‘The use of pronouns has entered the discourse of the technology industry in recent times’ | Photo Credit: Getty Images
artificial intelligence (AI) Chatbots are having a significant moment. Large language models (LLMs) are promoting chatbots that interact like human experts, sometimes doing a better job than the best of us at summarizing a complex idea or writing an essay. ChatGPT’s bulleted response reminds us of exam answers from A-grade students.
While internet search requires us to learn the art of keywords, LLM requires us to master the signs. Prompts are archetypal user-generated questions as well as instructions given by software programmers to elicit a desired response from an algorithm. Prompt engineering is becoming an in-demand job to train chatbots to act like skilled humans.
Computer scientist Alan Turing proposed a simulation game to test a machine’s ability to exhibit intelligent behavior that is indistinguishable from a human. In our voluntary suspension of disbelief, shall we forget that we are conversing with a machine? Yes, on some occasions, we will anthropomorphize the model. Other times, we won’t know and the machine will pass the Turing test. Even if we don’t fall into an emotional or financial trap, anthropomorphic chatbots will contaminate our sense of reality.
giving AI a penis
Lawmakers are divided on the question of whether to grant legal personality to AI. It gets more complicated with autonomous machines. But there is a general consensus that misrepresentation of identity by AI sounds like manipulation. Experts suggest that restricting AI from using first person pronouns as well as other human pronouns could reduce cases of AI misidentification. This way, it will be easy to identify text produced entirely by the machine. This is important because pronouns have everything to do with identity today.
Writers like me have a hard time using pronouns for AI in my writing. I use the inanimate pronoun ‘it’ even though ‘it’ is no longer strictly used for inanimate nouns. Fiction writers resort to traditional gender-based pronouns for AI characters who are written as self-aware. Yet, in the real world, AI is not a sentient being. Therefore, AI can easily avoid using ‘I’ in the first person. Even ChatGPT believes that ‘giving AI a distinct identity can help clarify its role and prevent it from being confused with humans.’ We also need to avoid attributing gender-based second and third person pronouns to AI.
Apple has typecast Siri into a female voice. Although Siri has optional masculine and gender-neutral voices, the default is feminine. We give AI a gender to foster an emotional connection with users. In the long run, this will reap the benefits of greater engagement, and therefore, a stronger revenue stream. Historically, the pronoun ‘he’ was commonly used for any student, creating a mental image of a male student. The early version of Siri betrayed our gendered assumptions about a preference for the more submissive female assistant in its default mode.
Still, technology companies are careful about pronouns. Google’s Smart Compose technology, which autocompletes sentences on Gmail, is careful not to predict pronouns, to avoid exposing unconscious gender biases to AI models. Google fired an employee who famously claimed that its AI models had become sentient and preferred pronouns.
AI does not require gender. Some argue that AI should use gender-neutral pronouns such as ‘it’ or ‘they’. It will impersonate AI. Nevertheless, a non-binary identity can be considered non-inclusive. Giving AI a contemporary pronoun will increase popular demand for more diversity in chatbots, as was the case with avatars and emoji. Furthermore, AI requires pronouns to establish an identity that is distinct from humans. Presumably, for ethical and security reasons, it is our right to know that we are actually interacting with a bot.
In the English language, pronouns have evolved based on changes in cultural norms. ‘You’ began to be used for singular and plural second person pronouns from the early modern period, when social interactions became less formal. In literature, the first known use of ‘they’ as a singular gender-neutral pronoun was in 14th-century French poetry.
Editorial | The Good and the Bad: On India and Artificial Intelligence
Historically, we were not completely satisfied with the pronouns used at that time. There have been brief attempts in the 1880s to invent and use gender-neutral third person singular pronouns such as ‘theon’ – meaning ‘that one’. Kelly Ann Sippel’s thesis in 1991 contains a long list of gender-neutral pronouns in the third person singular that had been proposed over the past 150 years. These include hes, hiser, hem, ons, e, heer, he’er, hesh, se, hesh, herim, ko, te, per, na, en, herm, em, hir, and she. This is not even a complete list.
use in technology industry
The use of the pronoun has entered the discourse of the technology industry in recent times, especially since models have become sophisticated enough to pass the Turing test. Suggestions include using first-person neopronouns such as ‘xe’ or ‘ze’. But they are already used by people who identify as gender neutral. Therefore, we must invent an entirely different set of pronouns for AI.
Explained | Are safeguards needed to make AI systems secure?
Regulators should not miss this opportunity to fix this problem at this early stage, before traditional pronouns for AI become prevalent. In order to have a structured approach, regulators need to work with lexicographers and linguists to set a standard for the major languages. Pronoun related mandates can be added to the style guide for faster engineering of AI models. Arguably, even if we lose the emotional connection with AI, we will build a trusted and transparent online environment.
Shalini Verma is an entrepreneur and author