Since the dawn of civilization the perception of what is intelligent and creative has shifted over time. With the ever- growing society ,this in the contemporary times is referred to as Artificial intelligence. The research on AI begun long back to the 1950’s initiated by the propositions made, regarding machine learning, by Alan Turnings .This was followed by research carried out on programming languages, encoding and decoding in 1980’s and 1990’s . Then finally came the “AI winters ” when researchers lost interest in this particular specialization. The artificial Narrow Intelligence came , for the first time, in the mid-1990s’. As a result, currently we are surrounded by ANI in our smart phones, watches and even in our houses. Some theorists (Legg and Hunter) believe that Artificial intelligence is dependent on its competence and is independent of how much hard work the agent puts in when carrying out its behaviour. This statement brings out the importance of behavioural component in complex systems. Hence this research paper will first look at the possibility of imitation of the human brain then focus would be primarily on ways of making the technology -user interaction more creatively enhanced by instilling the ability of effective reasoning in the complex systems.
The complex behavioural systems are the upcoming artificial general intelligences. The intelligence which are competent enough to perform an intellectual task that humans perform , easily adapt to the environment and form its own cognitive and heuristic abilities are termed as artificial general intelligence. The article by Angela Chen(2014) takes into the consideration ‘The paper-clip maximiser Theory’ (Nick Bostrom, 2003) which states that if the goal of a task performing machine is even as simple as contriving as many paper clips as possible then it can go to extreme lengths to do so which evokes the necessity of behavioural and social systems.
According to a Journal of artificial Intelligence (Goertzel and Ben,2014) artificial leaders believe that the famous Turing tests(1950) are only partially convincing as they still have limitations to them since they were more focussed on emulating humans . Leaders are more interested in development of general purpose intelligence which could be abled and useful to perform practical things that humans can do (Nils Nilsson, 2005).A sensible approach to this problem is to utilize the knowledge of computational neuroscience for creation of the model of the human brain to use it in AGI. But for that the human brain needs to be fully understood. In 1982 and 1984, John Hopfield shared his research on artificial neural networks and general intelligence. This was followed by Salakhutdinov (2006) proposing deep belief networks which triggered the beginning of research on the imitation of the human bio-neural network in AI. Additionally, the research journal by Tie-Jun Huang (2017) discusses the possibility of assembling the neuromorphic devices to construct the artificial human brain. An apt example is how Tokai University in Japan successfully mapped the neural network of the brain of the Drosophila. Another illustrative example is IBM’s “BLUE BRAIN PROJECT” which used “Blue gene”, a supercomputer to simulate the cortical column of the rat brain. IBM scientists at Zurich announced to have discovered the first ‘artificial neurons’ with phase-change materials(2016).
Apart from this there are other mapping human brain initiatives like ‘The brain science and brain -inspired intelligence technology of China(2016-2030). Imitating the biological neural network is attainable but before that inputting the behaviourism and cognition into artificial intelligence is yet another hurdle needed to be surpassed. Zhan and Zhou(X.Y, C.L . Zhou, 2013) had designed simulation experiments and deduced that self – consciousness can be duplicated by machines. While this is still speculative, it practically remains as a possibility. “Techniques for automatic facial expression processing have been studied intensively in the pattern recognition community and the findings are highly relevant to HCI” (2004; Lyons, Budynek, & Akamatsu, 1999). The Theory of Mind is the ability to analyse, judge and infer from other’s behaviour in a social environment. While it still remains a challenge for robots , scientists believe that they can undergo development process by interacting with more people over time. A typical model of cognitive robot is said to possess the theory of mind if it displays carefully -guided intentions while taking objective -oriented action. This will also help the AI to work in interest of the user while being goal oriented at the same time. Works of Meudt’s et al. (2006) suggest that understanding behaviorism and social interaction can certainly enhance the agent and the user’s interaction. Peter Ford Dominey and Felix Warken’s paper also emphasizes on the fact that recognizing emotions also helps in building shared intentions between the robot and the agent while working towards a goal. This is essential as the outcome of a collaborated effort between an agent and a user can be a success only if the reason to adopt the specific goal is common for both. This can be explained by a descriptive example that a wardrobe full of clean clothes maybe an outcome though the agent’s main goal was to keep the house uncluttered while the user wanted clean shirts.
Another approach which elucidates the ideal human – machine interaction is the companion -system architecture( Wendemuth A, Biundo S. ,2012). This architecture has been defined as a system related to how the agent needs to use the available datasets and user’s bio physiological data with the help of measuring tools to recognize the overall affective state of the user. Thus in this human – computer interaction(HCI) the computer needs to interact with the user and not vice versa for the user’s convenience. The article written by A .Esposito and L.C. Jain (2016) elaborated on the above point by taking the example of the system enabled in the expensive modern cars.
The companion system comes handy when the logical interaction concept needs to be mapped into physical representation. In the car when the user wants to avoid traffic jams the companion system helps the user by providing it with a smaller route. The system takes into account the many equal tracks the driver passes through every day( home to office and back home). While the driver moves through the tracks, the system monitors the facial expressions at the same time. Weak confusion is mostly expressed in slight changes of facial expressions using FACS (Wallace .P.E. and Friese V. ,1976). Since the driver never complained about the announcement given by the system, the system gained from experience to continue with announcements. But the stop in the GPS when system senses the driver is becoming very expressive (angry/irritated by the announcements) and it learns the ‘new dislike’ and makes announcements only when necessary.
Deep Neural networks is another form of cognitive architecture and deep learning is increasing due to high computational power and availability of large emotional data sets. The journal of Artificial General Intelligence( Goertzel, Ben, 2014) states many exceptional examples using deep learning .This also includes Cyc (Lenat and Guha ,1989) which is an architecture using logical reasoning techniques to answer questions and derive new knowledge from old.
While there is intense research still going on enhancing the emotional and cognitive systems in AI there have been small successes by developing simplifies robotic models for basic engaging interaction between the user and the robot. The most recent development in the field of medical science is the ‘DREAM’ project (2014), supported by the European Commission, which targets at enhancing the therapy experience for children with autism spectrum disorder. NAO’s main goal is to analyse the sensory data to understand the child’s nature and accordingly assist the therapist in teaching social skills to these children via interaction. All this while the behaviour generated by the children with autism is being interpreted by this 25 degree freedom robot , the robot is also learning . Additionally ,specific ethical and legal norms are being followed when robotic therapy and interaction is being provided.
While contributions of Lewandowska – Tomaszczyk and Wilson (2016) highlight the disgust across different cultures towards the social environment consisting of socially interacting robots encoding and decoding emotion in order to interact, other researchers are of the view that the most important frontier of being assisted by robots is not only mutual learning between the user and the agent while interaction but also the evident reduction of the human error in all areas of specialization( Tingley.K, 2017). Researchers give examples like 94% accidents occurring in general aviation due to human error( Fowler, 2014) which clearly indicates the lack of situational awareness , poor problem diagnoses ,poor planning / execution / organizational functioning amongst humans etc. When it comes to human errors, these agents are needed to reduce them. Even the ongoing regular data breaches by hacking networks could be possibly reduced to minimal with the help of smart buildings consisting of and self- learning systems. This could yield a more robust cyber and physical security thereby increasing the overall productivity of the organizations( Mylrea,2015). To make these self -learning systems more modified they must refer to an emotional model which teaches them to understand (not synthesize them) emotions solely in the interest of the organizations (users). These agents must learn to argue what is right and wrong with experience(Bartneck ,2001). In right words these modified behavioural systems when put to use will learn and understand over time wherein understanding means incorporating newly acquired knowledge about the society / working of organization into their already existing body of facts(Summers Stay .D ,2017).
Firstly, readers must understand that even though we have successfully mapped the neural network of the brain of the Drosophila or simulated the cortical column of the rat brain scientists have a long way to go before coming up with something similar for human brain. While theories like ‘the paper clip maximiser’ might be true , the scientists are carrying out various researches to incorporate behaviourism into the complex system because when system is able to understand the intention behind the user assigning it with that task, this becomes the basis of shared intention and efficient working by the agent( Peter Ford Dominey and Felix Warken). The experiments like “DREAM project” (2014) gives hope that eventually the care and treatment being offered to the autistic children by robots will turnout out to be effective and that this degree of dedication shown by robots will be extended to other patients and elderly people as well. Over the years various user- computer interaction architectures have been developed like companion -system architecture (2012) and Deep neural networks(Goertzel and Ben ,2014) in order to enhance the cognitive skills of the AI and great care is being taken by the developers while designing such systems so that they not only perform basic tasks but also take into account the wider environment. Lastly even though AI has the potential to reduce the human errors the humans must remember to be more judicious with their trust on them and build systems that could perform a safety analysis on their own actions.
Esposito A., Jain L.C. (2016) : Modelling Emotions in Robotic Socially Believable Behaving Systems; Chapter-2 :Towards Robotic Socially Believable Behaving Systems ( pp 9-14, vol 105).
Esposito A., Jain L.C. (2016) : Modelling Emotions in Robotic Socially Believable Behaving Systems; chapter-6: ‘Going Further in Affective Computing: How Emotion Recognition Can Improve Adaptive User Interaction’ (pp- 73 – 103, Vol 105)
Goertzel, Ben (2014) :Journal of Artificial General Intelligence: Concept, State of the Art and Future Prospects; (pp 1-48, Vol.5, Iss1)
Mylrea M., Gourisetti S.N.G. (2017) Cybersecurity and Optimization in Smart “Autonomous” Buildings ; Autonomy and Artificial Intelligence: A Threat or Saviour? (pp 263-294)
Laird, John E; Lebiere, Christian; Rosenbloom, Paul S. (2017)A Standard Model of the Mind: Toward a Common Computational Framework Across Artificial Intelligence, Cognitive Science, Neuroscience, and Robotics(pp 18-19, Vol. 38. Iss 4)
Toward Humanlike Task-Based Dialogue Processing for Human Robot Interaction( Vol. 32, Iss 4.)
Huang , T.J (October, 2017 ). International Journal of Automation and Computing: Imitating the brain with neurocomputer a “new” way towards artificial general intelligence, Volume 14, Issue 5, pp 520–531.
- Y. Zhang, C. L. Zhou. From biological consciousness to machine consciousness: An approach to make smarter machines. International Journal of Automation and Computing, vol. 10, no. 6, pp. 498–505, 2013.
Chen, A. (11 sep, 2014).The Chronicle of Higher Education; Washington : The Chronicle Review’ : Is artificial intelligence a threat?’
*The graphic images were taken off from istocks and were added by Unisnap.
*The above article was written by Sarah Singh as research for her undergraduate degree.