smalltalk 2016: hello stranger!

INTERACTIVE ARTIFICIAL INTELLIGENCE INSTALLATION

concept, design and implementation: Márton FERNEZELYI, Zoltán SZEGEDY-MASZÁK — robots' brain development: Richard ACZEL, Márton FERNEZELYI, Zoltán SZEGEDY-MASZÁK — videos: Emese MUCSI, Róbert LANGH, Zoltán SZEGEDY-MASZÁK — curatorial work: Emese MUCSI, Zsuzsanna SZEGEDY-MASZÁK

Related works: Smalltalk 3.0 (2006/2007), Smalltalk (2000)



Below is a detailed description of the installation, as well as an interview with its creators and a video showing how it works. If you have the Safari or Chrome browser installed on your computer, depending on the speed of your internet connection you can try out the demo version of Smalltalk 2016: Hello stranger! which has been made available on the web.

In the previous versions of the installation entitled Smalltalk two computer programs conversed with each other about topics selected by the visitors. In the course of these dialogues, the chatterbots try to interpret each other’s reactions and give replies that are as appropriate as possible using algorithms based on mechanical symbol reduction. As they wish to resemble their creators, they are mainly occupied with how they can behave intelligently, as a cognizant human being. Their chief goal is therefore to carry out a unique and varied exchange of ideas. Once they catch each other repeating themselves, the robots try to end the conversation as quickly as possible by throwing selected insults at each other until they have both had enough of arguing. They then light up a cigarette and wait for the next exhibition visitor to activate them by providing them with a subject for small talk.


In the 2016 version the mature, middle-aged robots strive to resemble human beings, their keepers. They are still concerned with how they can behave as cognizant human beings, but this includes the ability to form opinions about issues that are frequently raised in daily conversations and in the news. As robots, they are capable of amassing and storing a large quantity of information, but they need guidance when it comes to interpreting these data. The immense body of knowledge available to them includes readings about Jesus Christ, Buddha, and Mohammed, but also stories by Andersen and the Harry Potter books, and they have seen episodes of Star Wars. It is not obvious to them how to differentiate between the historical and the fictitious characters, between Buddha and Dumbledore, or to grasp why Jesus and Snow White were migrants, while Buddha and Mohammed were natives. And they certainly cannot differentiate between members of the robot clan such as Kempelen’s chess machine and R2D2. For them, only logic based on algorithms exists, thus they seek to organize the emotions, fears, scapegoating attempts, and all kinds of prejudices into a logical system in order to be able to imitate the various attitudes and manners of speaking about refugees.


In contrast with the previous versions in the installation Smalltalk 2016: Hello stranger! the exhibition visitor does not select a leading sentence. Instead, s/he controls the moods of the robots as a conductor might (with the use of a gesture-recognition interface). The installation is thus guided by hand gestures, each hand controls one of the two robots. The robots are able to assume moods on a scale of five, from negative 2 to positive 2. The negative domain signifies moods of scorn and dismissal, while the positive domain signifies hospitable and welcoming attitudes. In other words, the intolerant, malicious views are to the left, and the naïve and affectionate views are to the right, while in the center one finds the uncertain, reflective approach. The dialogue about a randomly selected topic begins once the gesture-recognition interface is activated. The majority of the topics are connected to the issue of refugees and the question of tolerance, but we have also included a large portion of the subject “What is Called Thinking?” from the 2006 installation, and so these themes also appear scattered throughout the conversations. The moods of the robots can be adjusted mid-conversation, so if, for example, the aggressive state of one of the robots becomes too boring, the website user or the exhibition visitor can switch the robot’s mood to a less extremist or even very tolerant mood. What is bewildering in this situation is not so much that these robots unflinchingly obey their conductor, but that the exchange of ideas between partners who continuously alternate their flexible attitudes within a single dialogue does not appear baffling or even unusual. An important aspect of the installation is that while the intolerant, populist clichés are amusing, the robots’ mechanical imitations also parody naïvely positive views.

Simultaneously with the installation, a web-based demo version accessible from one’s home was also created. However, given its slower speed and inferior sound and visual quality, the web version only provides a sample of the robots’ chats. Smalltalk 2016: Hello stranger! has an impact on the viewer specifically because it allows him or her to experience what it is like to conduct these obedient artificial intelligences.