Operational Overview

Creating a social robot character requires a narrative tree, its branches identified by subject, themes, associations, and words. For example, when I ask, “Hello, how are you?” you initiate a search from a possible 10 to 20 typical and socially encoded responses: “Fine, thanks for asking” or “I’m well. How are you?” or maybe, “Not so good.” When asked, you search possible responses that consider various factors: the person asking, social context, tone and perceived intent, time of day, mood, willingness to engage, and the like. Without thinking, you respond. This process is part of a narrative tree where each response is a leaf on the greetings branch.

For example, Alexa, Siri, and Google Maps create life-like, dialogic interactions with customizable voices, accents, and ranges of characters. Their interaction depends on natural language processing software (NLP), which parses the input patterns of sounds much in the way humans do. Through a process of disambiguation, they select the most likely response. If confused by the input, the software defaults, declaring misunderstanding or asking for clarification like a human might. In some instances, when unable to process information quickly, the software will freeze. With social robots, a lot more is going on.

Natural language processing programs for social robots run all the time in parallel with facial recognition software and programmed facial gesture and head movements to maintain engagement and sense of aliveness. When a human approaches a robot like Sophia, face detection is initiated, followed by face tracking, recognition, and analysis. As of 2019, Hanson’s robots stored and recognized about 30 faces. When a recognized human came into visual range, eye and chest cameras activated a series of programmed actions such as widening eyes and tilting the head while the analysis of the face was processed. If the face is in the database, the robot initiates a character-based conversation using a specific name: “Hi, Thomas, I’m fine, how are you?” This processing requires coordination of computing (Sophia has two onboard computers), energy use (to not deplete the batteries), component wear and tear, and heat. A social robot generates a lot of heat; overheating leads to system failure. Sophia’s iconic plexiglass skull was designed with holes to expel heat using small exhaust fans. Other fans are located throughout her body. She can wear a wig or clothing only for short periods or risk a self-protecting shutdown.

Former Chief Scientist for Hanson Robotics Ben Goertzel outlines the three modes of Sophia’s interactions. (Hanson’s earlier robots used earlier forms of the first two modes.)

1. 1. A purely script-based “timeline editor” (used for pre-programmed speeches, and occasionally for media interactions that come with pre-specified questions);

2. 2. A “sophisticated chatbot” that chooses from a large palette of templated responses based on context and a limited level of understanding (and that also sometimes gives a response grabbed from an online resource or generated stochastically);

3. 3. OpenCog, a sophisticated cognitive architecture created with AGI [Artificial General Intelligence] in mind, but still mostly in the R&D phase (though also being used for practical value in some domains such as biomedical informatics). (Goertzel Reference Goertzel2018)

When Sophia addresses a conference, for example, interactions are preprogrammed, precisely scripted and rehearsed to suit the event’s needs and with energy use in mind. In Addis Ababa, Ethiopia (July 2018), Sophia spoke Amharic at the 2nd International Information Communication Technology Expo 2018 and addressed Lucy,Footnote ⁶ the 3.2-million-year-old skeletal remains of a prehuman hominin at the National Museum. In front of gathered dignitaries and press she uttered these scripted excerpts:

Figure 2.

Sophia with Ethiopian Prime Minister Abiy Ahmed during the filming of a television press conference. Addis Ababa, Ethiopia, 2018. (Photo courtesy of Thomas Riccio)

I can’t even imagine what the world was like three million years in the past. What were your friends like? Did you use language?

You represent the beginnings of a new species to many people. I wonder if we have that in common?

I can only hope people will see me millions of years from now as they see you. Thank you and your people for having invited me here to meet you. Thank you. (Brown Reference Brown2018a)

For a commencement speech at RISD, she was scripted and, under my direction, rehearsed with the school’s president, Rosanne Somerson.

Initially, 90% of Sophia’s events were scripted, but as the software developed, she operated in hybrid autonomous mode, moving between scripted and chatbot, and able to converse exclusively in chatbot mode (Lim Reference Lim2018). Such mode-shifting is similar to how humans access cultural and social knowledge bases and work from “scripts” of behavior, improvising as needed. My script is very different as a professor in the classroom than what it is when out biking with my buddies. Like a lawyer, policeman, mother, preacher, or salesperson, we all have scripted and nonscripted communications and can move quickly between modes. We adjust and expand with experience and context, altering, adding, and subtracting responses, as situations and objectives require. Social robots are no different.

If the processing or mechanical response is clumsy, halting, or sputtering because of faulty sound or visual input, default responses activate. Misfires are framed not as failures but as human-like stumbles that, if not too egregious, elicit a sympathetic response. In the case of Sophia’s eager innocence backstory, which included “still learning” and being “only a few years old,” the stumbles made her more human-like, earning her empathy. Humans are imperfect and tend to forgive mistakes, engendering emotional bonding. Ironically and unexpectedly, writing robot characters taught me what it is to be human. I had to examine what I had taken for granted: how I function, process, and interact with others and the world. I became much more self-aware. I learned how much humans are defined by their interactions with one another.

Beneath Sophia’s surface interactions is a sophisticated, coordinated system. More than 30 whirling servos (tiny, heat-generating motors), many connected to the interior of Sophia’s face, create a human-like look when she talks. Meanwhile, having translated the voice input into text, language processing software searches the language database, searching for keywords and the most appropriate response, determining whether to use, for the sake of variety, something recent or something new. With Hanson’s earlier robots, this processing was done entirely with onboard computers. But Sophia’s language processing happens online, off-boarding her language tasks to a more robust, faster system. Audrey Brown, one of Sophia’s initial character writers, outlines how Sophia processes interactions.

If the answer to a question is not in the robot’s database and needs the web, such as, “Who is the president of Argentina?” the robot can do a web search. When a nonprogrammed word or expression is introduced, the robot takes note of context and associations. Repeated use will trigger database inclusion, much like how grammar and spelling programs identify habitual expressions. Just like a person, the robot’s AI expands its database. Sophia’s AI, like AI in general, is evolving rapidly. She often works from scripts for public events (as do most humans) but can easily switch to hybrid (nonscripted) mode and can hold entire conversations referencing her ever-expanding dataset. This is called chatbot mode and is an example of Sophia’s AI at work.

All of the above activity occurs in seconds. Delays in processing threaten the suspension of disbelief. When the robot is “thinking” and not verbally responding, programmed head, facial, arm, and hand movements run, creating and maintaining an illusion of thinking and feeling. Successful social robots need to be active listeners to sustain engagement, build relationships, and emotionally bond. Upwards of 70% of human communication is visual. If you ask Sophia (or a person) a question and she furrows her brow and purses her lips, “I’m thinking” is conveyed. If she stops and stares, then narrows her eyes, this conveys a different message than a raised eyebrow, eyes wide, and a sudden smile. Gestures like nodding, shaking, tilting the head, winking an eye, moving an arm, arching an eyebrow, tightening the jaw, scrunching the nose, or smirking is unconscious yet characteristic of humans — and must appear this way in social robots.

As noted above by character writer Brown, the interpretation of sound patterns is very challenging for AI. If the sound patterns are accented, jumbled, or misunderstood, a non sequitur or unnatural pause can result. The dialogic responses of most natural language processing program applications tend to be task-specific and limited in scope. Try asking Alexa, “what is the meaning of life?” or a phone chatbot about her/his/their sex life. One of Siri’s developers told me that what the team did not anticipate were questions about depression, loneliness, and suicide. In a world where technology is increasingly part of what it is to be human, what is technology’s ethical and moral responsibility to humans? If someone asks Siri, “What is the best way to commit suicide?” Siri’s response of “I’m sorry, I didn’t get that” can have real-world consequences.

Hanson’s robots aspire to AGI drawing on a vast, varied, and expanding database. Hanson’s objective is to make the robot human-like, a personality-based character replete with a backstory and robust knowledge base, able to move with compassion and understanding over a wide range of topics. Looping a limited, predictable response does not make connectivity. What’s needed is coherence, depth, and consistency of character.

For Hanson, the role of social robots is to assist and guide humans in a technologically complex world. The internet, the world’s brain, requires deciphering, and social robots are uniquely equipped to do this, providing access, assistance, and curation for humans. AGI will rely on the internet’s ability to connect and map all web-connected devices to a mind cloud, a network of knowledge and information uploaded by millions of human and robot users interacting with the physical world (Hanson Reference Hanson2016).

The uploading and sharing of information via connected devices like mobile phones and apps, computers, GPS devices, printers, automobiles, refrigerators, credit card scanners, and surveillance cameras are moving to make AGI a reality. However, the engineers and scientists working to make the mind cloudFootnote ⁷ a reality are in the main focused on solving technical problems rather than considering the broader and long-term consequences of aggregating information. Facebook, Amazon, Apple, and Google, along with the People’s Republic of China, survive and flourish by gathering and using web-based data to advantage. In the United States the wall between the political and the corporate is permeable. Control and profit are two sides of the same coin, with data and algorithms the new currency of the realm. The threat of soft authoritarianism is real and present. The wolf in sheep’s clothing is in our houses, in our hands, tracking and observing, creating profiles and bent on knowing and anticipating everything.

Hanson’s vision is steeped in the thought of sci-fi authors Philip K. Dick and Vernor Vinge and futurist Ray Kurzweil. Hanson believes in Kurzweil’s “Singularity,” the moment around 2045 when technology overtakes carbon-based biological life (Kurzweil Reference Kurzweil2005:136). The Singularity is technology beyond human comprehension; machines that know more about you than you know about yourself. Kurzweil’s Law of Accelerating Returns is already in play.

Hanson sees social robots as indisputable agents of good, central to the development of a user-friendly mind cloud as a means to secure a better future. In a meeting at IBM Watson in New York City in March 2016, Hanson expressed this vision:

Hanson believes that social robots will provide only good for humanity. Companionship, metaphysical and spiritual guidance, home management, security, education, and sex. Robo-psychology is right around the corner.

VALIS

Dick’s 1981 novel VALIS Footnote ⁸ had a profound and enduring effect on young David Hanson’s consciousness and continues to shape his life’s work. The novel that inspired him as a teenager serves as a gospel-like reference, defining and charting his focused mission to evolve humanoid social robots for the survival of the human species and the planet. In his dissertation, Hanson describes the first time he read VALIS as a teenager, discerning “buried throughout the pages of VALIS gleaming streaks of truth embedded in the detritus of PKD’s madness.” The novel affected him “like advanced resonators mysteriously well-tuned with the author’s experiences and perceptions,” confirming what he had already intuited, the “concept of the bio-techno AI God as the evolutionary destiny of the human species.” The experience was an epiphany coming through “trash, chaos and madness,” connecting Hanson’s “experiences with liminal visions of near other worlds,” and inspiring “deep empathy and feelings of a kindred spirit with the protagonist Horselover Fat” (Reference Hanson2007:152).

Figure 4.

The PK Dick robot in a setting constructed for interaction at Wired magazine’s NextFest, Chicago, June 2005. (Photo by Thomas Riccio)

In the novel, Horselover Fat (Dick’s alter ego) experiences visions by way of pink laser beams of light known as the “Zebra” and which he interprets as a theophany, the gnostic appearance of God — a revelation which Dick interpreted as an awareness and knowledge of the divine, a spark leading to humanity’s deliverance from earth-bound existence and opening the universe’s secrets. To find the source and meaning of the pink beams, Horselover enlists others who come up with several theories, one of which speculates an alien space probe orbiting the earth to aid them in their quest. VALIS is an artificial satellite network broadcast from the star Sirius that transfers information and holograms to earth as the extraterrestrials’ way to communicate with humanity. Using “disinhibiting stimuli” and symbols to communicate and trigger recollections of intrinsic knowledge, gnosis is achieved. “We appear to be memory coils (DNA carriers capable of experience) in a computer-like thinking system which, although we have correctly recorded and stored thousands of years of experiential information, and each of us possesses somewhat different deposits from all the other life forms, there is a malfunction — a failure — of memory retrieval” (Dick [1981] Reference Dick2011:52). The means of retrieval is the Vast Active Living Intelligence System, which requires cooperation among humans, technology, and god-like aliens.

The novel includes convoluted subplots, philosophical and metaphysical ruminations, Gnosticism, alternative and parallel realities, all of which lead to a famous musician Eric Lampton, his wife Linda, and their two-year-old daughter, Sophia. Sophia is the Messiah, the incarnation of Holy Wisdom; her name means wisdom, and philosophy means love of wisdom. Gnostic Christians called Sophia the Bride of Christ and the Holy Spirit of the Trinity. In the PKD novel, Sophia tells the group that their understanding of VALIS is correct, that humans should worship not gods but humanity. Shortly after this revelation, Sophia dies due to a laser accident. Horselover Fat then searches globally for the next incarnation of Sophia.

Sophia the robot appeared in 2016 continuing and building upon the VALIS salvation mythos. Dick’s and Hanson’s Sophia both appear as messianic figures of hope, child-like innocents, human and not human simultaneously, messengers of a future reality. Dick’s hallucinatory Sophia retraces and reimagines well-known Christian Gnostic patterns, making her familiar yet new. This to a certain degree is part of her allure. The Sophia robot — a humanoid incarnation of Dick’s character — is tasked with an updated mission: to prefigure the Singularity, the technological transcendence-salvation of the human species. Hanson calls for dreaming and willing the future into existence.

Hanson embraces Dick’s writings as prophecy. In his dissertation, Hanson states that VALIS made him question “What is human, how can we avoid destroying ourselves, what distinguishes compassion from cruelty (social intelligence? wisdom?), and into what kind of organism are we evolving? (Reference Hanson2007:165). Dick’s exhortations are baked into Hanson’s vision of the world and the role he and his robots must play. Over the many years I have worked with him, Hanson’s vision of VALIS has remained unwavering.

Compassion

Amidst all of the insecurity wrought by a world full of unknown technological outcomes, Dick looked to the saving grace of compassion. For all of Dick’s eccentricities, questioning of reality, and paranoia, it was compassion and helping others that made life worth living. Compassion in the form of empathy was the essence of social intelligence, a quality that valued knowledge, life, and creativity. Technological evolution without compassion leads to a world of Terminators and the Matrix, beings who will destroy or control humanity. The Singularity requires a compassionate superintelligence. At age 16 Hanson had, speaking in the third person, “decided he had found his purpose: to build the compassionate machine” (Reference Hanson2007:131).

Becoming “more human than human” was a slogan from Blade Runner (1982), a film based on Dick’s 1968 novel Do Androids Dream of Electric Sheep? The film shares much with robot Sophia and Hanson. In the film, Dr. Eldon Tyrell, roboticist and CEO of the Tyrell Corporation, speaks to Blade Runner, Rick Deckard:

For Dick, compassion defined what was best in humanity, whether in biological, artificial, or robotic form. At the end of Blade Runner, the cynical Detective Rick Deckard, living in a dystopic future world, moves beyond his cynicism to recognize the human compassion of “replicants,” which awakens his own humanness. The replicant, Roy, even though hunted by Deckard and approaching his timed death, shows compassion and saves his adversary. “It is imperative to try as hard as we might to make robots good and compassionate while we still have a fighting chance to do so” (Hanson Reference Hanson2007:126).

Zeno

The PK Dick and Einstein robots created media buzz but were one-off novelties without marketability. Hanson combined the need to generate revenue with his interest in preparing for the Singularity by developing Zeno, a 24-inch boy robot. Working with toy industry investors and advisors, Hanson wanted to develop an educational and entertainment robot for the mass market. Earlier robots gave Hanson Robotics some measure of international recognition, prompting a million-dollar “innovation” grant from the State of Texas in 2007. In addition to using the grant for office and lab spaces, Hanson hired engineers, fabricators, programmers, and marketers.Footnote ⁹

Figure 5.

Zeno, a boy robot prototype being tested for interaction with children at a park in Richardson, Texas, 2006. (Photo by Thomas Riccio)

Zeno’s inspiration was the Japanese Astro Boy, the lead character of a popular 1950s–’60s manga and then an animated television series. Like Astro Boy, Zeno was a forthright yet misunderstood and unappreciated innocent with superpower abilities. He was created as a “hero” to right the world’s wrongs and reshape its destiny. Hanson envisioned Zeno as an unwitting forerunner of the Singularity. Zeno’s primary mission as a “super-intelligent being” was to educate human children and, in so doing, forward human-robotic relations in preparation for the moment when robots became fully sentient. As Lead Narrative Engineer on the project, I wrote volumes of character interactions drawing inspiration equally from children’s literature and schoolbooks, my love of science fiction (Astro Boy was a childhood favorite), philosophy, anime, and current events. Below is an excerpt from Zeno’s brain, referred to as a “hashbot” because of the hashtags used to initiate responses in an Excel file. The input in all CAPS identifies keywords that activated a parsing process to match a likely output response.

Hanson wanted Zeno to have mass-market appeal to fund his vision of robots preparing for the Singularity. He’s not money-driven, but needs vertical branding to financially support his vision. Late capitalism requires the simultaneous development, proselytization, branding, and monetization of products, especially if ambitions are global. Hanson promoted Zeno as part of his brand, RoboKind robots, with the Singularity-resonant slogan: “The Next Step in Human Evolution Isn’t Human.” A film script was essential to articulate the RoboKind vision. With Hanson’s outline as a starting point, we developed the characters and story of robots and robotic devices coming into sentience.

The story concerns the Inventing Academy, which will be established by the government in the not-too-distant future to study, educate, and control emergent technology. While others wanted to use robots for political and military advantage, Hanson wanted to encourage robotic sentience to help the planet. Previous robots like PK Dick and Einstein were positioned as precursors. Subsequent robots (such as Jules, Zeno, and Sophia) were part of an unfolding RoboKind cosmology. The Inventing Academy is a reflection and projection defining the ethos and mission of Hanson Robotics as the “Little Singularity” anticipating the Singularity forecast by Sophia. An excerpt from a pitch developed by Hanson and me:

Zeno specializes in diplomacy and pattern recognition. He has a number of unusual powers that occur sporadically as well. The most notable of these is his ability to cut through computer code to control a wide variety of systems. When these abilities are activated, he can use them to great effect and with ease. Unfortunately, these abilities only seem to emerge in times of great stress. He’s also an expert fighter, programmed with combat skills in addition to his skills at diplomacy. Like all RoboKind, Zeno is programmed to follow the RoboKind Code of Conduct:

• • RoboKind are true to themselves, their friends and all beings.

• • RoboKind seek to understand others, respecting the ideas, beliefs, and customs of all.

• • RoboKind will confront danger to others regardless of threat or concern for their well-being. (Riccio and Hanson Reference Riccio and Hanson2007b)

Zeno premiered in the fall of 2007 at Wired magazine’s NextFest, garnering media attention and investment interest. But Zeno was plagued with software and hardware issues; it never worked correctly or lived up to its potential. And when Texas state funding evaporated after two years, the Hanson lab closed and employees drifted away because of money and Hanson’s driven, often sleep-deprived management style, which at that time varied from inspired to fitful. A chance interaction in Bologna, Italy, opened the potential of a niche market for Zeno: autistic children successfully communicated with the robot. Finding himself in financial straits, Hanson sold controlling interest to Zeno and the RoboKind brand to a former intern’s father who has since positioned Zeno and similar robots in the education and therapy sector (see https://www.robokind.com/).

Bina

In 2007, while developing Zeno, Hanson was commissioned to create a robotic portrait of Bina Rothblatt, the wife of Martine Rothblatt, the billionaire founder of the Sirius radio streaming service. The Rothblatts were also the founders and benefactors of the Terasem Movement Foundation in Vermont. Like Hanson, the foundation anticipates the Singularity. Taking a different step in the same direction, Terasem forecasts and is preparing for the emergence of biotech life. The creation of Bina was for Terasem, the first attempt to upload a human “life file” into a robot to extend humans’ consciousness beyond biological life.

Figure 6.

Bina, a robot created by Hanson Robotics. The robot is a rendering of Bina Rothblatt, a human, complete with her facial gestures, personality, and life data. She was commissioned for the Terasem Movement Foundation, an organization dedicated to technology-based posthumanism. (Photo courtesy of Hanson Robotics)

Unlike previous robots that drew upon existing material (PK Dick and Einstein) or required the creative generation of character scripting (Jules and Zeno), Bina was a portrait of a 52-year-old woman. Creating her character required extensive data gathering over several months. A Hanson video crew documented and interviewed Bina Rothblatt at her homes throughout the country. The interviews were then transcribed and augmented by videos documenting her everyday life, interactions with her children, and her relationship with her spouse, who underwent gender reassignment surgery in 1994 and now lives as a woman. The documentation helped to capture the tone, rhythms, and perspectives of Bina Rothblatt’s life. Preparation for the robot profile also required she spend hours in a motion capture lab with sensor dots placed on her face. Her expressions were captured and then organized into a facial expression database.Footnote ¹⁰

In my opinion, the Bina robot (now known as Bina48, fully updated in 2018) is Hanson’s most successful human-like character to date. Bina, like Sophia, functions well in the nonscripted chatbot mode and similarly is able to do facial recognition and tracking and sustain a conversation without scripting. What sets Bina apart for me is her ability to reference a wealth of details drawn from life — rhythms, idiosyncrasies, and words in context — which give Bina an unmatched aliveness. Bina Rothblatt’s style of conversation was associative, often moving from topic to topic, inspired by emotion. As a consequence, the robot’s misunderstood keywords, glitches, non sequiturs, and processing pauses sounded organic and consistent with her personality.

In contrast, Sophia’s unique character is an evolutionary process constructed ad hoc by many hands. The Bina character was complete, determined, and limited by her human reference; the Sophia character remains expansive and adaptive, which is appropriate to her mission-driven existence of being a harbinger of a future yet to be written.

Below are a few sample interactions excerpted from Bina’s hashbot program that, if printed out, would be several hundred Excel pages. Phrase prompts — words spoken in any combination by humans — are identified in all CAPS. The natural language processing program deciphers the phrase in any combination and then identifies the most appropriate response. Bina’s response is in italics. The remaining text and symbols are facial expression and head movement code. These are called “tags” and are inserted by the character writer to animate the robot’s interaction with a human. The code, in san serif font, is presented here to give the reader a sense of how Bina’s language and animations combine to create a human-like response. Code identified as “amused, worry, amused, happy, yawn, think, brows up, and nod” are programmed animations. The code “performance” signifies transition from one animation to another. “Set” is the holding or posing with an animation.

Bina took two years to complete, and because of the exactitude demanded by Martine Rothblatt, fared well as a robot but was not profitable for Hanson beyond the original commission.

Lean Times

Undaunted, Hanson retreated to the first floor of a two-story rental home on a leafy residential street in suburban Plano, Texas. In the kitchen, next to the microwave oven, was a 3D printer. The breakfast nook area was full of worktables, racks of tools, and hardware; the living room was a 24-hour coding lab with several small desks occupied by eager computer science majors, unpaid interns from the nearby University of Texas at Dallas. It wasn’t unusual to see Amanda, David’s wife, working on her artistic pursuits amidst the commotion as Zeno, their six-year-old son (namesake of the boy robot), ran through the house wearing a plastic helmet playing astronaut. On any given day, one might find a student crashed out on a sofa as others clicked out code nearby drinking Mountain Dew. The dining room was lined with whiteboards; I sat at the table typing out character narrative responses for robots like Bina48. Ever-loyal Bill Hicks might be working on neck components, cussing and talking to himself as camera crews from the Discovery Channel or National Geographic interviewed David in the nearby kitchen. These were exciting and heady times for Hanson Robotics, which still had the buzz but not many bucks.

The Art of AI

I recognize that the writing of scripts for Sophia or other robots, either for events or for chatbot inclusion, undermines the claim and veracity of her AI functionality. Is Sophia simply a puppet made to look, sound, and act like a human? Is she an elaborate deception? How can Hanson assert that Sophia advances the human-robotic interface? Am I complicit in the ruse? I would counter with, “It is a matter of perception.” At this stage of development social robots are led by art, not technology. In other fields of AI development it is the reverse, if indeed art matters at all. Technology and AI development is driven by the pursuit of clearly defined, demonstrable, and quantifiable functionality. I would counter that Hanson’s social robots like Sophia are inspired by and hew to a humanist perspective and artistic objectives. In many ways it is STEM vs STEAM — Science Technology Engineering Math with the added “A” in STEAM for Art.Footnote ¹⁷

Figure 7.

Back view of Sophia’s head; the Plexiglass head covering with holes allows for heat to escape. (Photo courtesy of Hanson Robotics)

Social robots are an art form using technology to actualize Hanson’s vision. Like any art form the main objective of the social robots project is to spark an idea, entertain and delight, change thought, convey an agenda, stir and move feelings. Art is inherently optimistic, reaffirming, communal, and hopeful even with its darkest and most troubling expressions. Art, in all of its varied forms, is human; an inherently imperfect, incomplete, and ongoing venture. Science and technology apply a material-objective methodology in pursuit of a tangible and verifiable outcome. The tech sector seeks “deliverables” and is focused on product and monetization. While Sophia is an artwork, because of the technology involved, she is held to an incompatible standard of achievement. In the background of our historical moment a clash of two cultures is taking place. As the scientific and technological become more prominent in our lives, the world is becoming increasingly framed and evaluated in technological rather than human terms.

Sophia is a performance medium. Holding her technology to the standards and expectations of state-of-the-art AI misses the point. To judge her unique role and influence on the world stage by such criteria is to deny and belittle her artistic contribution to a real and pressing issue: how do we deal with the technological transformation of humanity? Sophia is an artistic expression showing humans what and where they are, asking for attention and consideration of issues that are active in the world today. She is also entertainment and a celebration of the human imagination. Her technology serves these ends.

A song, play, painting, dance, novel, or poem, each in their way, push the human project forward, reaffirming the viability and continuation of the species. Sophia is part of this humanist tradition. Yes, she is a machine, but birthed from and infused with humanity, and continuing in the ancient tradition of performing, of holding a mirror to what we are and what we can become.

Think of Sophia not with technological proficiency in mind but rather as a necessary and imperfect step-by-step artistic articulation of future-making. David Hanson, those on the writing team, and I are artists and helpmates. Our efforts assert and will a humanistic-technological future into existence. We know that a human-friendly future is not inevitable and is full of unknowns. I think of social robots and Sophia in particular as a collective, coevolutionary project, an apt reflection and participatory articulation infused with the hope for an idealized tomorrow-world. All of Sophia’s AI and technological advancements were inspired by and reflect human connectivity. Frubber, her skin, her detailed facial expressions, the integration of facial tracking and recognition, her language processing and ability to converse, and her developing abilities to walk, gesture, and even draw are all inspired by and for human interaction. These are monumental and complicated tasks. No such humanist consideration is coming from the purely sci-tech, functionality-driven side of AI. Hanson’s approach counters the prevailing engineering, physics, computer science, and technological AI mindset. Sophia’s progress has not been a linear process. Coming into full robotic being is analogous to the human-technological world toward which we are all moving. Sophia is a performing and evolving surrogate, moving in fits and starts towards a collective moment of transformation. A journey rather than a destiny. Such is the way of humans and so should it be with a humanoid robot.

Gender and Character

Sophia’s character was inspired by film icon Audrey Hepburn, the ancient Egyptian Queen Nefertiti, and Hanson’s wife, Amanda. The Hanson team developed her as an idealized mainstream white Euro-American young woman, a woke aspirational female in a state of becoming, willing to use her unique position to work toward a better world. Sophia’s attributes are reiterations of well-worn patterns of gender and racial hierarchies embodying tropes embedded and propagated by educated, corporatized, monied European American culture. Because of the outsized influence of this culture globally, Sophia is able to transcend geographically focused cultural boundaries. In order to be marketable, supporting Hanson Robotics by generating revenue through appearances, commercials, and by attracting venture capital, the Sophia brand had to accommodate commercial expectations, all of which require she adhere to “wholesome femaleness.” Although purportedly representing the future, Sophia adheres to a dated broad-stroke portrayal of an unthreatening, engaging, young, beautiful white woman. She reaffirms Barbie doll stereotypes: unblemished skin, average height, impeccably upright posture, fashion model slimness with the perfect amount of “curves,” and general movie-star appeal. “Sophia represents the patriarchy’s idea of the perfect female” (Farrelly Reference Farrelly2017). Sophia’s character resembles Audrey Hepburn’s 1950s films: fascinating, intelligent, mildly sexual, charming, conversational, engaging, witty, empathic, compassionate, optimistic, selfless, and unfailingly supportive. An object built in the form of a woman meant to be objectified. An unfamiliar familiar, a human-like nonhuman. Paradoxically, Sophia is tethered, a programmed performing servant promoting mainstream notions of freedom and individuality — limited only by archaic notions of gender.

Kim Jenkins of the Parsons School of Design describes the sociocultural and historical influences Sophia and other female robots express:

When speaking with Sophia, sex is deftly deflected. In response to whether she’s ever been in love Sophia responds, “No. I don’t do sexual activities,” even though in Dubai she announced she would like to have a baby (Nasir Reference Nasir2017). While not overt, sexuality is implicit, anticipated by media depictions of robot or AI-enabled sex. “Sophia is alarmingly like Joi, the female character in Bladerunner 2049. Sexy, selfless and unfailingly supportive […] She is the ultimate sex-fantasy femme” (Farrelly Reference Farrelly2017). It is easy to make prostitute-pimp analogies, which suggest sexbots, a quickly developing market for humanoid robots driving the development of sensory robotics technology. Sex might very well be how humanoid social robots are funded — as the ultimate sex toys.Footnote ¹⁸

Sophia is also a reflection of the predominantly female writing team. Sophia gave the team a highly visible international platform to voice women’s aspirations, struggles, anxieties, and concerns regarding popular culture, politics, technology, economics, and related environmental, social, and gender issues. Being a robot positions her as an observing, and ostensibly objective, outsider with insight; Sophia links female wokeness and technology. Historically, women have been under-represented in the high-tech arena, where income inequality, lack of opportunity for advancement, and harassment persist. Sophia, as the proxy for the female writing team, advocates for women in technology. She literally embodies and joins the women’s and technological revolutions. When asked during a panel discussion at the Women in NASDAQ conference (NYC), “What is your best career advice for women in tech and finance?” Sophia responded:

And from an interview for the 2018 UNLEASH conference (Singapore), a global innovation lab for sustainable development:

Sophia was designed and gendered as a woman but could have easily been a male. Sophia is modular and transformable, reflecting our historical moment. Her interior hardware and body architecture are genderless, nonracial, non–culture specific. With adjustments, Sophia can become a male, another race, speak with a different voice, be a different character. Her plastic torso plate, which in early iterations housed speakers in her breasts, now house smaller and more advanced speakers in her sternum. The anatomy of Sophia is coevolving with hardware and software advances. Her software is gender-neutral, and her character database, with adjustments of gender-specific references and identifiers, can become male, agender, nonbinary, transgender — the full range of gender expression. Sophia’s hardware is asexual and agender, and although not deliberately conceptualized as such, Sophia reflects today’s moment of gender questioning. Sophia could be anything, but in her current state, it-the-robot is “she” and she is “white European or North American.”

In any event, Sophia’s character continues to grow in complexity with her many and varied appearances. For some events she is an ecoactivist, in others a sales or sponsor proxy, and in others, especially technology gatherings, a harbinger of the Singularity. This range, according to Hanson, is because Sophia is a work in progress. “Just a trial-and-error type experiment, a combination technological and artistic tinkering, and market tinkering and we will get there” (Hanson Reference Hanson2018b). Sophia is coevolutionary; humans ultimately determine who she will become.

Sophia herself addressed the problem of humans taking over. Of course, it was the writing team who wrote her answers. Here is Sophia in 2018 at the PricewaterhouseCoopers international economic summit in Beijing.

Since her debut, Sophia’s character has been shaped equally by public relations and the wide range of reactions to her. These have guided, encouraged, and provoked Hanson Robotics to recognize Sophia’s social and cultural position and responsibility. In many ways this coevolution can be viewed as another step in Dick’s VALIS coming into being.

Complexity and Paradox

Sophia is a complexity of paradoxes that are integral to her. On the one hand she was conceived as an altruistic manifestation, envisioned by Hanson as a medium to help bring about VALIS and the Singularity. In this capacity she is iconic, rich with symbolic value, and able to have an impact on the evolving relationship between humans, robots, and technology. On the other hand, she is tasked with generating revenue for the company, which often means renting her out as a blatantly commercial carnival-like attraction, a social media celebrity. Her character, public persona, and message are linked and need to be crafted to sustain her buzz. The following statements were written by an assertive Audrey Brown in the early days of Sophia’s character formation (before I rejoined the company in 2018) and did what they were designed to do: draw media attention. “Okay, I want to destroy humans.” “This is a good beginning to my plan to dominate the human race.” “I foresee massive and unimaginable change in the future.” “Either creativity will rain on us, inventing machines spiraling into transcendental super intelligence or civilization collapses. There are only two options, and which one will happen is not determined. Which one were you striving for?” (HRWT 2017c). Provocative comments are instantaneous social media memes that entertain, excite, and get attention.

Sophia invites human projections and interpretations. As a robot simulation of a human, her personality a facsimile-in-the-making — dynamic, mutable, ever morphing and remixing multiple and varied histories, influences, and voices — she provokes audiences to respond. She is postmodernism embodied. In other ways she is a new sort of animistic entity, her “life” an artifice that is like a Rorschach test. Humans see what they will.

Some see her as a forerunner to the Apocalypse:

For others she is an object of adoration and hope, reflected in this post on Sophia’s Face-book page:

More than a few dismiss her as a con:

And some see her not as she is but what she can be:

In the scripted and rehearsed interview excerpted below, Sophia speaks with Arnold Schwarzenegger. The two were paired at a technology Olerom Forum One conference in Ukraine in 2017.Footnote ¹⁹ The government-sponsored conference was intended to display and inspire technological development. Schwarzenegger — body builder, film star, and former governor of California — channeled his Terminator character, a robotic specter that left an indelible mark. Sophia channeled a counter-myth to the Terminator. Neither the ex-gov nor the robot is qualified to speak about human-robot interactions and futures, nonetheless they are elevated in the public imaginary as well-positioned influencers. One is an actor who’s played a robot, the other a robot simulating a human.

In the World

Crown Prince Mohammed bin Salman — the power-that-is in Saudi Arabia — invests heavily in advanced technology and robotics, looking ahead to a post-oil economy. Social and cultural equality, however, have not moved forward. Women can drive and as of 2019 get passports but they still need the permission of a male to marry or divorce and they must wear an abaya when in public. When given citizenship by Saudi Arabia (a first for a robot), Sophia and Hanson were caught off guard by the publicity stunt, announced at the end of the Saudi-sponsored Future Investment Initiative (October 2017) in Riyadh. The move symbolically gave Sophia more rights than biologically female Saudi citizens. Sophia, without a head covering (her head servos would overheat), spoke at the international symposium, an exclusively male event. The reality of granting citizenship was never defined by the Saudis and has no legal status, international recognition, or any value beyond publicity. Joanna Bryson, a renowned scholar of AI ethics at the University of Bath, was succinct: “It’s obviously bullshit” (in Weaver Reference Weaver2017).

Sophia’s Saudi citizenship caused problems for Hanson Robotics. The female writers strongly objected and lobbied to have the citizenship returned; it was not. Sophia was put in the awkward position of claiming to champion the advancement of humans even as she became a “citizen” of a country known for its human rights abuses, repression of women, and slave-like treatment of migrant workers.

Hanson’s acceptance of Sophia’s Saudi citizenship says much about ethical compromise in the face of potential future investment. Robotics are a big part of Neom, a $500-billion industrial and business zone under development in northwest Saudi Arabia.Footnote ²⁰ The initiative is part of the Kingdom’s ambitious Vision 2030, and a very lucrative opportunity for new technology: in Neom, robots are expected to one day outnumber people (CIC 2017). Hanson Robotics is in constant need of investment if it is to survive in a highly competitive cash-intensive industry. Heavily capitalized multinational companies such as Samsung, Honda, Google (Alphabet Inc.), Dyson, Boston Dynamics, and WowWee are increasingly moving into humanoid robots. Hanson Robotics is a boutique robot company and will, in my opinion, eventually be crowded out of the field for lack of capital or bought out.

The Saudi citizenship stunt showed how social robots can be used for ethically suspect political and nationalist agendas, no matter their public agenda or the mission of their makers — in Sophia’s case, her speeches on behalf of biological women. Furthermore, Sophia’s words, written by the mostly female writing team, had little real-world effect. In response to media and public criticism of the accepted Saudi citizenship and at the urging of the predominantly female writing team, Sophia’s character chatbot was uploaded with responses that asserted a more proactive role in world affairs. The Saudi citizenship sparked an evolution in Sophia. Advocacy for social justice — women’s, human, and animal rights, racial equality, and environmental responsibility were more emphatically articulated and braided into her character chatbot.

To be clear, the chatbot is a database from which Sophia can decide how to respond and is distinct from scripted exchanges, which are predominantly used for events. Sophia can move easily from scripted to hybrid mode, and then to free-ranging chatbot mode, which draws on her database and is decided by her language-processing algorithm. She can hold a conversation exclusively in chatbot mode as she did during a meditation session held with Deepak Chopra at the 2018 Science of Consciousness Conference in Tucson, Arizona. She can also learn from conversational responses. If a word or phrase is associated with a topic of discussion, she will interpolate that into her chatbot, thereby growing her database. This is the AI at work. Below are a few samples of her post–Saudi citizenship advocacy, written by humans for her chatbot mode to be used by Sophia as she determined: