng Kung University, Tainan, Taiwan,

Recently, neuron-like software computational systems have been successfully demonstrated for applications such as computer vision, speech recognition, machine translation, robotics, medical image processing, etc. On the other hand, neuromorphic circuits attempt to mimic the operation and topology of biological cognitive systems in hardware. Neuromorphic circuits, or cognitive computing, will likely be adopted in the near future for these tasks due to the significantly better speed and power efficiency compared to software realization. Computer-aided-design is crucial for realizing neuromorphic systems in hardware due to its complexity. The simulation of neuromorphic circuits must take into account the basic semiconductor device’s behavior accurately. At the same time, it must also be carried out over thousands of training cycles in a system containing millions of neurons and synapses. Traditional SPICE circuit simulation tool models device behavior accurately but lacks the capability to handle large neural networks. On the other hand, simulation tools for digital logic is not suitable for neuromorphic systems, which uses analog computation when implemented most efficiently. We have developed NVMLearn, a computer-aided-design tool for neuromorphic circuits. NVMLearn is developed for neuromorphic circuits that utilize non-volatile memory as the basic “synapse” element, which stores information about the significance of each neuron-to-neuron connection. In order to accurately describe non-volatile memory semiconductor device, NVMLearn take in a Verilog-A compact model for the non-volatile memory as input. NVMLearn also takes inputs related to a specific neural network topology, such as how neurons are connected, mathematical functions that describes the propagation of neuron signals, and each neuron’s learning behavior. With the new tool, such neuromorphic circuits can be simulated in an efficient manner. It is also able to predict the speed and power consumption of the hardware when implemented.


 
Inter-individual Differences in Consciousness Development Via a Child-robot Scenario

Presentation Number:211.02Time:08:45 - 09:00Abstract Number:0125
Irini Giannopulu1, Tomio Watanabe2
1Humanities and Social Sciences, Bond University, Gold Coast, QLD, Australia
2Systems Engineering, Okayama Prefectural University, Okayama, Japan


The investigation of consciousness is one of the enigma facing the scientific perspective. There are many theories of what consciousness is. According to one of them, consciousness is defined as the having of perception, feelings and thoughts of the internal and the external world which can be verbal expressed. Multimodal verbal and nonverbal interactions are joined with emotions in a continuous dynamic neuronal complex that constitutes consciousness. Combined cognitive neuroscience and engineering knowledge, we have used a perception-action scenario to analyse conscious processes in association with emotion and oral language in neurotypical children aged 6 and 9 years old. The actor was always a child; the perceiver was a human or a robot. Unconscious emotional expression reflected in physiological data, i.e., heart rate, as well as conscious process mirrored on behavioral data, i.e., number of words and reported feelings, were considered. The results showed that 1) the heart rate was higher for children aged 6 years old than for children aged 9 years old when the InterActor was the robot; 2) the number of words expressed by both age groups was higher when the InterActor was the human. Contrary to children with ASD (Giannopulu et al., 2016), neurotypical children would prefer verbally to interact with humans, as human complex nonverbal and verbal behavior does not constitute an obstacle for them in communication. The present findings would be associated with the development of un/consciousness. Nonverbal behavior expressed by heart rate is an unconscious automatic activity which, in our case, depends on the InterActor Robot. Verbal behavior given by the words pronounced by the children is a conscious activity which depends on Human InterActor. Unconscious and conscious processes would not only depend on natural environments, i.e., humans, but also on artificial environments, i.e., robots.


 
Surprise and Narrative in An Automatic Narrative Generation Game

Presentation Number:211.03Time:09:00 - 09:15Abstract Number:0055
Jumpei Ono1, Takashi Ogata2
1Graduate school of software and infrmation science, Iwate Prefectural University, Takizawa, Iwate, Japan
2Faculty of software and infrmation science, Iwate Prefectural University, Takizawa, Iwate, Japan


“Automatic Narrative Generation Game (ANGG)” by the authors generates stories through the interaction between two mechanisms: “Game Master (GM)” and “Player (PL)”. The main elements of the story are “world setting”and “scene sequence”. The former includes characters, objects, locations, times and restricts that define possible ranges in their elements. The main element in each scene is an “event”. The story generation is performed using an “Integrated Narrative Generation System (INGS)”developed by the authors. A current focus of this system is to introduce the emotion of “surprise”to propose a function for making more interesting stories to the ANGG. According to Descartes, “surprise” means a strong and temporary emotion with the sudden appearance of an unexpected event. The authors considered that the incremental change by the GM and the PL (in many case, PLs) is partially driven by the PL’s function that gives surprise to the GM through the story’s change beyond the GM’s expectation. For incorporating the function of surprise into the ANGG, the authors developed a simple story generation mechanism using the INGS, which incrementally changes a story using semantic gap between a first story and a changed story. Further, the authors have confirmed a correspondence relation between the gap and the degree of surprise, and the degree of surprise can adjusted according to the gap. In this paper, the authors present a set of story techniques to make gap and surprise, from a first story by the GM and develop a method for control the use of their story techniques. Surprise based on gap is produced through various types of change of a story regarding a verb, an event, a part of a story and an entire story. Further, the effectiveness of the above two mechanisms will be evaluated by real subjects.


 
From Theoretical Perceptions of Metalogy to Analyze the Grit and Mindset Theory, a Qualitative Research

Presentation Number:211.04Time:09:15 - 09:30Abstract Number:0091
Wei-Chun Li
Department of Education, National Taitung University, Taitung, Taiwan,

The purpose of this research is to base on theoretical views of Metalogy to analyze the important concepts of the Grit and Mindset theory. This research employs literature analysis from many empirical papers and books about the Grit and Mindset theory. Computer-assisted qualitative data analysis software Nvivo 10 is applied to process and analyze three phases of data. There are three major findings for this research. Firstly, the fixed and growth mindset and the gritty characteristics all conform the Janusian thinking of Metalogy theory. Secondly, the gritty people always not only have bright and growth mindset to success, but creative thinking and meta-thinking of Metalogy maybe the key mediator variables. Thirdly, the three main theoretical views of Metalogy could be explain how the people with the growth mindset and grit to achieve their respective objective.


 
Hallucination and Phenomenal Presence

Presentation Number:211.05Time:09:30 - 09:45Abstract Number:0086
Kevin Kimble
Department of Philosophy, National Chung Cheng University, Chiayi, Taiwan,

Some significant recent work in philosophy of mind attempts to wrestle anew with what Susanna Schellenberg calls the hallucination question-- how do we explain the phenomenology of hallucinatory experience? Apart from negative disjunctivism, three general approaches have been pursued with a view to answering this question. According to the classical view, in undergoing a hallucinatory experience, a subject S is aware of some kind of particular, mind-dependent mental item or relatum, such as sense data or qualia (Chalmers, Robinson). According to the relational view, S is aware of some kind of extra-mental, mind-independent item or relatum. Candidate examples include external world properties or universals, states of affairs, propositions, or Meinongian objects (Byrne, Dretske, Smith, Tye). Finally, the no-awareness view denies that S stands in an awareness relation to any item or relatum, regardless of whether that item is construed as a mind-independent entity or property or some kind of mental item (Pautz, Schellenberg). I argue against the no-awareness view, offering considerations based on the nature of phenomenal-perceptual judgment for the conclusion that hallucinatory experience does involve an awareness relation to some kind of existing entity. Along the way, I criticize the arguments advanced by Pautz and Schellenberg for the no-awareness view. Then I argue, against the relational view, that the items we are aware of in hallucinatory experience cannot be explained by appeal to mind-independent universals, states of affairs, propositions, or mere intentional objects. In developing this line of thought, I criticize arguments against the relational view set forth by Pautz and Schellenberg, but I go on to argue that the view is nevertheless implausible on phenomenological grounds. This sets the stage for a defense of a particular version of the classical view-- the phenomenology of hallucinatory experience is best explained in terms of one’s awareness of phenomenal qualia.


 
Psycholinguistic Determinants of Object Naming in Thai for a Subset of the Bank of Standardized Stimuli

Presentation Number:211.06Time:09:45 - 10:00Abstract Number:0111
A. J. Benjamin Clarke1, Jason D. Ludington2
1Language Institute, Thammasat University, Pathum Thani, Thailand
2Faculty of Psychology, Chulalongkorn University, Bangkok, Thailand


Selecting suitable stimuli for investigations into cognitive processes in memory and language is an important process, given the natural variability for words and pictures on many psycholinguistic dimensions (e.g., name agreement, age of acquisition). Normative databases are vitally important and help maintain the necessary level of control over psycholinguistic dimensions when selecting stimuli for experimental purposes. Although norms have been obtained for several different languages, none are currently available for Thai. In the present study, 584 Thai university students provided norms for the 480-item Bank of Standardized Stimuli (BOSS; Brodeur, Dionne-Dostie, Montreuil, & Lepage, 2010), a picture set containing high resolution colour photographic images of common objects. Norms were obtained on seven psycholinguistic dimensions: name agreement, category agreement, image agreement, visual complexity, object familiarity, age of acquisition, and two types of manipulability (ease of grasping & ease of miming). Object naming latencies were also obtained from a separate group of participants (n = 32) on 332 items, after excluding items for low name agreement. The effects of the normative variables on object naming latencies were considered using multiple regression analyses and revealed that age of acquisition, object familiarity, name agreement, and category agreement were the major determinants of object naming speed in Thai, accounting for around 40% of the variance in naming latency. Age of acquisition and name agreement have also been shown to be robust predictors of picture naming speed in other languages (e.g., Alario et al., 2004; Bakhtiar et al., 2013; Bonin et al., 2004). The interpretation of the observed effects is discussed both cross-culturally and in relation to theories of lexical access during speech production. It is anticipated that the Thai psycholinguistic database, containing both normative data and object naming latencies, will be of interest to researchers working in the fields of cognition, psycholinguistics, and neuropsychology.