2018 · 13 citations · 7 references
ComputingComputer ScienceInformation TechnologyArtificial IntelligenceInternet Of ThingsMobile ComputingConversational InterfacesConversational User InterfacesHuman-machine InterfacesIntelligent AgentsMobile Human-computer InteractionMultimodal InteractionArtificial General IntelligenceChatbot InteractionChatbot SystemsChatbotsConversational AgentsHuman IntelligenceHuman-ai Interaction On Mobile DevicesInternet Of Things Devices
Soft artificial intelligence (AI) is defined as non-sentient AI designed to perform close to human level in one specific domain. \nThis is in contrast to “Artificial General Intelligence” (AGI) \nwhich solves the problem for human level intelligence across \nall domains. Soft AI is a reality now in the new generation \nof smart Internet of Things devices like Amazon’s Alexa, Apple’s Siri or Microsoft’s Cortana, giving rise to concerns about \nprivacy and how the technology is being used. This research \nis based around an experiment in “AI as a service” where fifteen chatbot agents using Google’s “Dialogflow” are deployed \naround the Queen Elizabeth Olympic Park in London for the \ngeneral public to interact with. The physical devices are 3D \nprinted representations of creatures living in the park, designed \nto fit with the park’s biodiversity remit. Park visitors interact \nwith the creatures via their mobile phones, engaging in a conversation where the creature offers to tell them a memory in \nexchange for one of their own, while warning them that anything they say might be repeated to others. \nThe scope of the work presented here is as follows. After explaining the details of the deployment and three month study, \nthe conversational data collected from visitors is then analysed. Following a review of the current literature, techniques \nfor working with the unstructured natural language data are developed, leading to recommendations for the design of future \nconversational “chatbot” agents. The results show distinct patterns of conversation, from simple and direct “verb plus noun” \ncommands to complex sentence structure. How users interact \nwith the agents, given that they are conversing with a mechanism, is discussed and contrasted with the memories that they \nhave agreed to share. \nThe conclusion drawn from this work is that, while the current \ngeneration of devices only listen for commands from users, \nthere is a danger that smart IoT devices in the future can be \nused as active information probes unless properly understood \nand regulated. We finish with observations on privacy and security based on our experiences here.
7
Sepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997 · 98.1K citations
Sequence to Sequence Learning with Neural Networks
Ilya Sutskever, Oriol Vinyals, Quoc V. Le · arXiv (Cornell University) · 2014 · 13.4K citations · Full text
Computing, Computer Science, French +14
Efficient Estimation of Word Representations in Vector Space
Tomáš Mikolov, Kai Chen, Greg S. Corrado et al. · arXiv (Cornell University) · 2013 · 11.7K citations · Full text
TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
Martı́n Abadi, Ashish Agarwal, Paul Barham et al. · arXiv (Cornell University) · 2016 · 9.8K citations · Full text
Software Framework for Topic Modelling with Large Corpora
Petr Sojka · 2010 · 3.8K citations · Full text