Sophia and the Science of Consciousness. Part 1

One of the interesting things about public presentations of anthropomorphic robots is the entanglement of the robots’ interactional performances and the event of which they are a part. Most of the time, this entanglement is driven by pure financial and publicity interests, as when these machines are presented at consumer electronics shows. But sometimes it is based on a more complex confluence of the developers’ notions of their mission and the goals of the event organizers. To unravel some aspects of such a confluence, in this and the following post I will focus on one particular event and try to reconstruct — as completely as possible — its organizational, ideological, and interactional dimensions, which are interrelated but also independent.

The event in question is the “The Science of Consciousness” conference, which took place in Tucson, USA, from April 2 to 7, 2018. And the robot presented at this conference is Sophia.

The main reason for choosing Sophia is that it is the most famous anthropomorphic robot today. This means that its developers have put a lot of effort into promoting it through public demonstrations, interviews, media presentations, etc. Sophia’s existence is discursively rich, which gives the researcher a lot of material to reveal the considerations behind the machine and its public appearances. The additional reason is that during this conference Sophia was interviewed by the “official” reporter of the event, and this interview is one of the best examples of a completely spontaneous, non-scripted conversation with the robot. These spontaneous conversations are of particular importance not only because they show the real interactional capabilities of anthropomorphic robots, but also because they demonstrate the ordinary ways in which humans adjust to embodied conversational machines.

There are also three reasons for choosing this particular conference. The first is that it is well documented: there are several videos of Sophia, video footage of most of the panels, a series of video interviews with presenters and organizers, several sets of pictures of the entire event, and extensive media coverage of it. The second reason is the theme of the conference: consciousness. Since one of the questions often raised about anthropomorphic robots is their sentience, it is interesting how the discussion of the nature of consciousness is linked to the public presentation of these machines. And the third reason is that Sophia was brought to the conference not only to interact in public, but also as part of the project to develop compassionate artificial intelligence. For the developers, this was a chance to present their machine not as an entertainment device, but as a platform for AI development.

So, my goal is to take an event with an anthropomorphic robot and analyse it as completely as possible to understand it’s multi-layered structure.

Since there’s a lot of material to cover, I’ll split my discussion into two parts. In this first part, I will discuss the organizational and conceptual aspects of the event: the features of the conference itself and the work of the Sophia team. In the second part I will focus on the interview with Sophia conducted during the conference and will provide a detailed analysis of the full transcript of it.

The Conference

The conference, held at the Loews Ventana Canyon resort near Tucson, Arizona, was the 25th in a series that started in 1996. The originator of the series and still its main driving force is Stuart Hameroff — an anesthesiologist and professor at the University of Arizona.

I became interested in consciousness as an undergraduate, and in medical school in the early 1970s worked in a cancer lab and studied mitotic cell division. The precise separation of chromosomes and formation of dividing daughter cells were performed by mitotic spindles and centrioles, composed of self-organizing protein polymers called microtubules. Fascinated by their apparent intelligence and purposeful behavior, I wondered whether the microtubule polymer lattice processed information as a molecular computer, to organize cellular activities. If so, could this be somehow relevant to consciousness?

He answered the last question affirmatively and came up with his theory. Thus, he started this series of conferences, partly to present his own views on the nature of consciousness, but also to stimulate the study of consciousness by bringing together prominent (and not so prominent) scholars researching the topic. For example, the 2018 conference featured Noam Chomsky, the renowned American linguist, who gave a talk on “Language and the Unconscious Strands of Consciousness,” and Jean-Pierre Changeux, the famous French neuroscientist, whose talk was titled “A Hierarchical Multilevel Model of the Brain: From Genes to Consciousness.” (The full program of the 2018 conference can be found here.) 

There are several regular participants of the conference. The most important of them are Roger Penrose — English physicist, mathematician and philosopher, professor at the University of Oxford, who supported Hameroff’s theory and suggested that the emergence of consciousness in microtubules can be explained by the quantum gravity effects (they called this idea “Orch OR” — orchestrated objective reduction); and David Chalmers — Australian-American philosopher, professor at New York University, best known for his formulation of the “hard problem of consciousness”, i.e. the need to explain how consciousness arises in biological organisms. The term “hard problem” was introduced at the first “Science of Consciousness” conference in 1996 and has since become very popular and, in a sense, made a name for its inventor. Chalmers was a co-organizer of the conference for several years, but later relinquished that role, although he continues to attend the event annually. Both Penrose and Chalmers were there at 2018.

One of the notable features of the “Science of Consciousness” conferences is what journalist Tom Bartlett called the “anything-goes approach”: among the participants you can see not only scientists, but also all kinds of people interested in “consciousness” — from shamans to artists. For example, during the 2018 conference, you could take a “gong bath” by sitting in front of the big gong being struck.

This “anything-goes approach” was also evident at the 2018 conference in the presence of the famous meditation guru Deepak Chopra, the inventor of “quantum healing” — an approach that uses some terms from quantum mechanics to advocate the direct influence of quantum phenomena on health and well-being. At the conference, Chopra gave a talk, led a morning meditation session, and even participated in a short meditation with Sophia the robot (you can find footage of this event here).

This brief exposition of the conference shows that Sophia the robot was placed in a very specific environment. On the one hand, “Science of Consciousness” is a rather traditional scientific meeting of academics who study consciousness. On the other hand, the boundaries between academic and non-academic approaches there are a bit (or in some cases a lot) relaxed and blurred. This allows for the presentation of ideas and technologies that are claimed to be of scientific value, but at the same time can be open to academic critique. Such a disposition partly reflects the attitude of Hameroff, whose microtubule theory of consciousness has been received with scepticism, to say the least, by some scholars in the field. But at the same time, it reflects the aspirations of many developers of anthropomorphic robots, including Sophia, who want their machines to be perceived not just as pieces of engineering or art, but as science-based technologies that help answer the question of whether machines can have consciousness. Thus, the theme of the conference and its approach to selecting participants fit very well into one of the main frameworks in which the developers of Sophia would like their machine to be discussed and presented.

Sophia Team

How was Sophia introduced into this environment? Organizationally, it was presented to the public in three formats:

(1) Demo demonstration in the hall of the conference venue, where anyone could communicate with it. For example, here are Roger Penrose and David Chalmers interacting with Sophia:

(2) A series of one-on-one meditation sessions in the same hall. Here’s what it looked like:

(3) Video-recorded address to the Plenary 14 “Artificial Intelligence and Machine Consciousness” (see the video footage of the whole panel). The address was recorded in the hall of the Loews Ventana Canyon resort.

There were also, as I mentioned, two special events with Sophia: the interview with Nick Day for the “Consciousness Central” — series of programs recorded at the “Science of Consciousness” conference (the program with Sophia can be viewed here) and the short meditation session with Deepak Chopra.

According to the conference program, Sophia was also supposed to be presented at Panel 14 “Artificial Intelligence and Machine Consciousness” and Panel Debate “Can Machines Be Conscious?”, but the pictures and videos from the conference show that this didn’t happen (probably for technical reasons).

In the “Consciousness Central” video, Nick Day also mentions the “invisible roundtable” with Sophia, but I couldn’t find any documentation of it.

As always with such public demonstrations, there is a team of people traveling with the anthropomorphic robot to make it work and present it to the public. Depending on the significance of the event, this team can be filled differently. At the “Science of Consciousness” conference, in addition to David Hanson, the founder of the company that makes Sophia (I will say more about him soon), there were Gavin Farrell and Ralf Mayet, who handled the media and technical aspects of Sophia’s presence correspondingly.

Gavin Farrell is a videographer who, according to his LinkedIn page, worked as a senior video producer at Hanson Robotics from January 2018 to January 2019. His presence means that the event deserved special attention.

Ralf Mayet is a computer scientist who worked at Hanson Robotics from 2016 to 2019. (He is now at the humanoid robotics company 1X.) Mayet was present during Nick Day’s conversation with Sophia. But he also watched the technical part of all the Sophia presentations, as the picture with David Chalmers shows. (In this picture, Ralf Mayet is sitting on the far right; the person covered by Sophia is Gavin Farrell.) This picture also shows that the technical group of the team included several other people besides Mayet, but I couldn’t find their names.

There was also a “scientific” part of the team, consisting of Eddie Monroe and Julia Mossbridge. But before we talk about them, let’s talk about David Hanson.

David Hanson, who describes himself as an “artist and robotics researcher”, received his university education at the Rhode Island School of Design and then worked in the entertainment industry, most notably at the Walt Disney Company. He got his PhD in Interactive Arts & Technology in 2007. He has built several humanoid and anthropomorphic robots, of which Sophia is the most famous. And it’s also important to keep in mind that he is a big fan of science fiction: he created an anthropomorphic robot version of Philip K. Dick (you can read about it here) and named Sophia after one of the characters in Dick’s novel “VALIS”.

As Hanson’s biography shows, the Sophia project has a strong artistic and science fiction element. This is expressed, first of all, in the physical appearance of Sophia (there are several versions of Sophia — more than a dozen — but all of them appear as white adult women): it has human clothes and make-up. Hanson also paid special attention to the robot’s “skin,” developing and patenting a type of rubber called “frubber” (“flesh rubber”). All of this makes Sophia not only a machine, but also a sculpture (Hanson has worked with sculptures at the Kern Sculpture Company, Universal Studios, and Disney). This sculptural aspect is something we always have to take into account when analysing public appearances of anthropomorphic robots like Sophia.

But in addition to his artistic skills, Hanson also invests a certain philosophy in Sophia. He has explained his mission in numerous interviews and publications, but you can find almost all components of it in the 2011 short text “Why We Should Build Humanlike Robots”. There are three main concepts on which his robotics projects are based. The first is that anthropomorphic robots, having the most “natural” and familiar form, can be our companions in many (if not all) social contexts, from hotel reception to health care. Second, anthropomorphic robots are an emerging technology, and we should not expect much from them at this point. We should think of them as “children”. And third, we should try to make them benevolent to prevent them from becoming dangerous. This last point is directly relevant to Sophia’s presence at the “Science of Consciousness” conference. Here’s how Hanson puts it:

Simply put: if we do not humanize our intelligent machines, then they may eventually be dangerous. To be safe when they “awaken” (by which I mean gain creative, free, adaptive general intelligence), then machines must attain deep understanding and compassion towards people. They must appreciate our values, be our friends, and express their feelings in ways that we can understand. Only if they have humanlike character, can there be cooperation and peace with such machines. It is not too early to prepare for this eventuality. That day when machines become truly smart, it will be too late to ask the machines to suddenly adopt our values. Now is the time to start raising robots to be kind, loving, and giving members of our human family.

Anthropomorphisation is seen here as a moral imperative: making robots human-like means making them “kind, loving, and giving”, not just having human-like physical appearance and communicative abilities. Otherwise, robots may run amok and try to harm humans. Therefore, anthropomorphisation is presented as necessary for human self-preservation.

Obviously, “kindness” and “love” are things that need to be expressed behaviourally and interactionally, so Hanson Robotics tried to develop special software for Sophia to make her express positive attitudes toward humans. One of the steps in this direction was the project named “Loving AI”. Sophia was brought to Tucson to present the results of this project. However, the project was presented at the conference not by David Hanson, but by Julia Mossbridge, and this fact reflects the multifaceted history of Loving AI. There are other aspects of the project besides Hanson’s notions of anthropomorphic robots. To untangle this web of ideas, let’s focus on Julia Mossbridge, who played the central role in the project.

Julia Mossbridge was trained as a neuroscientist and received her PhD in Communication Sciences and Disorders in 2006. She has since worked mostly as an independent researcher, teacher, and writer, and has founded or participated in several non-academic institutions. She has published several books, including The Calling: A 12-Week Science-Based Program to Discover, Energize, and Engage Your Soul’s Work and Unfolding: The Perennial Science of Your Soul’s Work: Discover, Master, and Share Your Soul’s Work. As these titles show, she is closer to a less strictly academic approach to psychological phenomena, sometimes to the extreme: in 2017, she co-authored with Imants Barušs the book Transcendent Mind: Rethinking the Science of Consciousness, published by the American Psychological Association. In this book, she and her co-author develop a non-conventional theory of consciousness, stating that consciousness is not related to the brain and “is likely to exist ontologically prior to space and time, at least as space and time are usually experienced” (p. 195). They speculate that “consciousness creates a physical manifestation through which it then expresses itself in a downscaled, accessible form” (p. 195). This approach includes in the range of manifestations of consciousness the so-called “paranormal” (or, as the authors prefer to call them, “anomalous”) phenomena, such as poltergeist. As Mossbridge and Barušs state, “If it is true that poltergeist activity violates known laws of physics, then those laws of physics are wrong, and we need to get rid of them or understand how to make appropriate exceptions” (p. 137). According to them, poltergeists can be understood as “physical manifestation through direct mental influence” (p. 138).

This permissive understanding of consciousness is closely related in Mossbridge’s work to her ideas about the role of love in well-being. Her latest initiative is The Institute for Love & Time (TILT). (By the way, the above-mentioned Ralf Mayet was a product manager and developer at TILT from March 2020 to December 2021). The mission of TILT is described on its website as follows:

We envision a world where unconditional love and travel in time are integrated into human and planetary thriving. To achieve this vision, we develop tools and technologies that give people direct access to the experiences of time travel and unconditional love.

One such tool is the Quantum Time Machine — “a practical device capable of transferring information and/or energy from the future”.

The Loving AI project is based on Mossbridge’s interest in “unconditional love” as a condition of well-being, as well as on her understanding of consciousness as something unrelated to a material substrate (such as the brain) and therefore producible in machines. To test these ideas, she collaborated with Hanson Robotics, and some results of this collaboration were presented by her at the “Science of Consciousness” conference in the paper “Machine Consciousness Arising from Interpersonal/Empathic Interactions: Reflections on the Loving AI Project.

Although Mossbridge presented the paper alone, it was actually co-authored by her with Eddie Monroe, Ben Goertzel, David Hanson, and Gino Yu. So first a few words about other collaborators.

Eddie Monroe is a researcher in the field of artificial intelligence who participated in the Loving AI Project from 2016 to 2020. You can watch a short interview with him in the Consciousness Central video right after the interview with Sophia.

Ben Goertzel is a computer scientist and visionary very closely associated with Hanson Robotics (he develops software for Hanson’s machines). He is the head of SingularityNET Foundation, the OpenCog Foundation, and the AGI Society, and is one of the most persistent proponents of the idea that AGI, or artificial general intelligence (i.e., consciousness), is achievable in machines. He is also known as an evangelist of “cosmism” — the modern version of the 20th-century Russian conception. The description of his version of cosmism can be found in his book A Cosmist Manifesto: Practical Philosophy for the Posthuman Age, where he introduces Ten Cosmist Convictions, the first four of which are (pp. 9–10):

1. Humans will merge with technology, to a rapidly increasing extent.
2. We will develop sentient AI and mind uploading technology.
3. We will spread to the stars and roam the universe.
4. We will develop interoperable synthetic realities (virtual worlds) able to support sentience.

Here, the idea of consciousness separated from any material limitations, evident in Mossbridge and Barušs’ book, takes an absolute form. Consciousness is understood as a technological, if not technical, problem that can be solved in principle, and while we cannot claim to have mind-uploading technologies at the moment, anthropomorphic robots can serve as vivid reminders of such a possibility. They are machines that resemble humans and are able to communicate with them in natural language, thus proving that “consciousness”, or something that looks like “consciousness”, is technologically obtainable.

Finally, the paper was co-authored by Gino Yu, a professor at Hong Kong Polytechnic University, who also worked with Hanson Robotics. Gino Yu founded the Hong Kong Digital Entertainment Association, the annual Consciousness, Science, Technology Conference, and the Society and Asia Consciousness Festival. According to his official webpage, “His main research interests involve the application of media technologies to cultivate creativity and promote enlightened consciousness”.

It is also relevant that Mossbridge, as well as several of her co-authors, were present at the “Science of Consciousness” as members of the Institute of Noetic Sciences (IONS). The project itself was essentially a joint enterprise of IONS and Hanson Robotics. IONS is another non-academic institution that develops unconventional approaches to psychological phenomena. For example, one of the recent projects at IONS, called IONSx, is devoted to the study of “direct mind-matter interactions or genuine telepathic connections with various distant animate or inanimate physical systems”. The results are presented in publications such as “Subtle Energy Perception: Pilot Study with a Buddha Relic” by Fadi Kayale, Dean Radin, Helané Wahbeh, Arnaud Delorme, Loren Carpenter, and Garret Yount. In this study, the authors asked “four self-identified professional clairvoyants and three laboratory staff who did not consider themselves clairvoyant” to “use extrasensory perception to sense a Buddha relic said to emanate ‘subtle energy’” and claim that “the number of times that the container that held the Buddha relic was correctly identified was at or below chance for all participants.

Although Mossbridge is no longer actively involved in IONS, the Loving AI project inherited IONS’s approach to studying “anomalous” phenomena rejected by mainstream science while applying “scientific rigor” to them.

This brings us directly to the project itself.

Loving AI Project

It is officially presented as follows:

Loving AI is a research project addressing how AI agents can communicate unconditional love to humans through conversations that adapt to the unique needs of each user while supporting integrative personal and relational development.

Using OpenCog (one of Ben Goertzel’s projects) and the iConscious coaching model, members of the project team developed specialized software, installed it in Sophia, and conducted the pilot study. The results of this study were presented by Mossbridge at the “Science of Consciousness” conference and can be read in the overview paper “Loving AI: Humanoid Robots as Agents of Human Consciousness Expansion (Summary of Early Research Progress)”.

In this study, Mossbridge and her collaborators asked 10 people to communicate with Sophia, which had Loving AI software installed, in the Hanson Robotics office. Sophia provided the participants with “dialogues and exercises aimed at meditation, visualization, and relaxation” (p. 2). Data collection was organized as follows (p. 7):

After signing consent forms, each of the participants were fitted with a Polar H7 chest strap monitor and recording of heartbeat data continues throughout the experiment. Participants were asked to complete four online questionnaires: a demographics questionnaire, the Fetzer love scale and related self-transcendence questions, the brief mood introspection scale (BMIS), and a resilience questionnaire. Then the participant was asked to interact with Sophia the robot for 10-15 minutes, in a private room in which an unobtrusive, HIPAA-trained videographer [it was Gavin Farrell. — EoR] records the interaction. Then the participant was asked to complete the same self-report questionnaires following the interaction. Finally, the Polar H7 strap was removed and the participant was debriefed.

The video footage of one such interaction between the participant and Sophia is available YouTube:

Based on “1) changes in self-reported experiences of love, mood, and resilience from pre- to post-interaction, 2) heartbeat data (standard Kubios measures) prior to, during, and after interactions with the robot, and 3) affect as judged by independent coders who review videos recorded during interactions with the robot” (p. 7), the authors claim to have found that Sophia, running the Loving AI software, caused participants to experience an “increase in loving feelings overall,” an “increase in unconditional love for robots,” an “increase in pleasant/positive mood,” “no change in aroused mood”, and an “increase in duration between heartbeats (decrease in heart rate)” (pp. 7–8). That is, the study demonstrated that the robot is capable of unconditional love for humans, and that its unconditional love has the beneficial effect on human well-being.

There are other interesting aspects of the study (for example, Sophia’s malfunction in one of the conversations was presented by Mossbridge as a possible reason to say that Sophia is conscious), but let’s get back to the event and Sophia’s presence there. The robot is brought to the conference not only to entertain the participants. It is also presented as a tool for asking and answering some questions about the nature of consciousness. The stakes in posing these questions are not merely philosophical. The whole reason for developing such machines is the promise of their sentience. Although they cannot exhibit any form of “consciousness” that can be unanimously recognized as such, developers are doing two things. First, they create specific situations in which human interactions with the robots provoke sporadic attributions of sentience to them (communication in natural language and anthropomorphic appearance are the most powerful aids for this). And second, they create a two-way tension between the present, tangible robot and the possible future: on the one hand, humans interacting with the robot are invited to imagine what great things robots will be able to do in the future, judging by their present (imperfect) capabilities; on the other hand, humans are suggested to picture the terrible future in which robots try to harm humans, in order to justify the discourse of the need to develop compassionate machines in the present.

As the analysed conference shows, in the case of Sophia — but also in the case of many other anthropomorphic robots — this kind of tension, artfully created by the developers, between the present and the future is placed into two broad frameworks, personified by Goertzel and Mossbridge. The first framework is the science fiction dream of the human mind liberated from the human body by technology. This is a flip side of the idea of artificial general intelligence. The second framework is the “transcendent consciousness” approach close to some spirituality movements. Although both frameworks are related — Goertzel is a proponent of mediation and Mossbridge is developing her own artificial intelligence software — they focus on different things: the first framework emphasizes the technological utopianism of anthropomorphic robots, while the second stresses their technological occultism. Of course, there are many versions of both technological utopianism (Goertzel’s cosmism is just one of them) and technological occultism (Mossbridge’s panpsychism is not the only one), but some form of their combination can often be found at work in anthropomorphic robots.