In what sense can AI have a mind?

Workshop Description

This workshop aims to foster an interdisciplinary discussion over the question of whether machines like AIs and robots can have a mind. In particular, it proposes to approach this question by asking what types of AI can have what kinds of mind. 

On the one hand, there exist many different kinds of AIs and robots. If we classify them in terms of their functions, for example, we can find AIs for language processing, visual recognition, game playing. Some even aspire to design an AI with general intelligence comparable to human beings. Similarly, there are robots designed to assist human activities, move like humans and animals, engage in social interactions with people. One may try to design a robot that is behaviorally indistinguishable from human beings. There are many other ways to classify AIs and robots, for instance, in terms of their internal hardware structures, computational principles, or ways of interacting surrounding environments. 

On the other hand, there are many ways to characterize the mind. One may focus on its subjective aspect, stating that consciousness is a definitional feature of the mind. Others may focus on its intentional aspect, stating that world-directed cognitive faculties are the hallmark of mindedness. Some may focus on its agentive aspect, stating that autonomy is the distinctive feature of the mind. There may be other perspectives to capture characteristics of the mind. Given that it seems (at least conceptually) possible that one aspect of the mind is realized independently of other aspects, we can think of each aspect of the mind as a different kind of mind.

Given this twofold multiplicity, we can reformulate the fundamental question of whether AI and robots can have the mind as follows: what types of AI and robots can have what kinds of mind? This workshop addresses this question by facilitating an interdisciplinary discussion drawing on philosophy of mind, phenomenology, history of science, and AI/robotics.

Speakers (Affiliation): “Presentation Title”

Tim Crane (Central European University): “AI Fantasies and the AI Reality: Sceptical Reflections”

Can Artificial Intelligence create a genuine thinking machine? Philosophical discussions of AI have tended to adopt one of two extreme answers to this question: either that it is impossible in principle, or that its actual invention, using something close to current technology, will be with us within decades. In this talk I adopt a different approach: there is no reason to think genuine AI is absolutely impossible in principle, but there is also no reason to think that current AI will produce anything like it. My approach is based on two things: an understanding of the methods of AI, and scepticism about the idea of ‘general intelligence’ (and therefore about ‘artificial general intelligence’).

Tadahiro Taniguchi (Ritsumeikan University): “Symbol Emergence in Robotics: Towards Emergence of Mind through Physical and Semiotic Interaction”

Symbol emergence in robotics aims to develop a robot that can adapt to the real-worldenvironment, human linguistic communications, and acquire language from sensorimotor information alone, i.e., in an unsupervised manner. The study also aims to build a constructive model that explains the dynamic process of mental development based on an agent’s subjective experiences. This invited lecture introduces the symbol emergence system, which is a background philosophy of the symbol emergence in robotics, and the recent development of integrative probabilistic generative models for language learning, e.g., spatial concept formation with simultaneous localization and mapping, and vision of symbol emergence in robotics.

Mai Sugimoto (Kansai University): “Metaphor Guides the Direction of Research: How Computers Have Been Analogized to Brains”

The idea that machinery has its mind presupposes an analogy of the machinery to human, and early artificial intelligence researchers based their research prospects on arguments using metaphors at various levels, which have influenced the history of artificial intelligence research for more than 60 years. In this presentation, we will briefly examine how some of areas leading to the current artificial intelligence research have evolved through the use of metaphors, even before the term "artificial intelligence" was coined. This will help us understand why the machine mind debate has been going on for so many years.

Katsunori Miyahara (Hokkaido University): “On Habit and Intelligence”

Habit is often considered the paradigm of unintelligence. Yet habits do not always lead to unintelligent behavior. In fact, intelligent coping is often enabled by developing appropriate habits. But how is it possible that we cope intelligently with a situation non-deliberatively and near-automatically out of habits? Call this the puzzle of habitual intelligence. In this talk, I review several responses to this puzzle and suggest that the pragmatist approach based on John Dewey’s account of habit presented in his Human Nature and Conduct offers the most promising framework to explore the nature of habitual intelligence.

Takuya Niikawa (Kobe University): “Conscious AI and Cognitive Phenomenology”

In this presentation, I argue that disembodied AIs including my laptop and AlphaGo can think consciously. My argument is based on three premises: (1) A conscious experience of thinking can occur without any sensory experience co-occurring, (2) disembodied AI can calculate for solving equations, and (3) mathematical propositions are not about the environmental world. Considering the three premises, we can see that there is no reason to deny that disembodied AI can have a conscious experience of thinking about a mathematical problem. This presentation mainly focuses on the first phenomenological premise, examining several arguments for it.

Workshop Format

This workshop is held online using zoom on March 16th, 2021. Please fill out the registration form to participate at https://forms.gle/1C2EYa4uztxCZUtKA 

Each speaker records a video presentation (around 20 minutes) in advance, which will be shared by all the participants so that they can watch them before the workshop. The workshop is mainly dedicated to discussions. The links to each presentation file and the online meeting room will be sent to the registered email addresses until one week before the workshop date. 

Timetable

The workshop has two parts. The first part will start at 14:00 JST (5:00 UTC) and is conducted in Japanese (Professor Tim Crane does not attend the first part). The second part will start at 16:00 JST (7:00 UTC) and is conducted in English (every speaker attends the second part). The whole workshop will end around 18:00 or 19:00 JST (9:00 or 10:00 UTC). 

Workshop Organizer

Takuya Niikawa (Kobe University): https://researchmap.jp/niitaku11?lang=en

Tsuyoshi Matsuda (Kobe University): https://researchmap.jp/read0172483

Website

http://www.lit.kobe-u.ac.jp/mst/en/schedule.html

Inquiry

Please direct any inquiries (Japanese or English) to [email protected] (Takuya Niikawa).