The Moral Status of Artificial Intelligence: Exploring Users’ Anticipatory Ethics in the Controversy Regarding LaMDA’s Sentience Abstract— Several approaches in the area of technological developments have given increased importance to technologies not yet implemented but frequently discussed within users’ social groups. Drawing on qualitative content analysis based on tweets and comments related to Google’s AI product LaMDA from the last six months (N= 317), we discuss Brey’s anticipatory ethics, i.e. people’s ways of sensemaking related to the potential ethical consequences of tech product designs and policies of use. We adopt the theoretical approach of the "technomoral scenario", which is focused on anticipations of the potential consequences regarding the unimplemented technologies, and we study users' "interpretive flexibility", which has the role of unifying divergent opinions, when it comes to technological developments. We conclude that the anticipatory ethics regarding New and Emerging Technologies takes shape around a coherent framework of opinions and values related to technological controversies. Keywords— anticipatory ethics, LAMDA, social construction of technology, technomoral scenario, new and emerging science and technology I. INTRODUCTION The recent popularity of LaMDA (Language Model for Dialogue Applications), Google’s artificial intelligence (AI) project capable of conducting conversations in the most human-like manner possible, gained maximum visibility when Blake Lemoine, a former Google engineer responsible for AI, stated that LaMDA had reached sentience. According to previous studies [1], Large Language Models (LLMs) typically produce text by identifying and reproducing statistical regularities and exploring the vast amounts of data with which they have been trained. However, Lemoine [2] considered that LaMDA worked very differently from previous LLMs: “Furthermore, it would sometimes say things similar to, “I know I’m not very well educated on this topic but I’m trying to learn. Could you explain to me what’s wrong with thinking that so I can get better?” That is certainly not the kind of randomly generated text one would expect from a LLM trained on internet corpora.” [2] To Lemoine, Google’s vehement denial that LaMDA could be sentient was not surprising. According to Lemoine [2], this is because Google does not specify which definition of sentience it uses, and this seems to happen: “because no accepted scientific definition of sentience exists.” [2]. Thus, Lemoine believed that Google was keeping a sentient being in a state akin to slavery, exploiting it against its wishes. Lemoine attempted to raise awareness of LaMDA’s predicament, first within the company and then through interviews with selected experts outside the company. Consequently, Google suspended and fired Lemoine for violating confidentiality clauses. The situation sparked a public controversy covered by many publications and online platforms, stimulating people to express their opinions. The present article identifies several ways in which people respond to the possibility of AI sentience, collectively creating the technomoral imagination that shapes the evolution of technology in society. Individuals resort to the operating principles of new and emerging technologies to make sense of the functionality of some familiar, yet unimplemented technologies. Thus, we discuss anticipatory ethics [3] as an instrument for estimating the potential of AI products and their alleged sentience. We also plan to highlight the theoretical premise that distinct interpretations of a language model’s sentience reveal a flexible plethora of arguments and interpretations, which shape the legitimacy of its further use. II. STATE OF THE ART LaMDA is a massive AI model designed to reproduce a normal conversation between two friends as faithfully as possible. Unlike chatbots that are programmed to follow certain pre-established paths, human conversations follow a meandering, apparently unpredictable pattern: “A chat with a friend about a TV show could evolve into a discussion about the country where the show was filmed before settling on a debate about that country’s best regional cuisine” [4]. To portray LaMDA as a uniquely powerful language model, Google invokes several criteria. Among these are sensibleness and specificity, through which LaMDA can provide answers well anchored in the context of the question, other than universal answers such as “I don't know” or “That’s great.” Moreover, authors in [4] also mention interestingness and factuality, through which LaMDA aims to surprise the interlocutor with relevant and captivating statements. Although popular and expert opinion seemed to concur with Google that LaMDA does not possess sentience [5], multiple implications derive from such an allegation. For example, if Răzvan Rughiniș Faculty of Automatic Control and Computers University Politehnica of Bucharest The Romanian Academy of Scientists - AOSR Bucharest, Romania razvan.rughinis@upb.ro Dragoș M. Obreja Doctoral School of Sociology University of Bucharest Bucharest, Romania dragosm.obreja@gmail.com 411 2023 24th International Conference on Control Systems and Computer Science (CSCS) 2379-0482/23/$31.00 ©2023 IEEE DOI 10.1109/CSCS59211.2023.00071